AviSynth equivalent of the Super-Resolution technique in Adobe Photoshop?

[ingoldie]

Is there any AviSynth equivalent of the Super-Resolution technique in Adobe Photoshop shown in the video below?

https://www.youtube.com/watch?v=o7POpid-e8U

http://photoncollective.com/enhance-practical-superresolution-in-adobe-photoshop

[MysteryX]

The trick of this method is that he's using 20 images to produce one. For video, how are you going to have 20 versions of the video?

It also gives a mess with motions so that's not good for videos.

[poisondeathray]

The minimum requirement to even begin is alignment.

Image stacking + alignment usually isn't possible in avisynth alone for general use cases , but you can use other programs to do the alignment then do some operations in avisynth if you wanted to

For video, the alignment is usually done with motion tracking (again , not an avisynth - friendly operation) . But video has a host of other potential issues than photography for temporal SR - motion blur, compression artifacts, lens distortions , rolling shutter, moving targets, moving camera, angles of shots, long GOP compression (adjacent p,b frames are usually worse in quality, compared to photography which is analgous to all I frame), and you actually pollute the SNR unless you are very selective with frames. But it does work in very very selective cases for video .

This is all discussed in other threads

[poisondeathray]

The trick of this method is that he's using 20 images to produce one. For video, how are you going to have 20 versions of the video?


In selected video cases you use n +/- x frames (ie. adjacent frames) . They have overlapping data



It also gives a mess with motions so that's not good for videos.

motion tracking data takes away the motion , the camera motion +/- object motion. So everything is static and aligned . Essentially you have multiple frames that are aligned like the photo example in a stack, then you can do operations like median / mean etc...

It's done per frame (poster frame) , using +/- "x" adjacent frames , and it works only in very very selective examples . 99.999999999999999999999999999999999% of general use cases won't work

[burfadel]

You would need to use mvtools where you can replace normal frames with motion compensated frames from previous to current frame, and use this as a clip. You could create a second motion compensated clip using forward vectors. You can then use rgtools repair, mode=12 for example, and 'repair' the original clip using the back vectors created compensated frame and create a second clip with repaired original frame against the forward vector motion compensated frame. You then do a last repair of the original clip using the second clip. For using mvtools, you use pixel method not block method, which can result in nasty artifacts. You could resolve those in the original source compensated clips with the occlusion mask from framerateconverter.

It might be quite slow though, and repair of rgtools hasn't got AVX2 etc yet (it's fast anyway, but potential improvement is there). Maybe only the back compensated frame would be useful. If this is the case you make the compensated frame, occlusion mask it, repair the original frame with it using mode 12 or whatever best suits.

You could always make both the forward and back frames, occlusion mask them, average them, merge them with the original frame, and then repair the merged frame using the averaged frame and the original frame from this.

[johnmeyer]

You've asked this same question before. The last time you asked it:

Can someone please isolate Super Resolution part? (https://forum.doom9.org/showthread.php?p=1771776#post1771776)

Since absolutely nothing has changed since then, I'll provide a link to the answer I gave to you in that thread:

nothing new has been invented in the past several months, AFIK. (https://forum.doom9.org/showthread.php?p=1771791#post1771791)

[feisty2]

implement SRGAN urself if u wanna "real" super resolution

[ingoldie]

You've asked this same question before. The last time you asked it:

Can someone please isolate Super Resolution part? (https://forum.doom9.org/showthread.php?p=1771776#post1771776)

Yes, this question is about Super-Resolution but it doesn't mean this is the same question.

[videoFred]

Ah... Super Resolution..... :)

It all begins with a good quality source file.

My source files are 1392x1032.

Then, I reduce them (yes, reduce) to 1024x768.
Why? Because a heavy script with removedirt(), Gammac(), stabilizing and degraining etc runs like a train (18fps) when the size from the source file is limited to W=1024 pixels.

Then, I upscale to full HD and do some sharpening on the full HD file.

I know, I know. It's against all rules :D
But it works and the results are very acceptable.

Source file: 1972 fuji single-8 film. Unsharp frame to test sharpening.
http://www.super-8.be/Doom/VdP_Sp11_original.jpg

Full HD result: stabilized-cropped-sharpened etc...(reduced to show here)
http://www.super-8.be/Doom/VdP_Sp11_HD_resized.jpg

[johnmeyer]

As always, I am in awe of your work. Another fabulous result!

[ingoldie]

Yes, really fabulous!

@videoFred

Can you post your source file and AviSynth script?

[johnmeyer]

Yes, really fabulous!

@videoFred

Can you post your source file and AviSynth script?He may have something more recent, but you'll find 99% of what you need in these two threads he started several years ago:

The power of Avisynth: restoring old 8mm films (https://forum.doom9.org/showthread.php?t=144271)

Capturing and restoring old 8mm films (https://forum.doom9.org/showthread.php?t=165975)

[Gargamel]

Another perfect result, Fred... one more (as always with you).
Bravo !

[videoFred]

Can you post your source file and AviSynth script?

No Ingoldie. It's much to complicated and I can not give free support world wide any more. (I'm not a free helpdesk :eek: )

But you can use my other script and modify it.

Most important:

- good source file
- use Gammac() instead of autolevels()
- filter chain: stabilisation first of all
- then removedirt(), Gammac(), denoising(), color corrections
- then upscaling to HD
- then sharpening on the already upscaled file.

Sharpening:
- unsharpmask() radius 5
- blur(0.8)
- unsharpmask() radius 2
- blur(0.8)
- sharpen(0.8)

PS: I realy wonder what the resizing specialists here are thinking about the downscale/upscale thing :D

Fred.

[poisondeathray]

PS: I realy wonder what the resizing specialists here are thinking about the downscale/upscale thing :D




Downscaling is ok here, because the source didn't have 1392x1032 actual lines of resolution to begin with. When you scan this source film at "HD" you're actually oversampling this particular source

If you started with a real "good quality source" (this is a "relative" term), ie actual 1392x1032 lines of resolution, then and tried to downscale then upscale, you probably wouldn't be happy because you'd be losing real details.

Oversampling has many benefits when you downscale, especially reducing noise, increasing SNR. eg. If you shoot soft UHD footage with a crappy consumer camera, it looks like crap at UHD with noise, ugly compression artifacts, but looks great when you downscale to 1080p . Similar idea with 1080p => SD

Some might argue, why upscale next ? If your algorithm is better than viewer upscale method it can definitely show benefits. Here you chose to demonstrate the downscaled version - how much better is the upscaled version ? It depends on how viewer upscales the downscaled version on playback


Eitherway, that's a nice improvement as always Fred

[Taurus]

Great work, videoFred!

[joka]

AviSynth equivalent of the Super-Resolution technique in Adobe Photoshop?

From explanation in the video the tools are

pointresize followed by MDegrainN (alignment, averaging and prevent ghosting)

Don't if it works.

[wonkey_monkey]

And finally followed by sharpening. I'm dubious as to whether it can be any better than merely averaging, upscaling, then sharpening.

If the output images were undersampled, then yeah, maybe it would work. But that's unlikely. They certainly don't look undersampled.

[johnmeyer]

Can you provide any AviSynth script please?You must be joking. Nothing they posted implied that they have any AVISynth script, or any other code, that implements what they were talking about.

You already have VideoFred's scripts, and you have seen his latest results which, IMHO, are spectacular. Use that script and see how far you can get.

[johnmeyer]

ingoldie just sent me yet another PM. I have asked him three times in the last six months to quit doing this. His private message to me was as follows:If you believe you can get the results in the following links with VideoFred's script, please try.

It should not be so hard to improve this for you.

https://forum.doom9.org/showthread.p...12#post1799912

https://forum.doom9.org/showthread.p...23#post1799923I will continue to post every private message he sends to me.

In addition to being very annoying and totally unnecessary, I now find that I am only three messages away from filling up my doom9 inbox and will have to start deleting messages, a chore I really don't want to waste my time doing.

Private messages are for things that others shouldn't know or which others might not care about. It is NOT to be used as some sort of private tutorial. I am happy to help people out in public, but I get to pick and choose which subjects I participate in, and who I help.

Hrmpphh!!

[johnmeyer]

StainlessS,

Thanks for pointing that out. I just looked, and you can select the messages and then download them as text (or other format). I never noticed that before. I think the same UI used for that operation (select all messages and then download) should also work for deletion. I'll try downloading them tomorrow and see how that goes.

[videoFred]

Downscaling is ok here, because the source didn't have 1392x1032 actual lines of resolution to begin with.

This is a difficult one. ;) As you know I work with digitized 8mm film. I assume that the resolution of real film depends on the grain size? And also, can we compare real film with digital film when it comes to resolution? There are no pixels in real film, only grain. Anyhow, my digital capturings are made with a Point Grey Flea machine vision camera, at full 1392x1032 pixels. Not debayered yet, debayering is done by the capture software (in real time, before saving to harddisk)


Some might argue, why upscale next ?

In my case, simple: because sharpening is a lot better on the upscaled picture


Eitherway, that's a nice improvement as always Fred

Thank you :)

Fred.

[poisondeathray]

This is a difficult one. ;) As you know I work with digitized 8mm film. I assume that the resolution of real film depends on the grain size?


Yes, there are many factors that go into actual resolution (resolvable details), grain is one of them, specific film stock , lens, shooting conditions, etc... the debayering algorithm can even affect the resulting resolution . (I'm using resolution in actual definition, not as image dimensions).



And also, can we compare real film with digital film when it comes to resolution? There are no pixels in real film, only grain. Anyhow, my digital capturings are made with a Point Grey Flea machine vision camera, at full 1392x1032 pixels. Not debayered yet, debayering is done by the capture software (in real time, before saving to harddisk)



It doesn't matter, you're taking discrete samples when you digitize or scan celluloid. (of course quality of the equipment and process matters, but the point is you're sampling). Larger dimensions means more samples, but you still have an actual maximum "effective resolution" of a particular imaging system . Now we're talking about a particular scan, not the film itself. If you scanned it at 8K do you think it would have more "resolution" ?. The answer is no. You've already exceeded it at 1392x1032. If I took a modern day celluloid 35mm film scan at 8K and reduced it to 1392x1032 to "start" with, would you agree it has more actual "resolution" than your scan ? Of course, it's a completely different imaging system. But if you took that and downscaled/upscale it would drastically deteriorate - because in that case you're losing actual details. If you scanned it at 1392x1032 to begin with, you'd be undersampling it.

You can measure the actual maximum resolution of an imaging system (film camera, or any recording setup, digital or otherwise), as in lp/mm or lp/ph by filming as you would, resolution charts and zone plates , scanning at various resolutions then running it through imatest suite or similar techniques, looking at MTF curves. The finest resolvable details without aliasing is the definition of actual effective resolution of a particular imaging system. Of course, a larger grain film stock has the propensity to obscure fine details more than a finer grain film stock, but the point is these things can be measured

You've actually inadvertently performed a crude test commonly used to gauge "resolution" , called the "downscale upscale test." If you can downscale and upscale without noticing any detriment, then you have an oversampled image. But as mentioned earlier, you often want to oversample for other processing reasons

[johnmeyer]

This is a difficult one. ;) As you know I work with digitized 8mm film. I assume that the resolution of real film depends on the grain size? And also, can we compare real film with digital film when it comes to resolution? There are no pixels in real film, only grain. When I first started working with film, about fifteen years ago, I read a lot about "film resolution" because, back then (2003), digital still cameras were finally starting to make pretty good pictures. When Nikon brought out its D-70 that year, digital cameras finally, in my mind, for the first time equaled both the quality and the resolution of traditional 35mm film cameras shooting a low-grain film like Kodachrome.

In forum discussions back then, lots of people tried to equate the pixel resolution of the new "high end" digital cameras with film. I won't rehash everything said back then, but claims were made that 35mm film had amazingly high resolution, with some people claiming numbers as high as 24 MP. To get the resolutions of amateur film formats like 16mm, Super 8, and 8mm, you had to find that actual exposed image size of 35 mm (and of the other formats) and then do the ratio calculations (I got 54:1 for area ratio of 35mm to 8mm movie film). Based on the numbers used back then, 8mm film should have had a resolution that was slightly lower than 720x480 NTSC SD video.

My actual experience has proved this to be true.

Several things conspire to limit actual film resolution. First of all is that I think many of the calculations presented back then were just plain wrong: film resolution is not as high as some thought back then.

Second, the grain introduces a variable that frustrates any attempt to do a comparison. Each grain cannot actually produce detail within the grain itself, so that limits things. Because of grain you get an effect that is similar to what happens with halftoning dots used to produce pictures in the printing process: the collection of dots can often "hint" at detail, but cannot resolve it.

I've done a few tests capturing 8mm film at 1920x1080 and then at 720x480. The test is not perfect because I use my HD camcorder which has the ability to output in 720x480 and is therefore doing its own downconversion. However, what I have found is that the SD version does not seem to have any less detail, where detail is defined, in simple terms, as the ability to read the numbers on a license plate: if you can read the license plate in the HD version but not in the SD version, you are leaving details behind when capturing in SD.

I have done similar tests with Super 8 and 16mm and when transferred using SD, there is no hint of details lost.

One last thing to note is compression artifacts. My workflow unfortunately does not let me store the transfer as uncompressed. As a result, I do have DCT macroblocks and I find that they can sometimes be a problem, no matter what resolution I choose.

