madVR - high quality video renderer (GPU assisted)

[Paladin77]

My personal taste is:
For low end Systems (laptops for instance): bicubic 75 on all scalers.

high end Systems: Bicubic 75 in chroma spline 4 in luma.

[Boltron]

So I have a question for you guys who watch blurays on high end players and HTPCs with madVR. Which produces higher playback quality?

I have a fairly large bluray collection which I have ripped (transcoded, 1:1) to a NAS. I do not have a bluray player at all other that the one on my computer where I do the ripping. My movie playback quality is awesome... as far as I know however I have never watched a movie played on a stand alone bluray player. I just can't help but wonder, would a higher end bluray player possibly provide better PQ?

I am sure that in this community, many of you guys would have compared the two. So, anyone want to share your thoughts?


My setup: Win7, MC17, LAV Filters, madVR, ATI 7770. HDMI to Denon 4311CI receiver, finally to my TV an LG LE8500 (IPS with LED backlighting, full array, local dimming) and the set has been calibrated.

[ajp_anton]

Bug:
Open video, stop video, go to fullscreen (I use exclusive mode), hit play. The video will occasionally start playing in a small window in the top/left corner with the size of the previous non-fullscreen window with the rest of the screen black.
Haven't narrowed down exactly how to do it every time, but it happens about 2/3 of the time when I do it.

MPC-HC, Win 7 x64, Nvidia Geforce 9300.

[6233638]

So I have a question for you guys who watch blurays on high end players and HTPCs with madVR. Which produces higher playback quality?Just like most high-end AV equipment, high-end Blu-ray players are mostly snake oil and don't offer anything over low-end Blu-ray players as far as AV quality is concerned.

Advantages of using a HTPC:
Gets you better chroma upsampling than most players
Gives you better picture controls (3DLUT calibration etc.)
Gives you the option to play back DVDs at 24p to avoid film judder (do any stand-alone players even offer this?)
Gives you more flexibility with DVD upsampling (and better quality)
Lets you force film/video deinterlacing rather than hope the player gets it right
ReClock 50Hz discs to 24p
ReClock 24/1.001 discs to 24p
Upsample audio
Start the film directly, no load times, advertising, menus or other crap to wait through
It's a moving platform. You buy a Blu-ray player and image quality is not going to change. HTPC image quality has steadily improved over time as development of things like madVR and LAV Filters has progressed
Instant access to your library. If you're ripping discs (hopefully not compressing them) then you can instantly start watching something, rather than find the disc for it. Also means you don't have to store your library of discs within arm's reach of the TV and can keep them out of sight
You can do more. My HTPC is my computer, my blu-ray player, my media streamer, and my games console. I only need one device hooked up to my TV and it's kept out of the way in another room (HDMI cable routed through the walls)
What you lose:
Simplicity. You need to do some initial configuration, and I've yet to see a way to automate things to make it as easy as simply putting a disc in a stand-alone player. (though I suppose if you're playing back off the disc on a HTPC, it can be that easy)
High bit-depth output. HTPCs are limited to 8-bit output with madVR right now, stand-alone players offer 10+bit output. (though many will "output" higher bit-depths than they are actually operating at internally) With madVR's 16-bit internal precision and dithering, I'd argue that it can still produce better results. It depends on the display device and the player you're comparing to. And hopefully at some point, madVR will output 10-bit or greater too.
Bitstreaming. PCs do not run their video and audio hardware off the same clocks, and so it's not possible to have them perfectly synced up. You either have to sacrifice bitstreaming (but if you're upsampling, this is moot) to have audio synced to the video without any dropped/repeated frames, or suffer dropped/repeated frames to have bitstreamed audio. This is a non-issue in my opinion, but some people love that light on their AVR and insist on wanting to bitstream.
Power efficiency
Noise level. Even a quiet PC is usually louder than a modern Blu-ray player, though I don't keep my tower in the same room as my display
3D support. I don't think there's any way to output 3D when running MPC-HC and madVR (though you can always have another player I guess?) I've had a 3DTV for close to two years now and I never use 3D any more so I don't really care.
In my opinion, the HTPC is a vastly superior solution, but it's not right for everyone.

