madVR - high quality video renderer (GPU assisted)

[madshi]

Originally Posted by kalston

I'm curious because I've started using Smooth motion (as I've heard it's great with high refresh rates)

FWIW, if you can set your display's refresh rate to a straight multiply of the source frame rate (e.g. 23.976Hz, 48.952Hz or 71.928Hz for 23.976fps sources) then you should probably disable smooth motion FRC. The real purpose of smooth motion FRC is to remove pulldown judder when movie framerate and display refresh rate don't match. Ok, in theory you can also use smooth motion FRC as a Reclock replacement to get endless playback without frame drops, and for that using a very high refresh rate is very useful. But I still believe it's better to that the display to a multiply of the source framerate and to disable smooth motion FRC. You might get a frame drop/repeat every couple of hours, but nobody has died from that yet (well, as far as I know)...

Originally Posted by iSunrise

I think you just did too much at the same time, when you introduced too many options in one of the last updates, that's why people really get confused and try out 2 million different settings and they never really provide meaningful feedback for only one option.

Probably, but I would have felt bad, "hiding" already existing features just to get better feedback.

Originally Posted by iSunrise

When we did the dithering tests, there was great communication, screenshots and lots of example shots where you could very clearly detect the positives and negatives and you also reacted very fast to improve them. With finesharp, this was not the case, since no one other than TheLion, me and you even provided screenshots. That's 3 people out of maybe 20 that basically just said "I like A better than B", but they could not tell us, why.

True, but the way things were going I didn't expect the feedback situation to change much, so I made the decision based on the feedback I got. If I had known that 2 days later suddenly contradicting feedback would show up, obviously I would have delayed my decision. Anyway, no harm done yet. Easy enough to revert my decision. Which doesn't mean I will, though. Not decided yet...

Originally Posted by iSunrise

One of the main reasons why a lot of DVDs were so bad, was not because of their low native resolution, but because of their excessive ringing that was applied in post. They are almost completely unwatchable when you upscale them to HD or 4K. Finesharp with the current LL will amplify that even more and that doesn't make any sense when you can just use finesharp with no LL without these negatives.

But in the left half of the squirrel test pattern, actually there were less ringing artifacts with LL turned on compared to off. So I'm not sure if the situation is that simple as you make it sound.

Originally Posted by aufkrawall

It's definitely LL that makes contoures look too thick and dark with FS.
not LL:


LL:


Sample (1 frame):
http://www52.zippyshare.com/v/lTUNZizZ/file.html

So I vote to make not LL the new default option. In fact, I don't see any reason for having a LL option if it distorts images as shown.

LL is devilish. Always, it seems.

As for thinning: With low values I don't see any effect (at least no positive). Combined with SuperRes, it can even look worse because it already seems to introduce tiny bits of aliasing which then get stronger. With higher values, this aliasing also becomes visible without SuperRes. So I don't see any point in this parameter, thus I vote to set it to 0 by default and probably even remove this option.

Thanks for the sample. I'm not sure I can reproduce the aliasing problem, though. Using super-xbr for doubling, followed by FineSharp in upscaling refinement with rather high thinning setting. Can you post a screenshot (preferably without SuperRes) where the aliasing is visible, and maybe mark it in MS Paint with a red arrow or something? And please let me know which exact upscaling/processing options you're using. Thanks!

Originally Posted by aufkrawall

"Better" is not objective. LL scaling seems to always manipulate brightness in a way that leads to an inaccurate result compared to the source. At least from what I've seen.

Scientifically it should be the other way round, at least when downscaling. Do you have some samples?

Originally Posted by ryrynz

Madshi, are there any knobs to turn for Super-xbr with regards to sharpness outside of SuperRes?

Maybe a different kernel, which would make things softer. Not sure yet, though.

Originally Posted by aufkrawall

Ok. However, I for myself have already stumbled over three (with FS now four) more or less "real world" examples in which LL was clearly worse.
Well, then I suppose it might be best if madshi leaves us the option to choose, like with scaling.

Samples?

No, there is not going to be a LL option for sharpening, sorry. There are way too many options in madVR already. At some point I'll remove several (many?) of the current options. (Of course I'll ask for feedback first.)