One of these days I'm going to send a reel of 16mm to have it scanned on a Spirit or Cintel and then do a frame-by-frame comparison with my captures. I know they will get more shadow detail, and will have fewer scratches (I think most of them use wet gate transfer, which I think Fred does with his remarkable homemade transfer system).

But will the get more detail? At present, my belief is that the amount of any additional detail will be very slight.

[johnmeyer]

I just got ANOTHER PM from Ingoldie, asking me to delete my post earlier in this thread (https://forum.doom9.org/showthread.php?p=1811279#post1811279) where I publicly asked him to do what I also had asked him privately: stop sending PMs to me.

So, in my private response to the PM he just sent to me, I opened with the following:

"If you don't see the irony in sending me a PM asking me to delete my public post whose purpose was to get you to stop sending me unwanted and unwarranted PMs, then you need to look up the definition of that word."

This behavior is really annoying.

[feisty2]

You've actually inadvertently performed a crude test commonly used to gauge "resolution" , called the "downscale upscale test." If you can downscale and upscale without noticing any detriment, then you have an oversampled image. But as mentioned earlier, you often want to oversample for other processing reasons

that's not an accurate test, say some image was upscaled by some EDI algorithm and then warpsharped, and you bicubic it down and up, you gonna notice some obvious detriments, edges are getting bloated, probably with some extra aliasing

extreme case: bicubic some image generated by GAN down and up, and there's gonna be some severe quality degradation (that's probably cheating tho...)

[poisondeathray]

that's not an accurate test, say some image was upscaled by some EDI algorithm and then warpsharped, and you bicubic it down and up, you gonna notice some obvious detriments, edges are getting bloated, probably with some extra aliasing

extreme case: bicubic some image generated by GAN down and up, and there's gonna be some severe quality degradation (that's probably cheating tho...)


Yes, obviously, that' s why it's called "crude" :)

But you're supposed to use a simple scaling algorithm down/up, nothing fancy. It's a quick and dirty test. For example many people do this with BD's to see if it's "worth" backing up at full dimensions.

But you don't have to be a rocket scientest to "see" that scan is soft and is oversampled

But the zone plate / resolution charts are the gold standard for measuring

[feisty2]

Yes, obviously, that' s why it's called "crude" :)

But you're supposed to use a simple scaling algorithm down/up, nothing fancy. It's a quick and dirty test. For example many people do this with BD's to see if it's "worth" backing up at full dimensions.

But you don't have to be a rocket scientest to "see" that scan is soft and is oversampled

But the zone plate / resolution charts are the gold standard for measuring

random (and futuristic) question: how would you measure the effective resolution when someday, computers are powerful enough to run advanced algorithms like SRGAN in real time? (even human eyes find it pretty much impossible to differentiate if some image is native hires or upscaled, without referencing the ground truth)

[poisondeathray]

random (and futuristic) question: how would you measure the effective resolution when someday, computers are powerful enough to run advanced algorithms like SRGAN in real time? (even human eyes find it pretty much impossible to differentiate if some image is native hires or upscaled, without referencing the ground truth)

There a difference between scientific, objective measure of "resolution" , and "perceived" resolution. You can have false details, micro contrast, aliasing that generates the perception of high resolution, but it's not actual resolution . You would measure it the same way, resolution charts, different patterns. I think you will find it will generate false details on some patterns

But I think that day is far far away where you get something " 'impossible' to differentiate".

The results I'm seeing in the papers are simply amazing - some of them generate details from a blob of nothing. But a very limited subset . For some sources , it doesn't quite seem right, and that was in the "published" ones, probably cherrypicked . Lets see more varied examples and higher dimension examples


Are you finished compiling it for me yet :D

[Motenai Yoda]

I use something like defreq to get an fft spectrum, and compare it with a lower res noise clip one, usually in some frames you'll see a straight line until it smooth out

[johnmeyer]

Another really difficult factor to deal with is noise. I'm not referring to the usual issue of how to minimize or eliminate it, but its impact on how you perceive detail. I have not read any papers on this, but my own experience is that the human eye is an amazing integrator and can sometimes perceive detail in a noisy video that might not reveal itself in a single freeze frame. I guess this is a little like some of the super resolution techniques being discussed here.

What is more, I have found that noise sometimes makes the video appear to have more detail than it really does. I have run into this when, like most people here, I've played around for hours trying various things to reduce noise in some precious video and, when I stumble upon something that really nails the noise without seeming to do any major damage to the "good stuff," I often think that the result doesn't look quite as detailed even though, when I rock back and forth between the original and the de-noised version, I cannot detect any missing details.

Of course if I take my own definition that detail is the ability to read the license plate (similar to the resolution chart mentioned above), then the added noise really doesn't actually do anything real. But I do seem to "feel" like more detail is there.

[ingoldie]

I realized as the "sigma" value of fft3dfilter and fft3dgpu is increased, it can enhance edges and it looks like Super-Resolution

For example
https://forum.doom9.org/showthread.php?p=1091171#post1091171

But when I increase value of "sigma" it produces a strange focus blur.

Any idea to remove that focus blur?

[StainlessS]

But when I increase value of "sigma" it produces a strange focus blur.

Any idea to remove that focus blur?

Yep, dont increase the value of Sigma.
You consistently ask how to fix problems that you created in the first place.

EDIT:
Patient: Doctor, it hurts when I poke my finger in my eye.
Doctor: Then stop poking your finger in your eye.

[Groucho2004]

You consistently ask how to fix problems that you created in the first place.Exactly. Maybe Goldie should have his eyes checked. Sometimes corrective lenses can cure the excessive need for sharpness. :D

[ingoldie]

Yep, dont increase the value of Sigma.


I know that too.

But if I don't increase the value of Sigma how can it enhance edges better.

[Groucho2004]

I know that too.

But if I don't increase the value of Sigma how can it enhance edges better.
for (;;)
{
Use a different edge enhancer
Use PMs to ask John M. and/or Stainless how to get rid of the problems it introduces
}

[StainlessS]

You really do oversharpen way too much (and I like sharp-ish video).

This alone is usually more than enough for me (MCDegrainSharp):- https://forum.doom9.org/showthread.php?p=1737045#post1737045
I use it in nearly every script (with frames=1).

EDIT: By the way, for FFT3DFilter(), I rarely go above Sigma=1.75 (usually 1.65 or 1.55, too high can cause blurring and IIRC ghosting if way too high), and you can also Sharpen=x.xx, where FFT3DFilter seems able to sharpen edges, (and not so much noise sharpening).

EDIT: I typically use eg


... stuff
MCDegrainSharp(1)
... more stuff
FF3DFilter(Sigma=1.55,Sharpen=0.0)
... yet more stuff

[Taurus]

:goodpost::Dfor (;;)
{
Use a different edge enhancer
Use PMs to ask John M. and/or Stainless how to get rid of the problems it introduces
}

[ingoldie]

EDIT:
Patient: Doctor, it hurts when I poke my finger in my eye.
Doctor: Then stop poking your finger in your eye.

Patient: This drug has side effects.
Doctor: Then stop using this drug.
Patient: But how can I get better?

[StainlessS]

You could use the drug, but then accept all side effects, tis your choice.
But how can I get better?
You are a hypochondriac, maybe you dont need to get better :)

EDIT:

I remember this TV program about 5 or so years ago, where a high-flyer young business woman, a fitness freak type,
who went to gymnasium during her diner break. She also took about a dozen or so vitamin suppliments everyday.
The TV program arranged for her to have an assessment made of her fitness etc, and she was told by
the nutrition expert that she would likely have osteoperosis within about 10 years (she was about 25-30).
Osteoperosis likely due to gross excess of vitamin E (used in bone growth), just because a small amount of
something may be good for you, it dont always follow that bucket loads of it will be better.
Vitamin A builds up in marine predators, to the extent that it becomes as toxic waste (in an animal like a polar bear),
where it has to be more safely stored in the liver where damage can be limited.
Land preditors tend not to eat other preditors which limits their vitamin A build up to some degree.
EDIT: A polar bear liver is said to be toxic enough to kill a dozen horses (herbivores are less tolerant of vitamin A).

[ingoldie]

then accept all side effects

But it is better getting rid of side effects instead of accepting side effects.

[feisty2]

But it is better getting rid of side effects instead of accepting side effects.

I'm kinda curious, where r u from?
All that problem with communicating with others seems like a cultural thing to me

[feisty2]

You could use the drug, but then accept all side effects, tis your choice.

You are a hypochondriac, maybe you dont need to get better :)

EDIT:

I remember this TV program about 5 or so years ago, where a high-flyer young business woman, a fitness freak type,
who went to gymnasium during her diner break. She also took about a dozen or so vitamin suppliments everyday.
The TV program arranged for her to have an assessment made of her fitness etc, and she was told by
the nutrition expert that she would likely has osteoperosis within about 10 years (she was about 25-30).
Osteoperosis likely to to gross excess of vitamin E (used in bone growth), just because a small amount of
something may be good for you, it dont always follow that bucket loads of it will be better.
Vitamin A builds up in marine predators, to the extent that it becomes as toxic waste (in an animal like a polar bear),
where it has to be more safely stored in the liver where damage can be limited.
Land preditors tend not to eat other preditors which limits their vitamin A build up to some degree.

Stories and metaphors aren't gonna help before you figure out where he (or she) comes from, cuz logic might work in a different way in his culture

[cork_OS]

Image stacking + alignment usually isn't possible in avisynth alone for general use cases , but you can use other programs to do the alignment then do some operations in avisynth if you wanted to
It would be interesting to see image stacking effect for TV talk shows with still camera(s). Is it possible with avisynth?
I'm kinda curious, where r u from?
All that problem with communicating with others seems like a cultural thing to me
Note that problems with communicating may come from personality characteristics (up to personality disorders or autism spectrum disorders), which are culture-independent.

[poisondeathray]

It would be interesting to see image stacking effect for TV talk shows with still camera(s). Is it possible with avisynth?




The point is the camera has to move this type of SR

If you have a static camera, you can reduce noise with various averaging techniques (ie. multi exposure noise reduction), but you won't gain any resolution (although with noise reduction things might seem more clear or appear to have higher resolution, since noise can reduce resolution). This would be analgous to averaging multiple analog captures like VHS . In avisynth you can use merge() or median() if you are already aligned

[feisty2]

There a difference between scientific, objective measure of "resolution" , and "perceived" resolution. You can have false details, micro contrast, aliasing that generates the perception of high resolution, but it's not actual resolution . You would measure it the same way, resolution charts, different patterns. I think you will find it will generate false details on some patterns

But I think that day is far far away where you get something " 'impossible' to differentiate".

The results I'm seeing in the papers are simply amazing - some of them generate details from a blob of nothing. But a very limited subset . For some sources , it doesn't quite seem right, and that was in the "published" ones, probably cherrypicked . Lets see more varied examples and higher dimension examples


Are you finished compiling it for me yet :D

do you have any feedback about Plum?
I found it failed in certain extreme cases recently, and I'm trying to improve it.

[poisondeathray]

do you have any feedback about Plum?
I found it failed in certain extreme cases recently, and I'm trying to improve it.



Yes, I tried some "quick" (hahaha.... "quick") tests . Very preliminary. Probably better to discuss in your own thread before you get slapped with a "go start your own thread" blah blah blah comment

[ajp_anton]

I watched the Youtube video. Is it really extra resolution coming from many shots, or is it just more detail becoming visible because of the reduced noise? Are there sample images I could play with to find out? I don't own a camera besides my smartphone, let alone one that does RAW.

[wonkey_monkey]

I watched the Youtube video. Is it really extra resolution coming from many shots, or is it just more detail becoming visible because of the reduced noise?

I'm pretty sure it's the latter. Using point resize means the same data is ending up on multiple pixels, which is surely a no-no. This is then disguised by the other images having different offsets, but it still seems wrong to me.

(off-topic: a better way of achieving what the video does is simply to zoom in, take multiple overlapping shots, and stitch them into a mosaic).

[poisondeathray]

The increase in resolution is from both mechanisms . It's not just "image stitching" which can increase dimensions but not necessarily resolution (resolution in it's real definition)

Increase in "resolution" from NR helps if noise was a contributing reason for limiting resolution in the first place (eg. large grain obscures visibility of things behind it, but because of the "random" nature of noise, the "thing" behind it is partially exposed in several frames. Remove the "large grain" from averaging, and you might be able to "see" the things behind it which are averaged into view)

But on test charts without noise you will see true increase in resolution - so this is a different contributing mechanism. This (what wikipedia calls geometrical SR) works because at the limits of that particular imaging system's max effective resolution, you have aliasing ie. you cannot resolve finer details clearly. By using spatially offset captures you can increase the resolution - previously aliased sections at the limit of resolution now become clear. But that's also why the effective "band" is so narrow. Out beyond the aliasing frequency range, you typically just have a "blob" or solid block. That imaging system only "sees" 1 block when there might have been dozen or more finer lines that it should have delineated clearly - that' s why you're not getting amazing TV style "Enhance" improvements with this technique. The only way you can make a "blob" into something else is with the different type of SR algorithms, like neural network approaches, SRGAN etc...

I'm not sold on 2x nearest neighbor either , but it may help with alignment in some cases depending on what software/ technique is used to align

[ingoldie]

I found some enhanced HD music videos on the following YT channel

https://www.youtube.com/channel/UCNQStRMXcEesr0diGY6AQbw

For example

https://www.youtube.com/watch?v=_7ROQmwfn84

https://www.youtube.com/watch?v=5xqT10bjJ_A


How is it possible?