[Hprd]

Hprd, are you able to produce a screenshot for this "bleeding" effect?

Yes, scaling to "fullscreen" in windowed mode shows the difference. Although it's obviously not terribly huge, "MN" is slightly cleaner/less blurry than "BC60":

MN:

http://thumbnails57.imagebam.com/20505/77cd47205042506.jpg (http://www.imagebam.com/image/77cd47205042506)

BC60:

http://thumbnails57.imagebam.com/20505/b75b1e205042503.jpg (http://www.imagebam.com/image/b75b1e205042503)

You can also see this on 6233638's comparisons with the text on the previous page. Anything sharper than MN is a little less defined, and although SC100 is a little cleaner, the colors are too dull (as was stated by him as well).

[Neeto]

What you lose:
[list] Simplicity. You need to do some initial configuration, and I've yet to see a way to automate things to make it as easy as simply putting a disc in a stand-alone player. (though I suppose if you're playing back off the disc on a HTPC, it can be that easy)
Noise level. Even a quiet PC is usually louder than a modern Blu-ray player, though I don't keep my tower in the same room as my display


Both Simplicity and Noise level are fixable.
Using Windows Media Centre, MyMovies & MPC as an external player I've been able to make it so simple that even my family can use it - ripped or off-disk playback - only one remote - no reconfiguring of inputs :D
I've managed to have a completely noiseless HTPC using a HFX case - expensive but worth it to improve acceptability :D

[ryrynz]

[ryrynz]

[innocenat]

I am having problem deinterlacing MBAFF with LAV Filter (w/o software deinterlacing) and madVR, both latest problem. Is there anything I should set?

[renq]

Power efficiency
Noise level. Even a quiet PC is usually louder than a modern Blu-ray player, though I don't keep my tower in the same room as my display
Define a quiet PC. It's quite easy to build a quiet, if not silent PC, although it's more expensive. Usually the loudest thing is the HDD- use an ssd and have a NAS/home server and then you're left with only fans from video card, cpu and PSU.
Modern low-end CPU (2-core Sandy, APUs) would be powerful enough and able to be run (almost completely) passively.
PSUs are a bit trickier, cheaper ones usually have loud(ish) fan, but one could probably get away with changing or completely removing the stock fan OR getting a proper PSU with proper fan control (some don't start the fan under 20% load).
Vid card- it's quite easy to find a powerful enough vid card that's passive

[madshi]

Here's a new test build:

http://madshi.net/madVRantiringing3.rar

Same as last test build, just with a slightly modified anti-ringing algorithm. I hope that directional artifacts are noticeably reduced. But that comes at the cost of some ringing which isn't removed. So please test:

(1) Are there still any sort of artifacts caused by the anti-ringing algorithm? Or is it artifact free now?
(2) Is the anti-ringing itself still (nearly) as effective as before?

[digitech]

Here's a new test build:

http://madshi.net/madVRantiringing3.rar

Same as last test build, just with a slightly modified anti-ringing algorithm. I hope that directional artifacts are noticeably reduced. But that comes at the cost of some ringing which isn't removed. So please test:

(1) Are there still any sort of artifacts caused by the anti-ringing algorithm? Or is it artifact free now?
(2) Is the anti-ringing itself still (nearly) as effective as before?

Thanks for keeping us updated and fresh madshi, a big thank you for giving us this wonderful piece of software art that MadVR is :)

[secvensor]

Here's a new test build:

http://madshi.net/madVRantiringing3.rar

Same as last test build, just with a slightly modified anti-ringing algorithm. I hope that directional artifacts are noticeably reduced. But that comes at the cost of some ringing which isn't removed. So please test:

(1) Are there still any sort of artifacts caused by the anti-ringing algorithm? Or is it artifact free now?
(2) Is the anti-ringing itself still (nearly) as effective as before?

Thanks for it!