Originally Posted by ThurstonX

FWIW, I just opened the Settings to confirm the above, and sure enough "Double Chroma Resolution" is checked. That must be the new default setting, as I've never checked that option in the past. I think that maybe "Double Luma Resolution" is also getting checked as a new default (both set to "NNEDI3, 16 neurons"), as I'm pretty sure I didn't check double luma on this laptop (it can't handle it).

There's a bug in the settings dialog which results in "Double Chroma Resolution" to always be checked when you enter the settings dialog with "Double Luma Resolution" checked. But that's just a bug in the settings dialog. Whatever settings you change and store, are saved and used correctly, as far as I can see.

For NEDI and super-xbr you cannot disable "Double Chroma Resolution" if "Double Luma Resolution" is enabled, becase those two algorithms always double both at the same time by design.

Originally Posted by Eyldebrandt

I know madshi wants some report about FS first, but it will be more practical to me to make an overview of the lasts features.

Finesharp
I used it years ago when Didée released it on avisynth after an user of doom9 ask him what was the best sharpener for HD source.
So here we have a important point.
Finesharp is definitively not an option to considering if we talk about sources < 720p.
FS is a factory for aliasing, distorsions, and artifacts. He has to be used only with very HQ content. Linear Light or not, he's too strong for low quality sources, the result is always horrible.
At least in my opinion, and my opinion is the only thing i can provide here.

Thinning control seems to be the equivalent setting to "xstr" in the Didée's avisynth script.
In others words, xstr/thinning is the setting than you absolutely want to turn down the most, unless you're an aliasing fan.

Finesharp is a very destructive sharpener, and to be efficient, i think he will be only used with Blu-ray or 1080p content.

And even on these types of sources, i think the strength should not exceed 1.2, in upscaling refinement, and no more than 0.7/0.8 in image enhancement.

Repair was set at 1.0, and to be honest, i don't know what to think about that. By the fact, i have not succeed to rule if repair fonction is equivalent to the "cstr" factor in the avisynth script. I think is some of, but not completely, cos i suspect the madvr version is quite different in some ways than the original script.
For fact, if repair have some strong common point with cstr, so 1.0 seems to be too high for me.

Thanks for your feedback - good 2nd post on the forum!

Yes, "thinning" is "xstr". However, "repair" is an algorithm tweak implemented by -Vit- when converting the AviSynth script to GPU pixel shaders. Basically "repair" tries to avoid/undo the aliasing problem that occurs when using "thinning". So your experience with the AviSynth script introducing aliasing when using "thinning" needs to be re-evaluated. There is still some aliasing, but it's less a problem compared to the AviSynth script. I'm still playing with the proper "repair" value. The "cstr" option is set automatically by madVR, depending on which strength you've selected.

Originally Posted by Eyldebrandt

LumaSharpen

I discovered this sharpener with a previous build of madVR.
I think it's an adaptation of a SweetFx option, and i welcome it with a sort of mistrust, because i hate SweetFX :d
Anyway, i tried to let him a chance, and after some tests, i have to say he can have some utility, on very HQ content, again.
With really moderate settings, and considering is just an eye candy toy, he adds some kind of pop effect wich is not unpleasant.

As he active on the luma, he brighter just some objects and pixels in a frame, and in some ways, i like it.

Maybe a lead for futures evolutions, add an usharp mask (like Unsharp HQ) to work with luma sharpen will open a new way for those who wants to build a very pop picture. If madshi let Luma Sharpen in madVR, and i'll like he will, this will be an option to hardly considerate.

An other opinion is if you're upscaling the content, you should definitively not turn on a sharpener in image enhancement, whatever it is.

I'm not a big fan of LumaSharpen. It is definitely quite fast, that's its big advantage. But it still introduces quite noticable ringing. At some point I'm probably going to replace LumaSharpen with a slower alternative, something similar to LimitedSharpenFaster, which doesn't ring as much. But for now I'll probably keep LumaSharpen because it's so nicely fast.