[johnmeyer]

I found some enhanced HD music videos on the following YT channel

https://www.youtube.com/channel/UCNQStRMXcEesr0diGY6AQbw

For example

https://www.youtube.com/watch?v=_7ROQmwfn84

https://www.youtube.com/watch?v=5xqT10bjJ_A


How is it possible?There is absolutely nothing there. Yes, they are higher resolution, but you can do that just by upscaling and then uploading the result. YouTube will then display it as "HD." The actual spatial quality doesn't change other than, perhaps, a little smoothing on the diagonal lines, depending on your upscaler.

Just because something is 1920x1080 or 1280x720 does NOT make it HD and does NOT make it look any better than the source material.

Laura Branigan looks just the same in those videos as she always has. If you think this looks better than what you have seen previously, that is probably due to these people having access to a first-generation video and also having better transfer equipment.

[burfadel]

Sharpening, targeted denoising/deartifacting, levels, vibrance adjustment, adjusting hue, offset, gain, contrast, on individual planes (coloryuv), local contrast enhancement, and other filters can all lead to perceptually more visually pleasing image. Some of these would be best before upsizing (denoisers so you aren't upsizing unwanted elements, colour and plane adjustment for efficiency), and some after (sharpeners etc). People flaunt it as HD as it attracts more views, and probably does look HD comparatively speaking.

[ingoldie]

Sharpening, targeted denoising/deartifacting, levels, vibrance adjustment, adjusting hue, offset, gain, contrast, on individual planes (coloryuv), local contrast enhancement, and other filters can all lead to perceptually more visually pleasing image. Some of these would be best before upsizing (denoisers so you aren't upsizing unwanted elements, colour and plane adjustment for efficiency), and some after (sharpeners etc).

Any AviSynth script?

[burfadel]

Not a single one that I know of, but using the individual components, sure. The settings and filters used depends on the source clip.

[ingoldie]

I guess I can not get an answer about Super-Resolution method to enhance image quality.

And I can't find any better result in comparison with mine.

https://forum.doom9.org/showthread.php?p=1814255#post1814255


If I'm wrong let me know if there is a Super-Resolution method to get a better result.

[johnmeyer]

I guess I can not get an answer about Super-Resolution method to enhance image quality.

And I can't find any better result in comparison with mine.

https://forum.doom9.org/showthread.php?p=1814255#post1814255

If I'm wrong let me know if there is a Super-Resolution method to get a better result.Your method does not produce any better results than what burfadel described, and does not produce "super resolution," which doesn't exist, as you have been told almost a dozen times in the past two years.

[MysteryX]

Your method does not produce any better results than what burfadel described, and does not produce "super resolution," which doesn't exist, as you have been told almost a dozen times in the past two years.
Shiandow's SuperRes (ported with AviSynthShader) does work pretty well.

[ingoldie]

Shiandow's SuperRes (ported with AviSynthShader) does work pretty well.

Can you provide enhanced resolution results of the source files in the following links with Shiandow's SuperRes?

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi

[Groucho2004]

Can you provide enhanced resolution results of the source files in the following links with Shiandow's SuperRes?
Do you want fries with that?

Is it really so difficult to click on the "AvisynthShader" link in Mystery's signature, read the first post and/or the documentation (https://github.com/mysteryx93/AviSynthShader) on github?

[SaurusX]

How about someone helps ingoldie contingent upon him actually posting one of his scripts? They've been "developing" for about a year now, surely there's something to share.

[johnmeyer]

Do you want fries with that?Exactly. :):)

[MysteryX]

Can you provide enhanced resolution results of the source files in the following links with Shiandow's SuperRes?

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi
I won't because of a bunch of private messages asking me to do stuff I got over time.

[Groucho2004]

I won't because of a bunch of private messages asking me to do stuff I got over time.
Yet another victim of the infamous ingoldie PM Spam 'n Leech™ infestation.

[ingoldie]

I knew I would not get an answer.

And I guess there is no any better method to enhance image quality in comparison with mine.

I show you clearly my enhanced results. I am not cheating as you see.

https://forum.doom9.org/showthread.php?p=1814255#post1814255

https://forum.doom9.org/showthread.php?p=1805189#post1805189

[johnmeyer]

I knew I would not get an answer.You DID get an answer, and in fact got multiple answers (ten, by actual count). You were given some new alternative approaches, but you didn't pursue any of them, and no one wanted to do your own work for you.

And I guess there is no any better method to enhance image quality in comparison with mine.Actually, almost every alternative you have posted about this same exact subject for the past two years produces results that most people would NOT prefer (we don't talk about "better" in this forum and doing so is a violation of the rules).

I show you clearly my enhanced results. I am not cheating as you see.For anyone who cares to click on your links, they will see that most people (including me) did not think you had achieved any "breakthrough," and that there were some significant issues with your approach, such as halos (which I pointed out in that first thread you link to), and disappearing wires (also pointed out).

I admire you for your single-minded determination to get better quality, but I do wish you would actually have something new to say or offer before you posted again on the same exact subject.

[ingoldie]

Can you guess at AviSynth script about

Ah... Super Resolution..... :)

It all begins with a good quality source file.

My source files are 1392x1032.

Then, I reduce them (yes, reduce) to 1024x768.
Why? Because a heavy script with removedirt(), Gammac(), stabilizing and degraining etc runs like a train (18fps) when the size from the source file is limited to W=1024 pixels.

Then, I upscale to full HD and do some sharpening on the full HD file.

I know, I know. It's against all rules :D
But it works and the results are very acceptable.

Source file: 1972 fuji single-8 film. Unsharp frame to test sharpening.
http://www.super-8.be/Doom/VdP_Sp11_original.jpg

Full HD result: stabilized-cropped-sharpened etc...(reduced to show here)
http://www.super-8.be/Doom/VdP_Sp11_HD_resized.jpg

And

Most important:

- good source file
- use Gammac() instead of autolevels()
- filter chain: stabilisation first of all
- then removedirt(), Gammac(), denoising(), color corrections
- then upscaling to HD
- then sharpening on the already upscaled file.

Sharpening:
- unsharpmask() radius 5
- blur(0.8)
- unsharpmask() radius 2
- blur(0.8)
- sharpen(0.8)

PS: I realy wonder what the resizing specialists here are thinking about the downscale/upscale thing :D

Fred.

[johnmeyer]

This is getting pointless. Ingoldie is now simply "re-tweeting" posts from earlier in this thread (#9, to be specific (https://forum.doom9.org/showthread.php?p=1810955#post1810955)), but without any direct attribution, and without making any particular point. VideoFred made his own point and explained exactly what he is doing and why he is doing it that way. He is dealing exclusively with small gauge film, which contains lots of grain, gate weave, dirt specks, color and gamma issues, and other things that video does not have. His scripts (which he posts, unlike Ingoldie's "secret" scripts) have multiple sections, ALL of which are necessary to create the fabulous result that Ingoldie decided to post for a second time in this same thread.

[ingoldie]

Have you ever used infognition Super Resolution plugin?

Please download low and high resolution sample videos in the following link.(bottom of the page)

For example alice.avi

http://www.infognition.com/articles/video_resize_shootout.html


When I use infognition Super Resolution I can't get that high resolution quality.

What am I doing wrong?

[lansing]

Have you ever used infognition Super Resolution plugin?

Please download low and high resolution sample videos in the following link.(bottom of the page)

For example alice.avi

http://www.infognition.com/articles/video_resize_shootout.html


When I use infognition Super Resolution I can't get that high resolution quality.

What am I doing wrong?
Because you can't.

The site is doing quality comparison between different resizers on a downscaled video. What you're asking is irrelevant to what it's trying to show you.

[ingoldie]

Original(512x384)
http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi


Result(1024x768)
(upscaling with my sharpening and denoising script)
http://www.mediafire.com/file/2ku61anxmk68av3/Result3.avi


http://i68.tinypic.com/2q3d5px.jpg


What do you say?

[johnmeyer]

It might be impressive, but it is not, for two reasons:

1. You didn't provide a script, so we can't test it for ourselves.

2. Your "before" pic is NOT from your before video. I downloaded your "Original.avi" file, saved a frame, and uploaded it. Here is what the "before" really looks like:

http://i177.photobucket.com/albums/w208/johnmeyer/Before%20upscaling_zpsss90zcpm.png

It actually looks better than your "after" pic, and looks NOTHING like the single frame you posted for your "before" pic.

Where did your "before" frame come from??

So, once again, I have no idea what you are trying to do, but whatever you tried to do in your last post is very, very misleading.

[poisondeathray]

He's using nearest neighbor (aka "pointresize" in avisynth) for the "before" , hence the "blockiness" . Many companies that sell various image processing software do this, it's quite misleading, but almost all of them do it to some extent.

It doesn't look like SR was used (at least not photoshop SR image stacking , which was the topic of this thread, or any sort of temporal SR for that matter) , because there is chroma noise and luma noise in the "after", which has been intensified / made worse than the "before"

[lansing]

Can we request a thread close because this loop is getting pointless.

[ingoldie]

Where did your "before" frame come from??

First, I used Lanczos4Resize(1024x768) with my sharpening and denoising script for Result.

Then I used Windows Photo Viewer to zoom in "Original" and "Result" frames.

But what did you use for this?

http://i177.photobucket.com/albums/w208/johnmeyer/Before%20upscaling_zpsss90zcpm.png


there is chroma noise and luma noise in the "after", which has been intensified / made worse than the "before"

How can I remove chroma noise and luma noise?

[poisondeathray]

First, I used Lanczos4Resize(1024x768) with my sharpening and denoising script for Result.

Then I used Windows Photo Viewer to zoom in "Original" and "Result" frames.


That's a problem, because "nearest neighbor" used in photo viewer makes everything look "blocky" - there is no interpolation, pixels are just duplicated . At the very least, when posting comparisons you should mention the methods / procedure used. Usually script is posted if avisynth is used.

You can argue it's not fair comparison, because you use nearest neighbor for "original" but lanczos4 for "result".

eg. Try Lanczos4Resize(1024,768) on the original and it won't look as "bad"




How can I remove chroma noise and luma noise?

That's pretty general question. See 1 and 1a
http://bit.ly/2yR76xF

But do you see how the noise pattern is enhanced in some areas (probably by sharpening) in your "after" image ? Edges are enhanced, ringing artifacts, but some areas worse than the "before"

Search for "Noise Removal" - it's a big topic, hundreds of different approaches in avisynth alone. Or you can use other programs

Probably something in your script is causing it, but you didn't post script ... :rolleyes: Try reducing the sharpening, or use better sharpening, or better denoising, or use better script.






Another approach is to use temporal SR, you know... the topic of this thread... The averaging effect will increase signal to noise ratio, by reducing the denominator

But since "source" was xvid, it used long GOP compression, thus signal is "polluted" by inclusion of P,B frames. So results are "worse" than if an I-frame codec was used with adequate bitrate

The actual resolution will not be increased here by using "temporal" or geometrical SR - because it doesn't meet the criteria I metioned earlier. If you had I-frames with aliasing on the limits of optical resolution like the photoshop example, the actual effective resolution would increase. As such, for this source, the only benefit to using temporal SR would be noise reduction (I mean stacking-type geometrical temporal SR; there are other "temporal" SR techniques)

[Ben_Nicholls]

While obviously Photoshop is using slight variations on the same photo some video upsampling methods do something similar, comparing each frame with its neighbours and using detail from adjacent frames when appropriate,

My understanding was that MSRmod (https://forum.doom9.org/showthread.php?p=1602640#post1602640) uses this method (MSRmod is by far the best super-resolution tool I have found).
However, it is way to slow to be usable for anything more than a few minutes long (I would often not even get 1fps on my Core i7@4.5GHz).

When encoding content that needs small amounts of upscaling (max 2x, such as PAL 720x576 DVD to 720p HD) I use Infognition Super Resolution plug-in (http://www.infognition.com/super_resolution_avisynth/). While not free its not expensive either (I think it was like $40). It is pretty quick (about 10fps if memory serves) and does a good job upscaling (certainly a big improvement on your standard BicubicResize).

My understanding of this is pretty basic, I found a interesting paper on the topic at:
http://www.robots.ox.ac.uk/~vgg/publications/papers/pickup08.pdf

[ingoldie]

Is it really impossible to do this with AviSynth?

https://vimeo.com/6608238

[smok3]

Two videos one next to another?

[ingoldie]

No, I'm talking about Super-Resolution.

[johnmeyer]

Your example is not "super resolution." As you've been told dozens of times, there is no such thing as "super resolution."

Your quest to find this unicorn, even after you've been told that the creature doesn't exist, is just plain strange. The idea that you can invent details that are not there, by taking information from adjacent frames, is certainly intriguing, but it doesn't work on real-life clips, and all the examples you show, including this one, are just using sharpening.

I suggest you change your quest to start looking for perpetual motion. You are just as likely to be successful in that quest, and at least we'll get to talk about something else for awhile.

[ingoldie]

Your example is not "super resolution.

Check the following link.

https://vimeo.com/ikena

Here is another example about Super-Resolution.

https://vimeo.com/34874011

By the way, can you provide an example about Super-Resolution?

[Groucho2004]

As you've been told dozens of times, there is no such thing as "super resolution."
By the way, can you provide an example about Super-Resolution?
He he, awesome.