[zerowalker]

what settings should i use to speed up the Seek, i currently use EVR as i totally love the Seek, it´s so fast, but MadVR is so damn slow (not complaining about the Renderer itself!).
So i wonder if there is a way to speed it up?

I got a pretty fast PC overall if that changes anything:)

[BetA13]

Here's a new test build:

http://madshi.net/madVRantiringing3.rar

Same as last test build, just with a slightly modified anti-ringing algorithm. I hope that directional artifacts are noticeably reduced. But that comes at the cost of some ringing which isn't removed. So please test:

(1) Are there still any sort of artifacts caused by the anti-ringing algorithm? Or is it artifact free now?
(2) Is the anti-ringing itself still (nearly) as effective as before?


:goodpost:

nice.. Well i just updated it and from the first little testing with some different videos, (xvid, mkv,flv, etc) it all looks a little better...the Artefacting is deffo a little less, i can clearly see it on the History channel Sign..its Red and looks damn good to me with lanczos 3 + anti ringing..

i do, on the other hand see some aliasing..But could be the movie doh, so i need to test that a little bit more...

But all in all it looks better a bit...ALso same gpu/cpu usage..

greetz BetA

EDIT..

also i have a little question :)

I use fullscreen Exclusive mode..And when i go fullscreen, (no matter what kind of movie, mkv, xvid, high low bitrate) i get Presentation Glitches..1 - 2 in 30 secs..Rest is fine..no drops or skipped frames...Just those Presantation Glitches are itching me a little...i tried a lot now..using dx11 render path, or teh other things sugessted in ur settings menue..

maybe someone could point me into the right direction so i kno wwhere to start to look for probs..Also, What exactly is a "presentation Glitch" ?

Thanks..

[Mangix]

what settings should i use to speed up the Seek, i currently use EVR as i totally love the Seek, it´s so fast, but MadVR is so damn slow (not complaining about the Renderer itself!).
So i wonder if there is a way to speed it up?

I got a pretty fast PC overall if that changes anything:)

Changing the resizers to Bilinear will speed things up really nicely.

[madshi]

what settings should i use to speed up the Seek, i currently use EVR as i totally love the Seek, it´s so fast, but MadVR is so damn slow (not complaining about the Renderer itself!).
So i wonder if there is a way to speed it up?
Which splitter and which video decoder are you using? Can I have a sample which shows faster seeking with EVR compared to madVR? Do you have the madVR option "rendering\general settings\delay playback start until render queue is full" activated? If so, try deactivating it.

nice.. Well i just updated it and from the first little testing with some different videos, (xvid, mkv,flv, etc) it all looks a little better...the Artefacting is deffo a little less, i can clearly see it on the History channel Sign..its Red and looks damn good to me with lanczos 3 + anti ringing..

i do, on the other hand see some aliasing..But could be the movie doh, so i need to test that a little bit more...

But all in all it looks better a bit...ALso same gpu/cpu usage..
Finally some test results with the new build - thanks!

That aliasing, does it go away if you disable the anti-ringing filter? If not, then it's not the fault of the anti-ringing filter.

I use fullscreen Exclusive mode..And when i go fullscreen, (no matter what kind of movie, mkv, xvid, high low bitrate) i get Presentation Glitches..1 - 2 in 30 secs..
I guess you're an NVidia user? This problem mostly affects NVidia, unfortunately. For me using the DX11 render path in combination with the "limit render times to avoid glitches" helps quite nicely.

[Netuser]

Hi,

I have pass to madvr and Lav filter setup and I have amazing quality and no more problem with Full HD playback :D

Thanks so much madshi :)

But I have 2 problem and I don't know if in futur version of madvr you can implement them.

1- I watch a lot of Vob of music videos and often old videos are 4/3 and boxed, before with ffdshow I use to use a script called LameAutoCrop which automatically crop the black border so that I can watch the clip in full screen and don't have to zoom each time. Is there anyway you can add this option ? or a way to adopt it ?