Originally Posted by Eyldebrandt

Super-xbr

As i play almost only 1080p content on 3440x1440 monitor and 4K VP, and as i have the rigs to motorize anything, i was a faithful user of NNEDI3 in chroma upscaling and Double resolution.
But super-xbr have insinuate the doubt in me.
On some aspects, super-xbr outclasses NNEDI. Which are aliasing control, clarity and natural look. But he's beaten by NNEDI3 on cleanliness, ringing control and precision.
This is worth for almost all types of content.

Then, super-xbr is incredibly less greedy for a result which is different but fairly comparable in terms of quality to NNEDI3, event at 128 or 256 neurons.

And more than that, it seems to me than super-xbr produce fantastic result in chroma upsampling.

I'm glad you like it that much! To be honest, I was a bit disappointed by the feedback so far. Most users didn't seem to like it much. Probably my fault, due to choosing test images which already had a lot of ringing in them, so it looked like super-xbr would ring a lot.

Originally Posted by Eyldebrandt

1080p content on 1080p TV
- Custom Res at 2560x1440 in Nvidia panel Control (no DSR)
- Chroma upscaling : super-xbr (without SuperRes)
- Luma/chroma doubling : super-xbr
- Upscaling refinement : Super Res (NNEDI3 defaults, 4 passes), FS (strength 0.4, thinning 0.007), LS (strength : 0.25, clamp : 0.045, radius 0.3)
- image downscaling : C-R AR LL

Who's doing the downscaling here? The GPU driver, I suppose? Does this really look better than straight 1080p playback? Oh well, SuperRes doesn't activate for 1:1 playback, maybe that's the reason?

[RainyDog]

Originally Posted by madshi

I'm not a big fan of LumaSharpen. It is definitely quite fast, that's its big advantage. But it still introduces quite noticable ringing. At some point I'm probably going to replace LumaSharpen with a slower alternative, something similar to LimitedSharpenFaster, which doesn't ring as much. But for now I'll probably keep LumaSharpen because it's so nicely fast.

Have you looked at Adaptive Sharpen madshi?

[iSunrise]

Originally Posted by madshi

...Scientifically it should be the other way round, at least when downscaling. Do you have some samples?

While that may be true, the real-world results with madVR are the opposite though, thats why LL was only recommended together with Catmull-Rom downscaling. At least that's what we concluded after we did several tests years ago and e.x. 6233638 and cyberbeing were two people that also strictly recommended against it. LL on some downscalers was full of artefacts compared to gamma light.

[har3inger]

Originally Posted by madshi

I'm glad you like it that much! To be honest, I was a bit disappointed by the feedback so far. Most users didn't seem to like it much. Probably my fault, due to choosing test images which already had a lot of ringing in them, so it looked like super-xbr would ring a lot.

Well, here's another vote saying super-xbr is a lot better than NEDI and a fair bit better than jinc3ar for luma. Maybe it's because it's the first doubling algorithm I've ever tried (nnedi3 doesn't work on my laptop), but it looks extremely clean and less granular compared to jinc3ar. I haven't really noticed any substantial ringing or haloing that wasn't there with jinc3ar already. It's my new default scaler for everything now.

I can't comment on super-xbr for the initial chroma upscale, however. In principle I don't feel great about using "smart" scalers for chroma, because the "other artifacts" it can introduce might create mismatches with the unscaled luma channel, which is expecting a conventionally upscaled chroma. This might just be a fear founded on ignorance though.

I dislike NEDI mostly because it likes to destroy small text (0.5-1.5 pixel thickness letters) in credits and title screens on lower quality sources where doubling/quadrupling is most useful. The algorithm somehow makes letters lumpy and can introduce a patchiness in their luminance, SuperRes or no.

[Barnahadnagy]

Originally Posted by madshi

I'm glad you like it that much! To be honest, I was a bit disappointed by the feedback so far. Most users didn't seem to like it much. Probably my fault, due to choosing test images which already had a lot of ringing in them, so it looked like super-xbr would ring a lot.

+1 for S-XBR. Its pretty awesome quality for the performance. You can say NNEDI is better all you want, I still wont be able to run it .

[madshi]

Originally Posted by RainyDog

Have you looked at Adaptive Sharpen madshi?

No, I wasn't aware of that one. I might give it a try next weekend.