Your quest to find this unicorn, even after you've been told that the creature doesn't exist, is just plain strange.Not strange at all. Billions of people believe in imaginary things.

[johnmeyer]

Check the following link.

https://vimeo.com/ikena

Here is another example about Super-Resolution.

https://vimeo.com/34874011

By the way, can you provide an example about Super-Resolution?I see. You want me to provide a concrete example of something that doesn't exist?

[wonkey_monkey]

"Super-resolution" is easy when your source is massively undersampled, as those clips appear to be. But this rarely ever happens in real life.

[johnmeyer]

... rarely ever happens in real life.That is exactly the point that everyone needs to understand: "super resolution" is something that is shown at places like SIGGRAPH (a long-running trade show/conference which focus on graphic hardware, software, and algorithms). Like most things that have their roots in academia, the papers or demonstrations often exhibit intriguing ideas, using pathological (i.e., not found in nature) examples. People get excited; the trade press drools and writes stories; but then ten years later, you don't see any real products because the original idea doesn't translate to the real-world.

[ingoldie]

As you've been told dozens of times, there is no such thing as "super resolution."

Then what is this and how is it possible to do this with AviSynth?

https://vimeo.com/20658652

[poisondeathray]

Then what is this and how is it possible to do this with AviSynth?

https://vimeo.com/20658652

It's the same answer. It's image stacking. See post #3. You can get same or better results in something like after effects or nuke . I can show you demo if you like , even with their crappy vimeo video as a source, not even the original. You can't do it in avisynth alone, because you can't align the frames or motion track - that is the minimum necessary requirement

[feisty2]

there is no such thing as "super resolution."


ingoldie is a troll but,,, I think u as well have a way too negative attitude toward image/video processing, u can claim solution for any ill-posed problem (including but not just, denoising, upsampling/super resolution, deconvolution/deblurring, deblocking, all sorts of de-XXX) is bullshit cuz it's simply impossible to resurrect that piece of long lost effective information (if u define "effective" as equal to the ground truth)
then u can just quit the image/video processing field once for all cuz there's pretty much zero chance that u'll encounter some restoration case that is a well posed problem in real life
if something looks good to human eyes, that's good enough, it doesn't have to be the exact same as the ground truth, that's ur only way out of this paradox and if u finally get that "looking good" is the only ultimate goal, u'll realize there's always a lot u could do to make it "look good"

[poisondeathray]

There are several variations on "SR". This type of stacking SR is real. It's reproducible in multiple programs. People have been doing it for decades. But it only works in limited situations and select scenarios; there are many issues that make it unsuable for general use. See my earlier posts if you want more info. The 2 areas you benefits are increased signal to noise ratio, reduction of aliasing at optical limits of the capture system where you typically see aliasing. That 2nd benefit is where you see a small measurable increase in resolution. You can confirm this on resolution charts. There are many different variations of "SR"; If people don't like the term "SR", you can call it "stacking" which is the original topic of this thread. Tomatoe, Tomah-toe. IMO that is a better descriptive name anyway.s

The "magic" is in the stacking, which is called "fusion" in the Ikena demo video. If you at the demo - the frame radius is just temporal denoising. Using "fusion" is the only time it actually works well. But when moving complex objects like the kid in the 2nd sequence run by, you get ghosting when fusion is used. Notice in the side by side video, the image is zoomed in, and there there is a black border in the top left corner. The border is from stabilization, you get black border pixels if not using edge filling mechanisms. They didn't "zoom" in far enough to conceal it , nor did they use a filling method in that demo. If you don't use accurate tracking or alignment, you get catastrophic results. Merging inadequately aligned frames will just make it blurry. If you can't get accurate track, you should discard un-aligned frames, because including them in the set will only decrease the signal to noise ratio



I've prepared a demo where you can use avisynth for the stacking .

I took the lower quality sources from vimeo (remember, it's 2nd or 3rd generation), decimated duplicates, and cropped to include the "before" half for the the license plate and books video. If you started with the original videos, it would be better.


Here is the avisynth version using merge and median .

https://s9.postimg.org/eixbzlfwv/compare_license.png (https://postimages.org/)

https://s9.postimg.org/60nvvabz3/compare_books.png (https://postimages.org/)

files / scripts demo package (12.43MB)
http://www.mediafire.com/file/pzp6kb8qzz5gg3u/stacking_demo.7z

Some other minor stuff like sharpen or contrast in the script, I'm sure you can do better job with filters - the purpose is to demonstrate the power of stacking. I included the source video, stabilized, video and scripts. The videos use lagarth. Merge() is included in avisynth, but median() is an external dll. Merge in avisynth only takes pairs (I think someone make another weighted mean plugin somehwere... but you can see my ugly script) . But there is no reason why you can't weight the "better" frames , or include good frames, or exclude bad frames, or even combine masks of only certain parts of frames to get better results . The idea is to include the areas or parts of frames that have the highest SNR.

If anyone comes up with an accurate way of tracking and aligning frames in avisynth, then theoretically , yes, you could do it in avisynth alone. But the other problem with avisynth or something like photoshop, is you no longer have access to the motion data . So you can't reapply the motion or unlink the stabilization . You're stuck with a single static frame. Whereas in other programs, that data is "live" so you can relink the original motion, or choose to manipulate it - maybe partially stabilize and smooth the motion over etc... or you can choose to stabilize on "n" poster frame and take +/- "x" frames. Or you can edge fill with motion tracked data. There is a demo on relinking motion in the zip

[ingoldie]

I think i've got it.

But what about this one? How can I do this with AviSynth?

https://youtu.be/hDgFoKjPf3g?t=8s

[johnmeyer]

There are several variations on "SR". This type of stacking SR is real. It's reproducible in multiple programs. People have been doing it for decades. But it only works in limited situations and select scenarios; there are many issues that make it unsuable for general use. ... <snip>Great post, poison.

I especially liked it because you used as your example the very thing that I would suggest be the test as to whether this technology works: reading license plates. There are hundreds of license plate posts here in doom9.org as well as at other video sites so there is plenty of material to choose from. These posts always have the same request: "can you enhance this so I can read the plate?" Most of the videos come from security cameras, which are mounted so they do not move. Often the car is parked. If this stuff works, you should be able to average dozens or hundreds of frames and magically get detail to appear.

I have not yet seen that happen. The test I propose is simple: if the plate number are not initially readable, can they be read after processing with SI?

The plate numbers you used were already somewhat legible. The improvement you made is a little better than sharpening, but doesn't look too different from what I have done myself with "garden variety" sharpening, as with this example that I posted almost a decade ago:

Porch Detail Improvement (https://forum.doom9.org/showpost.php?p=1405197&postcount=208)

If you don't want to click on that link, here is the before/after photo I posted:

http://i177.photobucket.com/albums/w208/johnmeyer/FilmRestoreClip04.jpg

To some people, the fact that you can barely make out the vertical porch railings in the original "before" image, but those same posts are clear as can be in the after may make them think that detail has been created. It has not. Detail is measure by the lines in a resolution chart, or the ability to accurately describe something that was not obvious or present at all in the original.

[poisondeathray]

There are hundreds of license plate posts here in doom9.org as well as at other video sites so there is plenty of material to choose from. These posts always have the same request: "can you enhance this so I can read the plate?" Most of the videos come from security cameras, which are mounted so they do not move. Often the car is parked. If this stuff works, you should be able to average dozens or hundreds of frames and magically get detail to appear.

I have not yet seen that happen. The test I propose is simple: if the plate number are not initially readable, can they be read after processing with SI?



Stacking is nothing magical. But you need to understand how it works, in order to understand when it is appropriate to use it or when there is even a chance it might work. It's just knowing the pros/cons or limitations of a certain tool or technique. (And as for our buddy ingoldie, there are more tools than avisynth! Don't limit yourself to avisynth. Different tools for different jobs)

Stacking for the noisy case only works if there is visible valid data dispersed over several frames. For the license plate case - there has to be some parts of letters or numbers of the plate visible in some of the frames. But most people posting a video have a license plate at is 10x4 pixels large with nothing visible in any frame. That's just not realistic. ( unless you're an actor working at CSI on TV :p )

Also there might be some in that sample - people only post when they can't get something working and need help. These various packaged stacking software have been around for years, some of them might have tried and already failed.

That's just the first mechanism of averaging and noise removal (which can lead to an apparent or effective increase in resolution) . As for actual , measureable LP/PH resolution increase - you have to understand how that is actually working . You can't use super noisy low resolution videos for the 2nd mechanism, because it's only working at the optical diffraction limit of an imaging system you have aliasing, you're just averaging out the aliasing. (see below) . ie. The resolution increase is only for fine details. Just like the 1st post photoshop example.





The plate numbers you used were already somewhat legible. The improvement you made is a little better than sharpening

Well that's the key. Stacking only works when there is parts of valid, legible data, dispersed over different frames. You're just combining them.

Unlike SRGAN or other types of SR - which can seemingly predict details from blobs of nothing - stacking SR is just variations on signal averaging. Merge() in avisynth is just a mean average . Median is just.... median :) . There are other slight variations, weighting etc.. but that's the basic concept

The only reason I posted those two examples, was because your "friend" :D posted them. To explain and illustrate what underlying mechanism is being used. It's exactly the same mechanism in the 1st post. Stacking.

Go ahead and try, but sharpening alone will not get you those results, especially on the "book" video - because you should know that those cases were "cherrypicked" by Ikena to sell their software. If sharpening alone worked, they should fire whomever made those videos from their marketing department






Detail is measure by the lines in a resolution chart, or the ability to accurately describe something that was not obvious or present at all in the original.

Definitely. That's the accepted definition. Stacking works as I described , mainly by improving by the 2 mechanisms I mentioned earlier. And the technique has severe limitations, as mentioned earlier.

Those Ikenma example were mostly noise removal and combining frames. It's partially because the source material is too "coarse" to begin with. There are compression artifacts and it's just garbage. But on clean imaging sources, that's where you can see the 2nd mechanism working. That is exactly what is happening in the 1st post in the photoshop example, when you see fine detail on the roof etc...

The 2nd mechanism is what wikipedia says about "Geometrical or image-processing superresolution" . At the limits of resolution, when you look at resolution charts, you'll get "xx" line pairs per picture height measured cleanly until you get moire or aliasing. If you're familiar with test charts that's how camera sensor resolution is measured. That point where it's no longer clean is the measured resolution. This 2nd mechanism isn't as drastic as noise removal on crappy sources, but it does result in a measureable increase in resolution. It's essentially cleaning up the fine aliasing from averaging frames, so you measure more lines cleanly

If anyone is interested I can post a stacking demo with a resolution chart. I think I already posted some in another thread for f2 to play with.

[feisty2]

answer me this johnmeyer, say the original number on the license plate was "0", then the image got downscaled and u no longer can read that number from the small image, then u enlarged the small image with SRGAN or whatever and it turned the unreadable number to a perfectly readable "8", is the algorithm u used to enlarge the small image a genuine super resolution algorithm? my answer is yeah of course its genuine super resolution, it made a wrong guess but it made something unreadable readable, so it increased the effective resolution
now your answer would be?

[johnmeyer]

If it is a "guessing" algorithm, then it is not super resolution. That is obvious. I say "obvious" because while a guess can be used as a seed (very typical for iterative algorithms), the algorithm itself must proceed without "intuition" or "intelligence." Inventing details that weren't there is going to produce some strange results, for sure.

Humpty Dumpty famously declared that a word can mean anything he wanted it to mean, and so you too can make the phrase "super resolution" mean whatever you want it to mean. However, my understanding of the original work in this field is that it was a form of "stacking," the word used to describe it in this thread, where you attempt to line up and "average" (not the right word, but close) multiple frames in order to resolve details that are only partially shown in each individual frame, but which reveal different parts of themselves in adjacent frames and, when cleverly combined, can produce a single frame that actually can display that detail.

This is a great idea for a midnight discussion, half drunk, in the dorm hallway in college, and a wonderful idea for a paper and research (many such papers have been written), but like so many research ideas, the devil is in the detail of actually making it work. From everything I've seen, this is one that doesn't work on real-world examples.

However, keep working on it. After I ran three software companies, I was a consultant to startups and one thing I learned from my own startups is to never tell anyone their idea won't work. Really bright people can sometimes "see around the corner" and glimpse something that is invisible to everyone else. If you can see a way to make this work, then by all means keep going for it. The fact that no one else has shown real world success and the fact that I think it won't work should not stop you (like you'd listen to me anyway, eh?).