2- problem is that I can not take screenshot in MPC-HC when I use madvr :( i have to switch to other output option in MPC to do so. are you planning to add this option in future ?

Thanks a lot :)

[BetA13]

Finally some test results with the new build - thanks!

That aliasing, does it go away if you disable the anti-ringing filter? If not, then it's not the fault of the anti-ringing filter.


I guess you're an NVidia user? This problem mostly affects NVidia, unfortunately. For me using the DX11 render path in combination with the "limit render times to avoid glitches" helps quite nicely.


hy there,

Thanks for your answer, and yes i am Nvidia user, hehe..
Im gonna test what u recommend for shure..

And you where also right on teh aliasing..it was the movie..it was an recording of an VHS tape and pretty old...Other Movies work fine (recorded Stuff from TV and STreams)...I like the anti ringing filter...
ALso im a little bit more away from my HDTV when i watch movies so the picture gets even better..Thats why noone can say these are teh best settings for Madvr and all..cause it has also a lot to do how far away or near the TV you are..

On close look with my CRT Monitor i can spot the differences..Still, im gonna keep lanczos - 3 - plus anti ringing, from my view its better then the Soft one..i can clearly read, picture is sharp and i have less or no artefacts.(depends on the source..BlueRays are awesome and clear as they can get..)

Anyway, THX for the Answer...gonna report back about the presentation glitches...

greetz BetA

[madshi]

And here comes the next test build:

http://madshi.net/madVRantiringing4.rar

This time anti-ringing is available for all upscaling algorithms, luma and chroma. Also anti-ringing should be quite a bit faster for NVidia users than before. I'm interesting in the following
FEEDBACK:

(1) Can you find any real-life situations where activating the anti-ringing filter negatively impacts image quality?
(2) How much worse is the performance with anti-ringing turned on vs. off?
(3) Is there any reason *not* to use anti-ringing?

Depending on your feedback, I might eventually always force on anti-ringing (for those algorithms which need it), and only allow to deactivate it via a "trade quality for performance" option. Thoughts on this?

[noee]

Testing with some run-of-the-mill SD sources => 1080, and a couple of animation movies (toy story, cars):

1. No, not with my setup, but I can't really see the difference on the chroma options.
2. With Luma upscaling only: Render time as reported by OSD, With AR: 7.35ms Without AR: 6.55ms. If I turn on chroma AR (bicubic 60), it adds another ms (8.35ms).
3. Not from where I sit.

My testing just consists of a handful of movies and I just watch them at certain points. I don't pause playback and all that micro-level analysis. I go by "feel". To my eyes, luma up with Bicubic 75 AR and Spline3 AR both look better (cleaner, sharper) on these particular sources. (HD6570, CCC12.7b)

and only allow to deactivate it via a "trade quality for performance" option. Thoughts on this?

Sounds like a good idea.

[Ver Greeneyes]

I have a question with regards to the color correction built into madVR. My monitor is calibrated using an X-Rite ColorMunki Photo, using a LUT together with a v4 ICC profile. If I setup my monitor in madVR so that it's set to "this display is already calibrated", will it use the ICC profile assigned to my monitor? Or do I somehow need to convert my ICC profile to a 3DLUT file? The LUT itself should already be loaded into my GPU.

Oh, and should I "enable gamma processing" on the color & gamma tab? Would it do anything if it's set to the same gamma as my ICC profile (2.20)?

If there's a readme with detailed descriptions for each setting out there, I'd love to read it.

[turbojet]

(1) Can you find any real-life situations where activating the anti-ringing filter negatively impacts image quality?

It blurs quite a bit, as a detail nut, hence the negative bicubic blur request earlier, this is concerning. f3kdb filter in my setup eliminates most ringing without blurring, in fact it looks a little sharper.

(2) How much worse is the performance with anti-ringing turned on vs. off?

nvidia 9500GT, bicubic 100, SD and HD source is same load
none - 18% GPU
chroma only - 20%
luma only - 30%
chroma+luma - 32%

(3) Is there any reason *not* to use anti-ringing?

heavier load, softer picture

Depending on your feedback, I might eventually always force on anti-ringing (for those algorithms which need it), and only allow to deactivate it via a "trade quality for performance" option. Thoughts on this?