Originally Posted by iSunrise

While that may be true, the real-world results with madVR are the opposite though, thats why LL was only recommended together with Catmull-Rom downscaling. At least that's what we concluded after we did several tests years ago and e.x. 6233638 and cyberbeing were two people that also strictly recommended against it. LL on some downscalers was full of artefacts compared to gamma light.

Are you talking about:

A) LL adding ringing artifacts? Or:
B) LL changing the brightness of the original source?

- Your recent posts were about A).
- Your screenshots showed A).
- aufkrawall talked about B).
- Your post is a reply to mine, replying to aufkrawall about B).
- 6233638's and cyberbeing's LL scaling recommendations are because of A).
- AFAIR, neither 6233638 nor cyberbeing have ever complained about B).

So far I've not seen a single screenshot yet supporting B). So what is it you're talking about? A) or B)? If it's A) then your post made no sense because aufkrawall and I have been discussing B). If you were now talking about B) then you mentioning 6233638 and cyberbeing made no sense because their scaling recommendations were motivated by A) and had nothing to do with B) at all.

Please don't misunderstand me. I appreciate your feedback very much. Having feedback that differs from other feedback can be helpful to find the best solution. But please don't mix different things together, or use arguments for one thing (A) while talking about a different thing (B). It's very frustrating for me to discuss that way, because simply just explaining how things are mixed up costs a lot of time and serves no real purpose.

If you do think that B) exists then please post some screenshots where LL downscaling modifies the brightness of the source (while not using LL doesn't). But don't use 6233638 and cyberbeing in that discussion then, because referencing them only makes sense when talking about A).

Originally Posted by har3inger

Well, here's another vote saying super-xbr is a lot better than NEDI and a fair bit better than jinc3ar for luma. Maybe it's because it's the first doubling algorithm I've ever tried (nnedi3 doesn't work on my laptop), but it looks extremely clean and less granular compared to jinc3ar. I haven't really noticed any substantial ringing or haloing that wasn't there with jinc3ar already. It's my new default scaler for everything now.

I can't comment on super-xbr for the initial chroma upscale, however. In principle I don't feel great about using "smart" scalers for chroma, because the "other artifacts" it can introduce might create mismatches with the unscaled luma channel, which is expecting a conventionally upscaled chroma. This might just be a fear founded on ignorance though.

I dislike NEDI mostly because it likes to destroy small text (0.5-1.5 pixel thickness letters) in credits and title screens on lower quality sources where doubling/quadrupling is most useful. The algorithm somehow makes letters lumpy and can introduce a patchiness in their luminance, SuperRes or no.

Thanks for your feedback.

Originally Posted by Barnahadnagy

+1 for S-XBR. Its pretty awesome quality for the performance. You can say NNEDI is better all you want, I still wont be able to run it .

Cool, thanks.

[aufkrawall]

Originally Posted by madshi

Thanks for the sample. I'm not sure I can reproduce the aliasing problem, though. Using super-xbr for doubling, followed by FineSharp in upscaling refinement with rather high thinning setting. Can you post a screenshot (preferably without SuperRes) where the aliasing is visible, and maybe mark it in MS Paint with a red arrow or something? And please let me know which exact upscaling/processing options you're using. Thanks!

Thinning 0:


Thinning 0.05:


Strength: 1
Chroma: Jinc3AR
Luma: NNEDI3 64
SuperRes (mild custom values)
Should be independent from scaling settings, but in doubt I always use NNEDI3 to have the cleanest result.

Well, thinning value of 0.05 is rather high, but apart from strength it'S still doing the same as with 0.02.
So, what is the point of thinning? I can't observe a positive effect and it seems to be potentially harmful.
Unfortunately, I forgot what I did to see the aliasing also with lower values in combination with SuperRes.

Originally Posted by madshi

Scientifically it should be the other way round, at least when downscaling. Do you have some samples?

As I said, I will try to create one in the next days.
Some pages back I posted an image from Wikipedia and also fonts become harder readable.

Originally Posted by madshi

No, there is not going to be a LL option for sharpening, sorry. There are way too many options in madVR already. At some point I'll remove several (many?) of the current options. (Of course I'll ask for feedback first.)