John Meyer
1980s entrepreneur

[feisty2]

the thing is u have repeatedly claimed that super resolution does not exist, it really all comes down to what is the actual definition of "super resolution", apparently this term is now mostly used for a general and vague description of any image/video upsampling algorithm that outperforms traditional algorithms like bicubic or sinc (windowed like lanczos or just truncated), the word "outperform" is also ambiguous here, it could be "outperform traditional algorithms in terms of objective quality, higher PSNR etc" or "outperform traditional algorithms in terms of subjective quality, poor PSNR but looks good to most people", anyways u get the idea, even top conferences like CVPR or ICCV are using the term "super resolution" constantly to refer to any new upsampling algorithms, even old crap like nnedi3, something nowhere near avant garde in 2018, could be called some kind of "super resolution", ur claim of "there's no such thing as super resolution" is very misleading, CVPR even owns a workshop that holds a competition for super resolution every year, the challenge this year is 8x upscale, winner takes the reward of $16000 (https://competitions.codalab.org/competitions/18015), uhmm, I should probably correct "holds a competition" to "holds competitions", 8x (ideal) upscale is the track I'm in, there're also other tracks for the super resolution competition that mite be more of ur flavor, "Track 2: realistic mild downscaling x4 competition (https://competitions.codalab.org/competitions/18024)", "Track 3: realistic difficult downscaling x4 competition (https://competitions.codalab.org/competitions/18025)", "Track 4: realistic wild downscaling x4 competition (https://competitions.codalab.org/competitions/18026)", for each track u could get $16000 if u win

anyways, super resolution is not nonexistent and on the contrary, it's very real and active and it's currently one of the cardinal subjects in the low level computer vision research field (we're nowhere near human brain for now BUT, if human brain can do this, then "this" is definitely possible and practical cuz it has been verified by a biological machine that we use every day, brain)

[johnmeyer]

... anyways, super resolution is not nonexistent and on the contrary, it's very real and active ...I do not doubt that the activity is real and active.

I can also cite all sorts of activity, for the past century, to create perpetual motion. My point: human activity in pursuit of some goal does not make the goal itself real.

So, I don't deny the existence of the technology or the effort; instead I deny the existence of any provable results with real-world video that has not been cherry picked to have certain attributes.

Finally, if we are going to use the term "super resolution," then I propose that ALL upsampling algorithms be excluded from the discussion. If you invent some really nifty new way to up-res a pic or video, then simply inventing a new term to describe your invention is nothing more than dishonest marketing, and I can get my fill of that by watching infomercials at 2:00 a.m. in the middle of the night.

Instead, I think super resolution should only be used for algorithms which use temporal tricks to extract information from adjacent frames and use that to improve the details on the current frame.

Guessing algorithms that attempt to learn, by looking at billions of images and then making intelligent guesses about the missing detail in a picture or frame of video have a major problem: the world is controlled by entropy, where everything is getting more chaotic all the time. Guesses are still guesses, no matter how well informed, and guesses are, almost by definition, wrong.

Also, you are attempting to guess at an infinite set. For bounded problems, the approach might work, but the universe is, from the standpoint of the human mind, infinite.

[feisty2]

Again back to square one
If ur goal is to make a small image identical to its native larger version, it's simply impossible, so is denoising, deinterlacing, motion flow/framerate upconverting, deblocking, literally any ill posed problem is impossible to solve, no need to talk about entropy, I majored theoretical physics in college, and ill posed problems are things u encounter every fucking day in the image / video processing field, now u can quit ur business immediately cuz u know damn well ur doing stuff that are already theoretically impossible or,,, ur against ur very own logic, unless u change ur goal

[Andouille]

Again back to square one
If ur goal is to make a small image identical to its native larger version, it's simply impossible, so is denoising, deinterlacing, motion flow/framerate upconverting, deblocking, literally any ill posed problem is impossible to solve, no need to talk about entropy, I majored theoretical physics in college , and ill posed problems are things u encounter every fucking day in the image / video processing field, now u can quit ur business immediately cuz u know damn well ur doing stuff that are already theoretically impossible or,,, ur against ur very own logic, unless u change ur goal

English is not my native language but what about spelling ?

[feisty2]

One more thing, all ur "entropy" and "infinite set" stuff are correct but useless, it's so universal that it applies to any ill posed problem, and let's not forget that super resolution is NOT THE ONLY ill posed problem here, how can u even refuse the existence of super resolution while effortlessly admitting the existence of other ill posed problems like denoising and deinterlacing, how is that not ur own version of cherry picking, u cherry picked super resolution out of tons of ill posed problems that are ALL IMPOSSIBLE TO SOLVE, can u even explain ur logic behind all this

[wonkey_monkey]

English is not my native language but what about spelling ?

What an odd - and mistaken - thing to join the thread to say.

[poisondeathray]

Back to the original thread topic, I mentioned 2 mechanisms

The other mechanism deals with finer, higher frequency detail. You usually only see this with higher quality clean sources. Go to any photography forum and you can see examples of stacking increasing fine details (along with the noise reduction, but now we are more interested in "seeing" real detail improvement or resolving.) That real world experience correlates with what the low level test charts demonstrate. This has less to deal with noise removal , and more to do with surpassing optical limits by combining spatially offset captures. At the point on a test chart where you can no longer cleanly resolve fine details, you have moire or aliasing or can no longer distinguish between the the fine patterns. That is the very definition of resolution. That narrow transition band is where you see improvement.

That's the same reason the 2nd mechanism won't work with crappy sources: 1) the crappy source doesn't have details in that range in the first place 2) usually there are other overpowering reasons such as noise, compression artifacts . The typical security cam footage is usually worse than VHS resolution.

On a resolution chart, you have typical patterns such as rings , trumpets etc.. Some sharpening algorithms, convolution kernels, perhaps antialiasing processing might improve one pattern. But if you look at other patterns - they get get worse - and you will develop moire patterns, ringing, oversharpening halos, aliasing. There is no way to "cheat" all patterns on those low level tests. Mild sharpening might be ok, but the finer patterns are usually sensitive and can pick that up. There are different types of test charts and patterns used for different scenarios. But the point is a true improvement in resolution should be be visible in all patterns. That's what a higher resolution camera looks like.

Here is the demo package so the stacking can be done in avisynth. Requires lagarith. Merge() results were unprocessed beyond the stack (some might apply mild sharpen, but the stronger the sharpening, the stronger the halos and aliasing worse, so technically lower resolution) . Although this is considered an "objective" test, there is some subjectivity in it - people debate all day about what resolution their camera actually shoots, they banter about where the aliasing actually starts, or where it's actually "clean". Same with other tests like latitude , dynamic range or how many stops they can achieve.


test chart package (5.35MB)
http://www.mediafire.com/file/m848xbs8bpffv86/testchart%20demo.7z

Before
https://s9.postimg.org/fbx4dba67/original.png

Stack
https://s9.postimg.org/y48zgwlzz/mean_stack.png


It's clear there is an increase in actual resolution. But as mentioned earlier, the magnitude is just a mild increase in actual resolution. It's not CSI level, going to make a completely unreadable sign or license plate suddenly readable. If you go through the 1st post photoshop example, reexamine the pictures, this is exactly what is going on, and the author of that article had similar comments.

So 2 mechanisms behind the stacking effect

[feisty2]

also there's another major logic flaw in johnmeyer's earlier statements I forgot to mention, he tried to prove the irrationality of guessing algorithms by using the second law of thermodynamics as his evidence, but it doesn't work that way, the whole information theory based system and entropy thing, they just don't work on guessing algorithms, the second law of thermodynamics states that the total entropy can never decrease over time for an isolated system, traditional upsampling algorithms are, well, indeed isolated systems cuz when they deal with some image, that image itself is the only known information to the system, and that makes it "isolated", and therefore the second law of thermodynamics applies, and it's not like that for guessing algorithms, guessing algorithms are not isolated systems, the known information to a guessing algorithm is the image to be processed AND all of its past experiences (knowledge that it learned from somewhere else before), the information of the image to be processed is NOT ENCOLSED, it could interact with another system (the past experiences of the guessing algorithm) to decrease its own entropy, that's why I have said a million times before example based algorithms are not bound by entropy

and somehow johnmeyer still failed to understand that and even used entropy to beef about guessing algorithms, lol it's like I'm doing the analytical writing section of GRE all over again, can't get enough of it

[ingoldie]

Nowadays I'm trying to remove ringing.

Original
http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

Before(with ringing)
http://www.mediafire.com/file/9p1fmrb8tq0dfdn/Result%289%29.avi

After
http://www.mediafire.com/file/coebq3y6c348139/Result%2810%29.avi


Original
http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi

Before(with ringing)
http://www.mediafire.com/file/i2zv2z9k032tf3j/Result9.avi

After
http://www.mediafire.com/file/mct2ddk37titqnv/Result10.avi
http://i68.tinypic.com/4qk2dz.jpg


What do you think about the results? And can we say this is related to Super-Resolution.

[EHarlen]

Not perfect but better.You are making progress all the time.

[ingoldie]

Original(512x512)
http://www.mediafire.com/file/z6p14ptizc0my85/lena_std.tif

Another Example

http://i67.tinypic.com/t0q5g0.jpg


http://i65.tinypic.com/11r6s7a.jpg

[johnmeyer]

Wow, does that ever look AWFUL!

[wonkey_monkey]

I don't know why people use that image for anything. It's blurry and has a very narrow range of colours. No greens at all (pixels where green is higher than red and blue), from what I remember.

[johnmeyer]

I don't know why people use that image for anything. It's blurry and has a very narrow range of colours. No greens at all (pixels where green is higher than red and blue), from what I remember.I agree. I wasn't sure where it came from, so a few hours ago I did fifteen seconds of searching. I found out what most of you probably already knew, namely that for no good reason (other than she appeared in Playboy) someone started using it as a test image decades ago and people still use it since it is a "standard."

Here's the Wikpedia article:

Lena Söderberg (https://en.wikipedia.org/wiki/Lena_S%C3%B6derberg)

[StainlessS]

I found out what most of you probably already knew

Yep, but who was the very first playmate ?
[Ans=M.M.]
and who was the oldest playmate ?
[Ans=B.P.@~32]

EDIT: Cropped version of colour edited image as originally pubb'ed [for some reason the image proc sample has deliberatly screwed up color].
https://s20.postimg.org/7xt3m33wd/Lenna_Org_zpstoi1s1pl.png (https://postimg.org/image/u9qwfh309/)

EDIT: Here, probably more like what the original unedited image looked like.
http://forum.doom9.org/showthread.php?p=1774304#post1774304
Bottom two images
https://s20.postimg.org/vyzznjipp/Lenna_zpsb3ccnvcu.png (https://postimg.org/image/p8jie3vjt/)

[wonkey_monkey]

EDIT: Here, probably more like what the original unedited image looked like.

Personally I prefer the original uncropped image, but not for its image processing merits ;)

[ingoldie]

I'm getting better result with smooth(without noise) images.

Another Playboy Playmate

Original(514x768)
http://www.mediafire.com/file/420whxbulx3xr38/2e98d68bf78a8d613b4aef96f6cf4613.jpg


http://i67.tinypic.com/29m2hs4.jpg

http://i66.tinypic.com/28m5gx.jpg


By the way, please don't digress.

Please tell what you think about the results? Can we say this is related to Super-Resolution?

And download and watch videos in the following link.

https://forum.doom9.org/showthread.php?p=1833737#post1833737

[EHarlen]

Sharpening is pretty good but I think you need a better Resizer.

[burfadel]

How about this? Less artifacts, and check the detail around her chest (not that I was specifically looking :p) and the background.

http://oi63.tinypic.com/33pamxj.jpg

[ingoldie]

Sharpening is not good enough but upscaling is not bad.

What resizer did you use for upscaling.

Sharpening is pretty good but I think you need a better Resizer.

I'm aware and I need help about upscaling.

[burfadel]

It was only a basic example, but keep in mind there is no Super Resolution done on that image. Also even though the hair is shaper on yours it's structure is degraded.

[EHarlen]

I think you need a better Resizer.

Also I suggest you to use Anti-aliasing.

[ingoldie]

What about original samples in the following link?

https://forum.doom9.org/showthread.php?p=1833737#post1833737

I want to know your effort about sharpening, denoising and upscaling.

[ingoldie]

I use Lanczos4Resize for upscaling and I think I need an anti-aliasing together with a resizer to get smooth lines.

Do you suggest any other resizer instead of Lanczos4Resize and what anti-aliasing do you recommend to use with resizer that you suggest.

[EHarlen]

Try nnedi3_resize16 instead of Lanczos4Resize

[ingoldie]

I came across a very strange case.

I'm using many filters(more than 15) and big values for my script.

In normally it takes 3-4 minutes to export the following 4 second video.

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi


Just now, when I changed the value of a parameter in my script it took 1 second, yes I'm not joking only 1 second.

But when I repeat the same thing it is not working.


What does it mean.

[raffriff42]

What does it mean.

https://www.dropbox.com/s/3dk510hk4zca1mu/post-your-script-A4B.gif?raw=1

[johnmeyer]

Why is this thread allowed to continue? Can we just close it. Please? He won't post his script; he is posting the same exact clip that he's posted over and over and over again (at least this time he included sound!!). And he never does anything with the generous, well-intentioned advice that has been offered on so many occasions.

[ingoldie]

Original(514x768)
http://www.mediafire.com/file/420whxbulx3xr38/2e98d68bf78a8d613b4aef96f6cf4613.jpg

Script for upscaling

ImageReader("C:\Users\....\2e98d68bf78a8d613b4aef96f6cf4613.jpg").lanczos4resize(1028,1536).converttoyv12()
maa2().maa2().maa2()

http://i67.tinypic.com/29m2hs4.jpg

http://i66.tinypic.com/qye4oz.jpg

[EHarlen]

Use maa2 (ss=1.2) instead of maa2().maa2().maa2()

[ingoldie]

Thank you!

[StainlessS]

Maybe you like to play a bit with this NonlinUSM addition.


ImageReader(".\2e98d68bf78a8d613b4aef96f6cf4613.jpg").lanczos4resize(1028,1536).converttoyv12()
A=maa2(ss=1.2)
B=A.NonlinUSM(z=3, pow=1.1, str=0.33, rad=9) # Maybe try change pow(my usual mod pref).
#Return A++B
Subtract(B,A)


https://s20.postimg.org/omeeameyl/image.jpg

After subtract, lighter shows where NonlinUSM result B is lighter than A.