Would be nice to keep chroma and luma separate. I'll use it for chroma but not for luma, if given the chance. The blur in the chroma isn't noticeable and most of the ringing seems to be in the chroma.

Thanks for adding anti-ringing to all resizers, it was worth the wait.

[06_taro]

If using a "trade quality for performance" option, I still prefer a resolution based method as I suggested before (http://forum.doom9.org/showthread.php?p=1585533#post1585533).

Upscaling 720p to fullscreen on my 1366x768 screen does not cause any noticeable ringing, unless I set tap to 8, but upscaling 480p does, even with 3-tap. On my HD3000, with 4-tap lanczos and AR disabled, the rendering time is about 25~26ms for 720p and 13ms for 480p, while with AR(luma only) they increase to 30~32ms / 16ms, and 33~35ms / 18ms with AR(luma&chroma), so frame drops may sometimes occur on 720p30 contents but not on 480p contents. That's why I prefer to disable it for HD but enable it for SD. Maybe an option like this could make it more convenient:

anti-ringing algorithm:
- always on
- only when enlargement factor >= 1.5 (or maybe other value) or above
- always off

Besides, anti-ringing algorithm seems also faster on Intel HD3000 too in madVR-AR4 compared to madVR-AR3, thanks for your great work!

[fairchild]

As always, thanks Madshi for continuing development of MadVR.

I personally agree with turbojet that if you do decide to make AR On, being the standard, then having an option to separate it by chroma and luma would be great. I currently am using the settings below and the picture quality has never looked better:

Chroma: Bicubic75 (AR on)
Luma: Bicubic75 (AR off)

[madshi]

It blurs quite a bit
Can I please have a sample, or maybe even 2 or 3, where this blurring is visible?

[Hypernova]

(1) Can you find any real-life situations where activating the anti-ringing filter negatively impacts image quality?
(2) How much worse is the performance with anti-ringing turned on vs. off?
(3) Is there any reason *not* to use anti-ringing?

Depending on your feedback, I might eventually always force on anti-ringing (for those algorithms which need it), and only allow to deactivate it via a "trade quality for performance" option. Thoughts on this?

Finally got some time to test this out. Please note that I've only test it briefly.

(1) From a couple files I tried, no.
(2) On Intel HD4000, 1080p->2560x1600 scaling, the rendering time are ~32 when on vs. ~28 when off. The video is still smooth so it's ok
(3) No reason from what I see so far

I agree with AR on by default, but I would still keep the deactivation checkbox at where it is now. I think it make more sense there.

[mbordas]

It would help if there were some agreed upon test clip so you could compare reactions from different users to the same source material. I've tried the new version and it looks great to me, and the rendering times are low on my gt 440 (6-9 ms). But I don't know what the standard is to look for, so can't really give an informed opinion.

[turbojet]

Can I please have a sample, or maybe even 2 or 3, where this blurring is visible?

http://180upload.com/tosenohlwins

1= 720x304
2= 1280x544
both= chroma+luma anti-ringing
chroma = chroma anti-ringing
luma = luma anti-ringing
none = no anti-ringing

The differences are more noticeable at 24 fps but unsure how to show that. The results are the same for every source and frame compared.

[chros]

I've tried the new version and it looks great to me, and the rendering times are low on my gt 440 (6-9 ms).
I have noticed that, too (using Lanczos 4 tap). Is that mean that you're (madshi) using the older algorithm or the new had speeded up?

And doesn't the antiringing option has any meaning at downscalers?

Thanks for your effort!

[madshi]

@turbojet, screenshots don't help me, I need *samples*. If I can't reproduce the problem on my PC, I can't fix it.