You also offer different options for deinterlacing for animation and filmed content, I don't see any difference in the situation...
Btw: Also game-captures profit a lot by FS due to chroma subsampling blur and they often show structures that aren't common and thus LL shouldn't be used with them neither.
You really make FS less usable with such content if you keep the option removed.
Already two real world cases...

[toniash]

Originally Posted by RainyDog

Have you looked at Adaptive Sharpen madshi?

It's very good and ONLY ONE KNOB to adjust!

[StinDaWg]

Super-xbr reminds me of running NNEDI3 with a light sharpen afterwords, like SuperRes. I quite like it.

I assume Super-xbr shares code from SuperRes due to both the look and the pixel shift it creates?

With clean sources it looks quite sharp and nice. I think I may actually prefer it over NNEDI3 for blu-rays and other clean video.

The GPU hit is very minimal compared to NNEDI3 too.

Good job!

Now I just need to figure out if super-xbr or NNEDI3+SuperRes (after upscale) produces a better overall picture.

[RyuzakiL]

Originally Posted by nevcairiel

That is just false. Both LL on and LL off will cause problems in certain content.
How do you think we arrived on the decision for LL on at the first place (and incidentally noone objecting)? Because someone thought LL on looked better? Exactly!



If you run a CrossFire or SLI setup, you better know how to setup a profile to disable it, since its not going to work with everything.
Because if you don't, your in for a world of hurt.

Just an update, making a profile of Madvr on AMD CCC and set the crossfire mode to disabled, doesn't work at all.

I'm Using 2X AMD HD7850

If I use NNEDI3 128 Neurons in Chroma Upscaling crossfire still turns on and I had a lot of dropped frames, i thought AMD Cards runs fine using NNEDI3 compared to NVIDIA's. But when I use Super-XBR, Crossfire doesn't turns on whether i make a profile of Madvr or Not on AMD CCC. This is coming from a fresh clean install of OS + Latest Beta Drivers from AMD.

it seems AMD cards wasn't suited using NNEDI3 though i read somewhere that AMD's R9 270 can run NNEDI3 128 Neurons just fine. So wth? seems to me it's a driver conflict.

So for now I'll just keep on using Super-XBR and hope for further development for this algo., It's lightweight and Picture Quality is on Par with NNEDI3 and Crossfire isn't turning on when using Super-XBR so that's a plus on my book.

[Eyldebrandt]

Originally Posted by madshi

Thanks for your feedback - good 2nd post on the forum!

Yes, "thinning" is "xstr". However, "repair" is an algorithm tweak implemented by -Vit- when converting the AviSynth script to GPU pixel shaders. Basically "repair" tries to avoid/undo the aliasing problem that occurs when using "thinning". So your experience with the AviSynth script introducing aliasing when using "thinning" needs to be re-evaluated. There is still some aliasing, but it's less a problem compared to the AviSynth script. I'm still playing with the proper "repair" value. The "cstr" option is set automatically by madVR, depending on which strength you've selected.

Yah, i've figured it out, because a simple rewrite of cstr with a factor of 1.0 would have result of a ton of blur :d
For that matter,i will let a 2nd chance to thinning, to see where it goes.

Originally Posted by madshi

I'm not a big fan of LumaSharpen. It is definitely quite fast, that's its big advantage. But it still introduces quite noticable ringing. At some point I'm probably going to replace LumaSharpen with a slower alternative, something similar to LimitedSharpenFaster, which doesn't ring as much. But for now I'll probably keep LumaSharpen because it's so nicely fast.

Imo, I would have seen LSF instead of Finesharp. For the reason I think that lumasharpen and LSF don't do the exacts same things.

But FS had the advantage to be already wrote as a shader.

If you want something like LSF and something fast, maybe the LSF adaptation by Emmanuel Piat could interest you.
It's LimitedSharpenfaster amputee of most of his fonctions to let a simple sharpener in state, very effective, extremely fast, and with high artifacts control.
Let know if that what you search.

Originally Posted by madshi

I'm glad you like it that much! To be honest, I was a bit disappointed by the feedback so far. Most users didn't seem to like it much. Probably my fault, due to choosing test images which already had a lot of ringing in them, so it looked like super-xbr would ring a lot.