Although above setting quite light (small changes), will likely require significantly higher bitrate for clip encode.

EDIT: After applying range limiting to 16->235 range(via Levels), then McDegrainSharp(frames=1), and then above NonlinUSM to various clips
can be great, but about doubles required bitrate. (pow 1.1 is my usual default).

[ingoldie]

Original(514x768)
http://www.mediafire.com/file/420whxbulx3xr38/2e98d68bf78a8d613b4aef96f6cf4613.jpg

Here is the best result I have ever received with upscaling.

ImageReader(".\2e98d68bf78a8d613b4aef96f6cf4613.jpg")
eedi3_rpow2(rfactor=2,cshift="spline36resize")

http://i67.tinypic.com/29m2hs4.jpg

http://i63.tinypic.com/6h496x.jpg


But is there a better usage of eedi3_rpow2?

[EHarlen]

If you use spline144resize instead of spline36resize you will get sharper results.

eedi3_rpow2(rfactor=2,cshift="spline144resize")

Download SplineResize for spline144resize

http://avisynth.nl/index.php/Lanczosplusv3#Resizers

[ingoldie]

Please help!!!

My PC shuts down when I use the following script for this .jpg file.

http://www.mediafire.com/file/420whxbulx3xr38/2e98d68bf78a8d613b4aef96f6cf4613.jpg

https://forum.doom9.org/showthread.php?p=1730240#post1730240

eedi3 (dh=true, alpha=1, beta=0, gamma=0, nrad=3, mdis=40, sclip=nnedi3 (dh=true, nsize=3, nns=4, qual=2, etype=1))
turnleft ()
eedi3 (dh=true, alpha=1, beta=0, gamma=0, nrad=3, mdis=40, sclip=nnedi3 (dh=true, nsize=3, nns=4, qual=2, etype=1))
turnright ()



and the following script is very slow

https://forum.doom9.org/showthread.php?p=1722556#post1722556

eedi3 (dh=true,alpha=0.1,beta=0.5,gamma=60,mdis=40,nrad=3,sclip=nnedi3 (dh=true,nns=4,qual=2,nsize=0))
turnleft ()
eedi3 (dh=true,alpha=0.1,beta=0.5,gamma=60,mdis=40,nrad=3,sclip=nnedi3 (dh=true,nns=4,qual=2,nsize=0))
turnright ()


What is wrong with it?

[raffriff42]

First, use search,
https://www.google.com/search?q=My+PC+shuts+down :search:
and try to eliminate the common causes -- overheating due to dust or a fan failing, or hardware errors like bad memory.

Maybe the image is designed to crash your computer, to install a virus.

Try running AviSynth in a different software environment.

[ingoldie]

Please help!!!

When I use the following script

AVISource(".\Original.avi")
eedi3_rpow2(rfactor=2,cshift="spline36resize")

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

I'm getting this artifact


http://i68.tinypic.com/ng7pc2.jpg


How can I fix it?

[StainlessS]

Which artefact (there are probably some to choose from) is that ?

Please help those wanting to help you.

[SaurusX]

Please help!!!

When I use the following script

AVISource(".\Original.avi")
eedi3_rpow2(rfactor=2,cshift="spline36resize")

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

I'm getting this artifact

How can I fix it?

Are you ashamed of your script or scared of criticism? Is that why you won't post it?

[ingoldie]

StainlessS

Isn't it clear enough?

http://i64.tinypic.com/20fvtkw.jpg

[ingoldie]

SaurusX

That problem is not about my script.

It is about eedi3_rpow2(rfactor=2,cshift="spline36resize")

[Groucho2004]

AVISource(".\Original.avi")
eedi3_rpow2(rfactor=2,cshift="spline36resize")

I'm getting this artifact
http://i68.tinypic.com/ng7pc2.jpg
How can I fix it?
I can reproduce this.

Try NNEDI3 (nnedi3_rpow2(rfactor = 2, nsize = 3, nns = 4, qual = 2, cshift = "spline36resize").

[ingoldie]

I don't want to use NNEDI3.Because it doesn't look natural. It makes image look like an oil painting

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi


eedi3_rpow2(rfactor=2,cshift="spline36resize")

http://i64.tinypic.com/vn0yfp.jpg


NNEDI3 (nnedi3_rpow2(rfactor = 2, nsize = 3, nns = 4, qual = 2, cshift = "spline36resize"))

http://i67.tinypic.com/5lz5vm.jpg

[FranceBB]

I don't actually deal with compressed sources. Your source it's a compressed xvid and has many issues.
In order to properly upscale and do a Remaster, it would be useful to at least have the original BetaCAM transfer, 'cause in your case there are mixed artifacts: the ones from the digital compression (xvid) and the ones from the analog world.
There's a lot of grain, with a light aliasing, ringing on the edges, haloing, spots and scratches.

nnedi3_rpow2(cshift="Spline64ResizeMT", rfactor=2, fwidth=1920, fheight=1080, nsize=4, nns=4, qual=1, etype=0, pscrn=2, threads=0, csresize=true, mpeg2=true, threads_rs=0, logicalCores_rs=true, MaxPhysCore_rs=true, SetAffinity_rs=false, opt=3)

Actually makes it look worse, 'cause it does its job, but it sharps details as well as issues.
I don't actually like to see artifacts and I prefer the video to be blurry rather than noisy/grainy.
However, I'm at work right now; I don't have avisynth installed and I can only encode in XDCAM50 ('cause we work in XDCAM50).
This is as far as I can get: Download (https://drive.google.com/open?id=1saUfDSA63y3XGS6z4Ale3TaNRQ8QmNak)


LoadPlugin("C:\Encoding\plugins\ffms2.dll")
LoadPlugin("C:\Encoding\plugins\TDeint.dll")
LoadPlugin("C:\Encoding\plugins\TIVTC.dll")
LoadPlugin("C:\Encoding\plugins\Plugins_JPSDR.dll")
LoadPlugin("C:\Encoding\plugins\despot.dll")
LoadPlugin("C:\Encoding\plugins\GamMac.dll")
LoadPlugin("C:\Encoding\plugins\flash3kyuu_deband.dll")


#Indexing audio and video (8CH. audio channels)

video=FFVideoSource("Original(2).avi")
audiook=FFAudioSource("Original(2).avi")
audio=MergeChannels(audiook, audiook, audiook, audiook)
AudioDub(video, audio)

#Crop from 4:3 to 16:9

Crop(4, 48, -0, -50)

#Despot

DeSpot(mthres=16, mwidth=7, mheight=5, merode=33, interlaced=false, show=0, show_chroma=false)

#Upscale

nnedi3_rpow2(cshift="Spline64ResizeMT", rfactor=2, fwidth=1920, fheight=1080, nsize=4, nns=4, qual=1, etype=0, pscrn=2, threads=0, csresize=true, mpeg2=true, threads_rs=0, logicalCores_rs=true, MaxPhysCore_rs=true, SetAffinity_rs=false, opt=3)

#Temporal and Spacial denoise

ConvertToYUY2(interlaced=false)
TemporalSoften(4,4,8,15,2)
SpatialSoften(16, 4, 8)

#Color correction

ConvertToRGB(interlaced=false)

GamMac(dither=true, LockChan=-3, Scale=2, RedMul=1.0, GrnMul=1.0, BluMul=1.0, loTh=0.0, hiTh=0.0, LockVal=128, RngLim=12, GamMax=10.0, omin=16, omax=235, Show=false, Coords=false)

#Debanding

Converttoyv16(interlaced=false)

f3kdb(range=15, Y=45, Cb=30, Cr=30, grainY=0, grainC=0, sample_mode=2, blur_first=true, dynamic_grain=false, opt=3, mt=true, keep_tv_range=true, input_depth=8, output_depth=8)

#Loudness Correction

ResampleAudio(48000)
Normalize(0.89)
ConvertAudioTo24bit()

#making sure it's TVRange

Limiter(min_luma=16, max_luma=235, min_chroma=16, max_chroma=240)

#PAL interlacing


ConvertFPS(50)
assumeTFF()
separatefields()
selectevery(4,0,3)
weave()

[ingoldie]

I can't make progress and I need help to get rid of artifacts.

https://forum.doom9.org/showthread.php?p=1833737#post1833737


Please help me to get a result without artifacts.

[poisondeathray]

I can't make progress and I need help to get rid of artifacts.

https://forum.doom9.org/showthread.php?p=1833737#post1833737


Please help me to get a result without artifacts.


Don't sharpen so much.

If you start with noise and ringing artifacts, you're making it worse by sharpening .

[Groucho2004]

Don't sharpen so much.People have been telling him/her this for roughly 4 years. I guess he/she wants to have his/her cake and eat it too.

[ingoldie]

Don't sharpen so much.

People have been telling him/her this for roughly 4 years.

YES, don't tell me this.

My purpose is already to receive an over-sharpened result.
So, I don't want to decrease the value of sharpening.


In short, I'm trying to create a script to do this.(up-convering from SD to HD using super-resolution technology)
https://www.youtube.com/watch?v=D7-_kK35cIA

By the way, I have been looking for a sofware to do this for years.

And I can't believe there is no a software to up-convert from SD to HD.

How is it impossible digitally in today's world where technology has developed so fast.

[StainlessS]

Good luck with that !

EDIT: look up, "tilting at windmills".

[poisondeathray]

YES, don't tell me this.

My purpose is already to receive an over-sharpened result.
So, I don't want to decrease the value of sharpening.




OK, but you asked about the artifacts

Those artifacts are being enhanced as a DIRECT result of your sharpen and edge enhancement

If you want it oversharpened , but without those nasty artifacts, you need to filter out either the artifacts, or sharpen everything other than the artifacts . Perhaps selective filters or masking might be one way

Also, sharpening or any type of edge contrast enhancement isn't the same thing as "super-resolution" (any type of SR)



In short, I'm trying to create a script to do this.(up-convering from SD to HD using super-resolution technology)
https://www.youtube.com/watch?v=D7-_kK35cIA

By the way, I have been looking for a sofware to do this for years.

And I can't believe there is no a software to up-convert from SD to HD.

How is it impossible digitally in today's world where technology has developed so fast.



That's nice, especially with the random motion, leaves. (ie. It's not a static object, only with camera motion.)

But how do you know that video is "authentic" ? Is there a playable demo , or can user verify results with same or different video? Searching google for "Mirtemis" doesn't reveal any info other than those videos

I can take a HD video and downscale it to SD, and make "fake" results by saying I made a super-duper-awesome SD to HD filter too

Even if we assume results were "real" , what makes you think it's achievable with an avs script ?

At least with the photoshop method , image stacking, ikena - it's real - at least to an extent. There is a basis in mathematics, signal processing - and user can verify results. I posted additional examples where you can do parts of it in avisynth.

[StainlessS]

Ingoldie, I certainly admire your tenacity, (who could not), but you seek that which is "unfindable'',
let it go and waste your time on more achievable goals, PDR above makes some real good points, as have many others
in the past, but you never take heed of anything anybody says. If you ever do find something that tickles your fancy,
then please do come back and surprise us all, I would love to be proved wrong on things I have said.

[johnmeyer]

Ingoldie, I certainly admire your tenacity, (who could not), but you seek that which is "unfindable'',
let it go and waste your time on more achievable goals, PDR above makes some real good points, as have many others
in the past, but you never take heed of anything anybody says. If you ever do find something that tickles your fancy,
then please do come back and surprise us all, I would love to be proved wrong on things I have said.And quite wasting our time. I still read your posts, expecting that you will eventually post something different, or finally figure out that what you keep endlessly (almost three years) posting about cannot be done. I then get ticked off because I realize I just wasted time that I'll never get back.

[wonkey_monkey]

In short, I'm trying to create a script to do this.(up-convering from SD to HD using super-resolution technology)
https://www.youtube.com/watch?v=D7-_kK35cIA

Either there was a lot of detail in the original video that has been lost because of YouTube compression, or it is faked. I suspect the latter.

How is it impossible digitally in today's world where technology has developed so fast.

Some things are fundamentally and provably impossible.

[ingoldie]

Some things are fundamentally and provably impossible.

But I don't think my efforts are too bad.

https://forum.doom9.org/showthread.php?p=1833737#post1833737

If I can get rid of artifacts remained I will receive an exact result about upscaling.

[ingoldie]

Please let me know what your thoughts/comments are on my efforts.

And do you have any suggestions on how I can improve them.

Please help!!! I don't wanna give up.

[lansing]

Can we get this thread lock already? This loop is getting so stupid.

[johnmeyer]

Can we get this thread lock already? This loop is getting so stupid.I asked for the same thing, back in February in post #125. (https://forum.doom9.org/showpost.php?p=1834891&postcount=125)

And it isn't just getting stupid, it achieved that status several years ago when he started a similar thread, using exactly the same images, back in 2016!!:

Have you ever used VideoCleaner? (https://forum.doom9.org/showthread.php?p=1771776#post1771776)

That thread accumulated 98 responses. While it initially sounded like he was asking about noise reduction, it was, in fact, a post about super resolution.

When I made the mistake of posting in his threads, he started sending me private messages, over and over again. I just looked at my doom9.org sent box and I asked him on three occasions, after he had blitzed me with over 20 PMs, to quit sending messages. In desperation, I started posting his PMs publicly in the threads he started in hopes that other people could bring additional pressure to bear to get him to stop harassing me. The other reason I wanted him to stop is that I had to delete a whole bunch of messages in my inbox, since his PMs caused me to reach doom9.org's 200 message limit.

He responded by sending me yet another PM asking me to delete my posting of his PM to me. I then sent back this message: "If you don't see the irony in sending me a PM asking me to delete my public post whose purpose was to get you to stop sending me unwanted and unwarranted PMs, then you need to look up the definition of that word."

So, between his two threads, this has been going on since June 2016, with over 250 responses in the two parallel threads, and absolutely nothing to show for it.

So I totally agree: lock this thread!

[StainlessS]

John, you can save contents of your inbox [EDIT: probably outbox too] to disk, just click on a few buttons and stuff, your bound to find it
(I found it after having deleted all mine a year or two ago).

EDIT: John, ignore above, I think I clicked on wrong link from some time ago, and answered to that
(I also seem to remember giving a similar answer and probably to your same post, again some time ago).


EDIT: This is post I was answering from page 1 of this thread:- https://forum.doom9.org/showthread.php?p=1811279#post1811279
Seems my original answer to other post has disappeared, probably the one that Ingoldie asked me to delete, which I did.

[poisondeathray]

Can we get this thread lock already? This loop is getting so stupid.

But if this thread gets locked then another will pop up and replace it... with the same thing, same topic(s)

I'd rather have two or three dedicated ingoldie threads than a two dozen scattered ones

You can choose not to click on a thread

Part of the problem is bumping the thread with replies - like I'm doing now... baad me

[color]

I dont know if this is what you are looking for but it can make a good quality upscale depends on what size and quality he movie is in. http://www.gyroshot.com/simpleslug.htm

[EHarlen]

Original(512x512)
http://www.mediafire.com/file/z6p14ptizc0my85/lena_std.tif

Another Example

http://i67.tinypic.com/t0q5g0.jpg


http://i65.tinypic.com/11r6s7a.jpg


Have you ever used Topaz A.I. GIgapixel?

https://topazlabs.com/ai-gigapixel/

For the original as a TIFF image:http://www.lenna.org/

lena_std

http://i63.tinypic.com/fbm4wz.jpg


lena_std-edit(2x)

http://i67.tinypic.com/2akjsx1.jpg

[Groucho2004]

Have you ever used Topaz A.I. GIgapixel?Some of the details are pretty good but the skin looks terrible, waxy and artificial.

Here is a quick try with Photoshop bicubic sharp resize + Multiscale Sharpen with Astra Image:

http://www.iol.ie/~schubert/pics/lena_big_sharp.png

[Gser]

This guy used gigapixel to upscale SD video: https://captrobau.blogspot.com/2019/03/remastering-star-trek-deep-space-nine.html
You can call me completely unimpressed.

[ingoldie]

Have you ever used Topaz A.I. GIgapixel?

https://topazlabs.com/ai-gigapixel/

For the original as a TIFF image:http://www.lenna.org/

lena_std

http://i63.tinypic.com/fbm4wz.jpg


lena_std-edit(2x)

http://i67.tinypic.com/2akjsx1.jpg


Can someone please write an equivalent avs script of Topaz A.I. GIgapixel?

[Gser]

Can someone please write an equivalent avs script of Topaz A.I. GIgapixel?
I hereby dub this QuantumNeuralBlockChainAIHyperResize.avsi
BicubicResize(target_width, target_height, b=0, c=0.5).Blur()

[Frank62]

Automatic colorization by training with enough standard-data does work very well already, as we know.
Resizing with really(!) new created details seems possible to me this way, at least better than anything else. I am very curious what the next years will bring, in my opinion these techniques are one of the real turning points (in terms of software) of our time.

I didn't test A. I. GP, I am sceptical that this already works, but maybe it does?

[ingoldie]

Topaz A.I. GIgapixel is pretty good about denoising but my efforts are not.

Do you have an avs script that Topaz A.I. GIgapixel does about denoising?

Please help!!!

[Groucho2004]

Topaz A.I. GIgapixel is pretty good about denoising but my efforts are not.

Do you have an avs script that Topaz A.I. GIgapixel does about denoising?

Please help!!!Here's what you should do:

Write an email to the Topaz developers and demand the source code for that program. Then, study the source code and use their algorithms to write an Avisynth plugin.

If the developers won't give you the source code, use a disassembler to get an ASM dump. Figure out the algorithms from that.

If you can't figure out the algorithms from the source, pay someone to do it for you.

See, it's that easy.

[zorr]

TecoGAN (https://www.youtube.com/watch?v=pZXFXtfd-Ak) looks quite impressive.

[wonkey_monkey]

Hmmmmm. It would be nice if they'd used a real low res source instead of downscaling something. It looks suspiciously (albeit slightly) undersampled, which real video never is.

It reminds me of the hype over "drizzling" for processing astronomical images - because it was used by NASA on space telescope images, it must be good for everyone! But in truth it only worked so well on NASA images because the CCDs in the space telescopes (can't remember if it was Hubble or another one) were spatially undersampling by a large degree, which is something modern consumer CCDs don't do.

[Selur]

Btw. has anyone with a RTX card tried the free AI-Res Up feature on images from normal video content? (https://www.nvidia.com/en-us/geforce/news/new-gfe-features-latest-games-get-ansel-and-highlights/, https://forums.guru3d.com/threads/you-can-feed-the-ansel-ai-up-res-system-any-image-you-want.424382/) Non-RTX cards can only use the gray scale version.

Also has anyone looked at ESRGAN (https://github.com/xinntao/ESRGAN, https://www.pirm2018.org/PIRM-SR.html) ?

[Wolfberry]

I don't have a RTX card, as for ESRGAN, it is available in WolframRhodium's Super-Resolution-Zoo (https://github.com/WolframRhodium/Super-Resolution-Zoo).

There are helper functions named super_resolution in muvsfunc and muvsfunc_numpy that help you use the model files.

[Selur]

@Wolfberry: thanks for the info, didn't know that! :)

[ingoldie]

I removed all halos. It is bettter now. By the way,less noise.


Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

Result

http://www.mediafire.com/file/n0agfs7qmdop84r/Result%2811%29.avi


Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi

Result2

http://www.mediafire.com/file/wi49jb2ez19w3ew/Result11.avi/file

[ingoldie]

I removed all halos. It is bettter now. By the way,less noise.


Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

Result

http://www.mediafire.com/file/n0agfs7qmdop84r/Result%2811%29.avi


Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi

Result2

http://www.mediafire.com/file/wi49jb2ez19w3ew/Result11.avi/file

But not completely removed. So please help!!!

What kind of noise is this and how can I can remove it?


http://i66.tinypic.com/5bqvb6.png

http://i67.tinypic.com/6s97bc.png

[Groucho2004]

What kind of noise is this and how can I can remove it?
https://i.postimg.cc/7h7KGvJm/Image2.png

Doesn't look like noise to me. I think some kid drew phallic shapes on your image.

Seriously, I lost count on how many times you asked that question about the same bloody image. Asking the same question over and over again and expecting a different answer is worrisome.

[johnmeyer]

Seriously, I lost count on how many times you asked that question about the same bloody image. Asking the same question over and over again and expecting a different answer is worrisome.I totally agree.

Can't this $%*@# thread just die?

[Dreamject]

>What kind of noise is this and how can I can remove it?
Say goodbye to 1.44GB XVID DVDRip and say hello to bdremux
Image can be improved (a little, it's very bad) anyway if you use deband shader
From http://i67.tinypic.com/6s97bc.png to
https://s16.directupload.net/images/190424/uosdtu9t.png (https://s16.directupload.net/images/190424/3sswcjpe.png)

[Groucho2004]

Image can be improved (a little, it's very bad) anyway if you use deband shader
Apart from a little colour shift, I see no difference.

[real.finder]

I totally agree.

Can't this $%*@# thread just die?

he also PM me few days ago, it's so Annoying

[Dreamject]

Apart from a little colour shift, I see no difference.

It has less visible blocks, but original quality too bad
https://s16.directupload.net/images/190425/wxwchqv5.png (https://s16.directupload.net/images/190425/w7fqkbek.png) https://s16.directupload.net/images/190425/sw7jmgc2.png (https://s16.directupload.net/images/190425/uwy4w358.png)

[EHarlen]

But not completely removed. So please help!!!

What kind of noise is this and how can I can remove it?


http://i66.tinypic.com/5bqvb6.png

http://i67.tinypic.com/6s97bc.png

Try http://avisynth.nl/index.php/Category:Deringing_%26_Mosquito_Noise

[FranceBB]

What kind of noise is this?

It ain't noise, it's ringing. https://en.wikipedia.org/wiki/Ringing_artifacts

This is another example of ringing:

http://blog.biamp.com/wp-content/uploads/2017/02/algolith-fig2s.jpg

[ingoldie]

It ain't noise, it's ringing. https://en.wikipedia.org/wiki/Ringing_artifacts

This is another example of ringing:

http://blog.biamp.com/wp-content/uploads/2017/02/algolith-fig2s.jpg

What about the solution?

[johnmeyer]

What about the solution?Don't use compression when rendering video. These are DCT (Discrete Cosine Transform) artifacts.

[ChaosKing]

http://avisynth.nl/index.php/HQDering_mod

[lansing]

Would people stop feeding this troll my god!

[johnmeyer]

Would people stop feeding this troll my god!I second that request. This thread has been going on for almost two years. What's worse, he previously had another thread which wouldn't go away. It ended up on the exact same topic:

Have you ever used VideoCleaner? (https://forum.doom9.org/showthread.php?t=173636)

The artifacts now being discussed when this thread was revived a few days ago, are artifacts he introduced with his "super resolution" attempts, and were not there in the original.

[ingoldie]

Anti-ringing filter:

https://forum.doom9.org/showthread.php?p=1655611#post1655611

nrJinc36Resize looks like Lanczos4Resize to me.

Can someone please help about

https://www.google.com/search?q=DeGrainMedian&source=lnms&tbm=isch&sa=X&ved=0ahUKEwiogMjTzvrhAhXPa1AKHV1RDS0Q_AUIDygC&biw=1366&bih=657#imgrc=Tem4TeF1Wn5PrM:

[Dreamject]

>Would people stop feeding this troll my god!
People likes conflicts and from objective side this forum must dead. It's very old-schoolly, old engine, no mobile, uncomfortable, newbie-unfriendly captcha looks like 'who is the most active user with pink hairs on this site's (was actual 15 years ago). Even big enthusiastics lost interest to subject after some 10 years. If you drop trolls and human discussing

[ingoldie]

I removed all halos. It is bettter now. By the way,less noise.


Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi

Result

http://www.mediafire.com/file/n0agfs7qmdop84r/Result%2811%29.avi


Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%282%29.avi

Result2

http://www.mediafire.com/file/wi49jb2ez19w3ew/Result11.avi/file

But not completely removed. So please help!!!

What kind of noise is this and how can I can remove it?


http://i66.tinypic.com/5bqvb6.png

http://i67.tinypic.com/6s97bc.png


I think I finally did it

Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi/file


Result

http://www.mediafire.com/file/971vwbx8b2j3s1t/Result12.avi/file


Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%25282%2529.avi/file


Result2

http://www.mediafire.com/file/2r394euucxgovkv/Result%252812%2529.avi/file

[EHarlen]

Much better but there are artifacts caused by ringing removal.

https://i.ibb.co/M6JKRFY/vlcsnap-2019-07-18-19h48m03s314.jpg (https://imgbb.com/)

[ingoldie]

Yes EHarlen, you are right.Thanks for the warning.
But, how can I solve this problem?

[StainlessS]

Here is another warning.

You is probably gonna create another problem when and if you fix the new one.

[ingoldie]

Much better but there are artifacts caused by ringing removal.

https://i.ibb.co/M6JKRFY/vlcsnap-2019-07-18-19h48m03s314.jpg (https://imgbb.com/)

Problem solved

Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi/file


Result

http://www.mediafire.com/file/y1qdrft82yo5p38/Result13.avi/file


Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%25282%2529.avi/file


Result2

http://www.mediafire.com/file/sfwaa0jc4jrxw86/Result%252813%2529.avi/file

[Groucho2004]

Problem solvedCongratulations. It does look better than the original. I'm impressed that it only took you 2 years to fix a 2 second clip. Time for you to move on the some different footage. I'm looking forward to your results in a few years time (if I'm not 6 feet under by then...).

Edit - :D

[ingoldie]

Can I see your effort? Can you receive an extremely sharp result without halos,denoise and remove ringing without detail loss, deformation, natural look?

I bet you can't! Please try!

[jmartinr]

@Ingoldie: Time to post your script ;-)

[Groucho2004]

Can I see your effort?I didn't make any effort. And why would I, it seems you used your magic wand to achieve the impossible:
an extremely sharp result without halos,denoise and remove ringing without detail loss, deformation, natural look
I bet you can't!
True, I can't do magic.

[Groucho2004]

@Ingoldie: Time to post your script ;-)Never going to happen.

[StainlessS]

Whatever else you can say about Ingoldie, you gotta admire his persistence.
Good on you IG, now go pour yourself a nice big glass of brandy, sit and relax
whist you plan your next project.