Maybe.
But from what i can note in comparison of NNEDI3 w/e neurons and that in any settings, quality is equivalent, and performance gap is impossible to catch up.
So, to my opinion, unless you find an s-xbr/nnedi3 killer, all in on s-xbr.

Originally Posted by madshi

Who's doing the downscaling here? The GPU driver, I suppose? Does this really look better than straight 1080p playback? Oh well, SuperRes doesn't activate for 1:1 playback, maybe that's the reason?

Yep, GPU driver.
It's an application with only madVR of the manual supersampling method.
Initially, upscaling (and other process, like deblocking, debanding, sharpening, etc...) was made in avisynth, sometimes with multiples layers, then madVR was downscaling to the final resolution.

Now, willingly or not (not i suspect :d ), the same processing is made possible with only madVR, which is a very better way.

And yes, that look better, it's the exact same process than downsampling in videogame. Less aliasing, a more natural look pictures, more clarity, etc...

I finally found the cause of those 10-15 stutters per 45 minute video - Direct3D 11. It just does not work well for me and I have no idea why... Whether I tick the Sync Every Frame option or not makes 0 difference. Adjust all other settings does nothing.

[PetitDragon]

Originally Posted by Eyldebrandt

Yep, GPU driver.
It's an application with only madVR of the manual supersampling method.
Initially, upscaling (and other process, like deblocking, debanding, sharpening, etc...) was made in avisynth, sometimes with multiples layers, then madVR was downscaling to the final resolution.

Now, willingly or not (not i suspect :d ), the same processing is made possible with only madVR, which is a very better way.

And yes, that look better, it's the exact same process than downsampling in videogame. Less aliasing, a more natural look pictures, more clarity, etc...

Haha, finally I see the method as exactly the same as mine. I doubt it could be made with only madVR tho, due to very heavy GPU load.

[lanzorg]

I'm not a big fan of LumaSharpen. It is definitely quite fast, that's its big advantage. But it still introduces quite noticable ringing. At some point I'm probably going to replace LumaSharpen with a slower alternative, something similar to LimitedSharpenFaster, which doesn't ring as much. But for now I'll probably keep LumaSharpen because it's so nicely fast.

I don't like LumaSharpen and my laptop can't handle FineSharp.
I used LimitedSharpenFaster with Avisynth before and it could be really great to get it inside madVR.

[yukinok25]

Originally Posted by madshi

Do you see any problems with repair set to 1.0? If so, I'd love to see a screenshot. I was planning to remove the repair option and set it fixed to 1.0. So if you found a situation where repair 1.0 is harmful, it would be good to know.

Actually after playing with the repair setting for a couple of days on different video sources, I have to say that the image looks better at 1.0.
Still mode 3 is the best option for me thus.

At this point I think its good to set the repair fix at 1.0.

Questions out of topic: when I choose DX11 my PC fan ( I7 4700QM 970M GTX) spins like crazy. Do I have to consider it normal?

[AngelGraves13]

I've been using Jinc for Chroma and Image Upscaling for a while now. Is there any reason to use Image Doubling with super-xbr for both chroma and luma? I'm not seeing much of a difference.

Rendering times are below 15 ms on average with Jinc and about 20 ms with super-xbr.

[Warner306]

Originally Posted by AngelGraves13

I've been using Jinc for Chroma and Image Upscaling for a while now. Is there any reason to use Image Doubling with super-xbr for both chroma and luma? I'm not seeing much of a difference.

Rendering times are below 15 ms on average with Jinc and about 20 ms with super-xbr.

Image doubling is only useful for resolutions where the image is doubled (e.g. SD -> 1080p or 1080p -> 2160p). As for chroma upscaling, super-xbr or NNEDI3 should be better, but it will likely be difficult to determine the difference because the chroma layer is inherently soft and difficult to differentiate.

So, the impact of image doubling should be very apparent for scaling factors 2x and larger but less apparent for all other resizes.

[mindz]

I think alot of newbies in video processing like me are a bit lost (or maybe just me?) in when the new processing methods take place, and thus i havent used them - because i dont know what it will do to the picture and when.

I have three questions to clarify it for me.