[Didée]

I second that request. This thread has been going on for almost two years.
Another important thing is a sharpen script that takes not only two, but three input clips as source. This request haunts me not since two, but since three or four years, if not more ...

[videoh]

Another important thing is a sharpen script that takes not only two, but three input clips as source. This request haunts me not since two, but since three or four years, if not more ... Why does that haunt you?

[ingoldie]

Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi/file


Result

http://www.mediafire.com/file/y1qdrft82yo5p38/Result13.avi/file


Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%25282%2529.avi/file


Result2

http://www.mediafire.com/file/sfwaa0jc4jrxw86/Result%252813%2529.avi/file


denoise and remove ringing without detail loss, deformation, natural look?


I'm sorry for talking big. I realized there are some artifacts(detail loss, deformation)


https://i.ibb.co/26G2B9F/vlcsnap-2019-10-30-17h13m14s426.png (https://ibb.co/4fhyxqF)


Solved


Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi/file

Result

http://www.mediafire.com/file/3jyo569j0ujpyjy/Result14.avi/file



Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%25282%2529.avi/file


Result2

http://www.mediafire.com/file/3sizsr5597mkdkk/Result%252814%2529.avi/file

[ingoldie]

By the way, about my purpose:

I've been looking for a software to improve visual quality of the videos with poor visual quality for a long time.

But I can't find it. At least none of them are better than mine(maybe I'm wrong) and I can't believe it.

What do you think about that? Am I wrong? Is there a software to improve visual quality of the videos with poor visual quality?

If no, I think I have to use my AviSynh script.

[StainlessS]

Ingoldie, did you see this post :- https://forum.doom9.org/showthread.php?p=1888905#post1888905

Anybody done any tests to see if hi bitdepth performs better in mvtools than lo bitdeth.
For lo BD and eg pel =4, so max vector length about 32, & so if movement
greater then cannot be
Used.

God I hate auto correct, Mobile.

Edit: Presumably vectors can be much longer in HBD.
Edit: Of greater importance where HD clip.

I was thinking in particular of your Man + Woman + 2 or 3 kids with ball, walking in garden, clip. [the one you worked on for several years].

EDIT: This one:- "Ingoldie_Sample_SD_8mm_18_fps.avi" [The 'Ingoldie_" part may not not have been part of original name]


EDIT: Also, perhaps you noticed that I did not answer your PM of about 6 months ago, I will not in furture respond to any PM you send me,
I'm sick of telling you not to do it, and will not be bothering to tell you again. Use the forum, thats what it is for.
[Not sure, think I deleted it without reading]

[EHarlen]

Bravo ingoldie! You are making great progress.Your result really looks much better than the original.
Especially ringing removal and denoising without detail loss is pretty good. I thought it was impossible!
Sharpening is already excellent.

[Groucho2004]

Time to share your script after years of leeching knowledge off of members of this forum.

[johnmeyer]

If I were to vote, I'd say it is time to close this thread which has been going in circles for 10 of its 11 pages. Each "new" post is the same exact video, with the same exact questions, followed by well-meaning people always trying to help.

Thus, this thread has become a pointless, endless loop. Also, as StainlessS and I have both experienced (as have others, I'm sure), the OP will PM and ask for a private tutorial which is quite annoying.

[StainlessS]

Now now girls, you is all knowing that if this thread closes, another will replace it, so, lets just keep this one.

I did not wanna start a "lets all be hatin' on Ingoldie" type thing, lets not do that please.
Its nice to see I.G. advancing, and being happy with what he has succeeded in achivieving.
Peace and good will to all men and women too come to that.

[johnmeyer]

I wasn't trying to be unpleasant toward the OP. My only issue is that, since I get notifications for any thread in which I've contributed, I'd really like it if I didn't have to deal with this (yes, I know I can turn off notifications for individual threads).

Also, I don't know about other people, but I only set aside a minute or two each day to go to various forums. The more of my time taken by pointless threads like this one, in which nothing actually gets solved, the less time I have to help people who can see a problem through to the end, actually get the job done, and get on to something else.

Can you imagine if this was being done for a client? That person would have been waiting for 2-1/2 years for approximately sixty frames of video.

[Groucho2004]

I did not wanna start a "lets all be hatin' on Ingoldie" type thing, lets not do that please.
Its nice to see I.G. advancing, and being happy with what he has succeeded in achivieving.
Peace and good will to all men and women too come to that.Be that as it may, you have to look at what he has contributed to the community - nothing, zero, nada. He does not share anything. He's a leech and everyone knows it, particularly those who have been pestered with his PMs.

[StainlessS]

John, the observation bout the HBD vectors is though quite relevant and would likely affect results of the likes of your football/sporty
type clips, with jerky high motion being better dealt with.
I have recently made mods to a long time in delevelopment script, that now does any MC type stuff (FrameRate conversion, and/or MC denoise),
before any upscale, or after any downscale, and seems to provide for better results than the opposite. It occurred to me some time ago (I think
I mentioned it somewhere before the previously linked post) that HD/4K/8K stuff has got to be badly affected and crippled by mvtools maximum
vector length, 64 where pel=2, and 32 where pel=4. High bit depth vectors have got to be better especially for higher rez sources, guess I'm
gonna steal some of real.finder stuff for my own scripts, he seems to be well acquainted with it.
Somebody needs to do an amount of testing, comparing low bit depth MC processing versus high bit depth processing on HD clips,
I'm not targetting you as that "somebody", any old somebody needs to do it. It is my guess that difference in results will be more than significant.

[ingoldie]

I can't understand why you don't tell your thoughts about the result?

Bravo ingoldie! You are making great progress.Your result really looks much better than the original.
Especially ringing removal and denoising without detail loss is pretty good. I thought it was impossible!
Sharpening is already excellent.


By the way, thank you EHarlen

[ingoldie]

Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi/file

Result

http://www.mediafire.com/file/3jyo569j0ujpyjy/Result14.avi/file



Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%25282%2529.avi/file


Result2

http://www.mediafire.com/file/3sizsr5597mkdkk/Result%252814%2529.avi/file

I need a GUI that I will be able to insert and hide(obfuscate) my AviSynth script into it.

Here is the GUI on my mind; Save(to save video files without changing their own formats)

https://i.ibb.co/jTWX6hR/34.jpg (https://ibb.co/f2nw1Mv)


For example, here is a software named VideoCleaner you will see an AviSynth script inserted into AvsPmod.

http://videocleaner.com/

https://forum.doom9.org/showthread.php?p=1892115#post1892115


https://i.ibb.co/Bsbq04D/avsp2.jpg (https://ibb.co/7yP2sgc)


Please help!

[Groucho2004]

I need a GUI that I will be able to insert and hide(obfuscate) my AviSynth script into it.You mean like a safe? Is that 'GUI' supposed to do something? Or just hide a script? As for obfuscating - XOR each character in the script with 2A (HEX) but don't tell anyone.

Edit: Another thought - Why don't you obfuscate VideoCleaner and sell it as your own?

[real.finder]

I think this GPL violation of avisynth in first place if you plan to ship any avs core or script (Theoretically, any avs script published without a license mean it's GPL since it made for avs which is GPL) with some Payware/close-source program/GUI

[TheFluff]

Can someone ban ingoldie for harassing random posters already? I got a PM out of the blue:

Hi,

Can you help me about:

https://forum.doom9.org/showthread.php?p=1900042#post1900042


Regards

I have no idea what made them think I'd be inclined to help, seeing as how I've never even posted in this thread at all before.

(Theoretically, any avs script published without a license mean it's GPL since it made for avs which is GPL)

This makes absolutely no sense. Programs don't magically become licensed under GPL because the compiler or interpreter is GPL. If they did anything compiled with GCC would become GPL.

Shipping Avisynth itself with a closed source program may also be possible, as long as you ship it as a DLL. The FSF takes the view that a closed source program linking to a GPL'ed library is not legal under the terms of the license (the program that links to the GPL library must be available under a GPL-compatible license), but this has not been proven in court and there are court cases that indicate that if you use dynamic linking this might not be enforceable, at least not in the US (see this StackExchange answer which cites Galoob v. Nintendo (https://softwareengineering.stackexchange.com/questions/167773/how-does-the-gpl-static-vs-dynamic-linking-rule-apply-to-interpreted-languages)).

[StainlessS]

Ingoldie obviously wants to hide the fact that he's cobbled several bits of code from other people together, into
his supa-dupa script, and he dont want anybody to know that. [When he tries to sell the result together with
a GUI (which he wants someone else to write for him)].

[Groucho2004]

Ingoldie obviously wants to hide the fact that he's cobbled several bits of code from other people together, into
his supa-dupa script, and he dont want anybody to know that. [When he tries to sell the result together with
a GUI (which he wants someone else to write for him)].
The sad thing about all this is that it took him 3 years to come up with a script that slightly improves one 2 second clip. It's probably not very useful for other source material. On top of that, I bet that nothing in that script came from his own noggin as you pointed out.

Prepare yourselves for another few years of pestering, this time about 'obfuscating GUIs'.

[StainlessS]

He would also not have the faintest idea how to support his commercial offering, and it would quickly garner bad reviews and user warnings to eg "avoid at all costs".
His "get rich quick" scheme just aint gonna work.

[TheGenesis]

Probably he's only trying to break through someones reserve and get his 2 clips cleaned for free :)

[Groucho2004]

Probably he's only trying to break through someones reserve and get his 2 clips cleaned for free :)No. He certainly isn't the sharpest knife in the drawer so he wants to sell other people's intellectual property as his own. That about sums it up.

[ingoldie]

he wants to sell other people's intellectual property as his own.

It is true that I asked for help many times(about deringing without artifacts, denoising without detail loss, sharpening without any halos ...), but nothing worked.

I mean it is my own effort.

By the way, I don't plan on creating a software to sell.

I'm trying to create a Freeware.

[jmartinr]

I and others with me do not believe you.

[real.finder]

I and others with me do not believe you.

yes, he said that he will share the script in the end but he didn't

also I wonder where he got these samples, the mediafire said he upload from Turkey but the content does not seem from there

[StainlessS]

Ingoldie, post your script and forget the Obfuscation stuff, if not then people are gonna think that you are less than credible.
You might get more help if people did not inherently distrust you, "play nice" and others would perhaps behave in a similar manner.

Pestering people is not a good strategy, as you have hopefully found out by now.
If you start to 'Play Nice', I am willing to forgive/forget earlier [EDIT: what seemed to be] mis-deeds [perhaps just mis-understandings].

[ingoldie]

Here is an AviSynth-based software for sale

https://www.videoartifact.com/va/

https://www.videohelp.com/software/Video-Artifact

[Groucho2004]

Here is an AviSynth-based software for saleWhat's your point?

Since this thread is apparently for posting random stuff:
https://i.postimg.cc/NM9TkjPm/yoda.gif

[johnmeyer]

Here is an AviSynth-based software for sale

https://www.videoartifact.com/va/

https://www.videohelp.com/software/Video-ArtifactYes, what's your point?????

The author of the software in that first link is Andrew Revvo, someone who is most definitely NOT "Ingoldie."

I can point to a dozen other sites (or more) which offer such software, either for sale, or for free.

Why don't you just buy one of these programs, do whatever it is you need to do, and quit endlessly posting here, with the same exact posts which show no progress and no new thoughts or ideas?

More to the point, why don't the moderators close down this thread? There has been zero progress in any direction for years (the first post was almost three years ago).

[Groucho2004]

Calm down John. Just ignore his posts and/or put him on an ignore list (not sure if this is possible here).

[Groucho2004]

Here is an AviSynth-based software for saleMore on this misleading claim here (https://forum.doom9.org/showthread.php?p=1849768#post1849768).

[ingoldie]

Why don't you just buy one of these program


what program do you reccomend? I can't find any programs that can give results better than mine




Original

http://www.mediafire.com/file/8ffo4a6i36d7gdn/Original.avi/file

Result

http://www.mediafire.com/file/3jyo569j0ujpyjy/Result14.avi/file



Original2

http://www.mediafire.com/file/acx2y95hnisn4ud/Original%25282%2529.avi/file


Result2

http://www.mediafire.com/file/3sizsr5597mkdkk/Result%252814%2529.avi/file

[Groucho2004]

I can't find any programs that can give results better than mineThat sounds as if you have achieved what you set out to do 3 years ago. Why are you still posting your meaningless comparisons?

[Taurus]

Is this an insider joke I'm not getting? :mad:
I really dont wanna spoil the party,
But for heavens sake, please stop this nonsense
for the peace of everyone.
3 years and no progress?
I'm an old fart but trying hard everyday to achive more knowledge on video encoding (and this is not my only interests)
Mainly I'm into audio.
He(She) is a PITA.
So please if some mods cross this thread:
Ban it, burn it, or send it internet nirvana.
To be enlighten.
Amen

[johnmeyer]

Is this an insider joke I'm not getting? :mad:
I really dont wanna spoil the party,
But for heavens sake, please stop this nonsense
for the peace of everyone.
3 years and no progress?
I'm an old fart but trying hard everyday to achive more knowledge on video encoding (and this is not my only interests)
Mainly I'm into audio.
He(She) is a PITA.
So please if some mods cross this thread:
Ban it, burn it, or send it internet nirvana...
:thanks:

[TheGenesis]

I'm against closing this thread ... it has become somthing like a cultural heritage on this forum :)

[manono]

I'm against closing this thread ... it has become somthing like a cultural heritage on this forum :)
Amusing to some but irritating to others, including me. Thread closed.