1. What is the difference between FineSharp, LumaSharpen and SuperRes?
2. When are they applied, one pre scaling and other post scaling?
3. Will the processing methods work when you use DXVA for scaling, or is everything then ignored?

[iSunrise]

Originally Posted by madshi

...But in the left half of the squirrel test pattern, actually there were less ringing artifacts with LL turned on compared to off. So I'm not sure if the situation is that simple as you make it sound.

Somehow I really begin to think you are looking at different images than I am. With the shots I provided there is no way that anyone could argue that LL has less ringing and less artifacts than no LL. It is not even close. I have even looked at the shots on different screens (even at work) and there's no way in hell that the no LL shot has more ringing than LL.

finesharp with no LL (no additional ringing):


finesharp with LL (heavy additional ringing):


If you refer strictly to the diagonal lines test, please zoom into your own images and show me that there is actually less ringing, because I could not reproduce that. LL always produces darker edges with more heavy ringing. I suspect that's also why you and maybe some others probably prefer the LL look. When you look at the shots above quickly one after another, you can see that this is a very deceiving thing that is happening. The picture might be interpreted to look "better" for some, but the picture is just heavily altered with LL.

Originally Posted by madshi

Are you talking about:

A) LL adding ringing artifacts? Or:
B) LL changing the brightness of the original source?

I am talking both about A) as well as B), like I already explained. Not only does the finesharp LL add very heavy ringing artefacts (see the shots above) to the image, but it also darkens the image (on the squirrel image above, see the long hair at the left side, close to the image border, it will get darker). Also, I wrote several times that there is a certain "blackness" that consumes some darker image details, just look at all the square borders in the following examples. These are the original shots that I provided again:

finesharp with no LL (no additional ringing, darker picture details are still intact, therefore the picture is obviously brighter (look at the object edges)


finesharp with LL (heavy additional ringing, darker picture details are crushed, the ringing/borders consume some object edges):


The edges of the squares get completely destroyed by the heavy ringing.

To sum up my findings:
The higher the upscaling factor (the more upscaling is needed), the heavier the added ringing artefacts of finesharp LL become and they completely destroy image detail with it. Even if you set repair to 1.00 (see above). Possible solutions: Either find a way to remove the heavy ringing with finesharp LL, which doesn't destroy picture detail or please consider changing the defaults.

And if you really insist I can do countless other examples, which show exactly the same kind of problems, but that will have to wait for the weekend.

[Braum]

Originally Posted by mindz

I think alot of newbies in video processing like me are a bit lost (or maybe just me?) in when the new processing methods take place, and thus i havent used them - because i dont know what it will do to the picture and when.

I have three questions to clarify it for me.

1. What is the difference between FineSharp, LumaSharpen and SuperRes?
2. When are they applied, one pre scaling and other post scaling?
3. Will the processing methods work when you use DXVA for scaling, or is everything then ignored?

Check this thread out, it's very useful and you'll find the answer to (most of) your questions

Madvr options explained by Asmodian
https://www.doom9.org/showthread.php?t=171787

1 :

FineSharp: A sharpener originally by Didée that attempts to keep local energy close to the original.

LumaSharpen: Blurs the original pixel with the surrounding pixels and then subtracts the blur.

SuperRes: This is a post process method. From Shiandow
Calculate an initial guess (using the configured upscaler)
Downscale and calculate differences with original image.
Scale those differences to the final size (medium = SoftCubic50, high = Jinc3)
Improve guess by:
Softening the image
Subtracting differences with the original image
Sharpening (optional)
Removing aliasing (optional)
Removing ringing (optional)
Repeat steps 2-4 several times.

2 : If you activate Finesharp and Lumasharpen in image enhancements then it's applied before any scaling.
If you activate Finesharp, Lumasharpen and SuperRes in upscaling refinement, then it's applied after chroma upscaling, image downscaling, image doubling and image upscaling. Upscaling must be active (you can check with crtl+J) in order to have upscaling refinement to take effect.

3 : If you use DXVA and there is no need for scaling then only image enhancements is applied.
If you use DXVA and there is a need for scaling, then image enhancements and upscaling refinement are applied.

Sorry for my english.