[Owyn]

Please fix the resizing crash, it crashes pretty a lot T_T
*sent all the reports to mail*

[turbojet]

There's no problem, so to speak, it does what it's intended to do well but a side effect is it blurs the picture, too much in my opinion to use it on luma (except unbearable ringing sources). Thus a negative impact on image quality, maybe you were looking for image artifacts or issues produced by anti-ringing. If so, I misunderstood the question.

I hope the settings stay as they are now, more control is always better.

Also do you still plan on back reading the old posts (much offtopic) to find requests/issues or do you want us to post them again?

[ryrynz]

Also do you still plan on back reading the old posts (much offtopic) to find requests/issues or do you want us to post them again?

He'll definitely read back through them and reply, he's just playing with the anti-ringing at the moment.


Lanczos 3 is using 18% less GPU which now makes it useable for me as there aren't any dropped and delayed frames now.

Enabling anti-ringing with lanczos on luma for a 624x352 video upscaled to 1920x1080 is using an extra 15-17% GPU on my Intel HD 3000. Anti-ringing enabled on chroma is only an extra few percent.


Some image quality analysis shows build 4 being quite superior to build 2. However as Turbojet pointed out it's reducing detail enough that I won't keep it enabled.

Some screenshots of build 2 & build 4 vs no anti-ringing using Lanczos 3 tap luma. We're up close on Haruhi Suzumiya's hair and eyebrow.

Build 2
http://img339.imageshack.us/img339/1456/arbuild2.th.png (http://imageshack.us/photo/my-images/339/arbuild2.png/)

Notice the light patches, the old code destroys a fair amount of dark detail we'd rather keep, the eyebrow gets lightened considerably. Also notice the tip of the hair.

Build 4
http://img43.imageshack.us/img43/373/arbuild4.th.png (http://imageshack.us/photo/my-images/43/arbuild4.png/)

A hell of a lot better than build 2 but still very smooth.

No anti-ringing
http://img221.imageshack.us/img221/5789/noarg.th.png (http://imageshack.us/photo/my-images/221/noarg.png/)

Much better :)

[cyberbeing]

(1) Can you find any real-life situations where activating the anti-ringing filter negatively impacts image quality?
(3) Is there any reason *not* to use anti-ringing?
It does eliminate resizer ringing, but the luma anti-ringing also reduces fine contrast and blurs fine details significantly. Maybe you could borrow some ideas from Didée's DeHalo_alpha (http://avisynth.org/mediawiki/DeHalo_alpha) avisynth script to compensate for this. Something like a slider for Strong (current setting), Medium, and Weak anti-ringing would also probably be useful if there isn't a one-size-fits-all setting.

http://screenshotcomparison.com/comparison/140183
1280x720 -> 2560x1440 Spline36 anti-ringing VS. normal (200% crop)

http://screenshotcomparison.com/comparison/140194
1920x1080 -> 2560x1440 Spline36 anti-ringing VS. normal (200% crop)

Anti-ringing on chroma causes colors to have increased blur and desaturation compared to normal. I tested various sharper algorithms with anti-ringing in an attempt to compensate for this, but all seem to give worse results compared to normal Catmull-rom chroma, which I still find to have the most neutral colors and saturation.

screenshots don't help me, I need *samples*.
Edit: PM sent with the two samples used in the screenshot comparisons above.



(2) How much worse is the performance with anti-ringing turned on vs. off?


NVIDIA GT440 DDR5 (Overclocked)

1920x1080 -> 2560x1440 23.976fps
19ms Catmull-Rom Chroma | Spline36 Luma
26ms Catmull-Rom Chroma | Spline36 Luma Anti-ringing
28ms Catmull-Rom Chroma Anti-ringing | Spline36 Luma Anti-ringing


1280x720 -> 2560x1440 23.976fps
17ms Catmull-Rom Chroma | Spline36 Luma
23ms Catmull-Rom Chroma | Spline36 Luma Anti-ringing
24ms Catmull-Rom Chroma Anti-ringing | Spline36 Luma Anti-ringing


720x480 -> 2559x1440 23.976fps
12ms Catmull-Rom Chroma | Spline36 Luma
18ms Catmull-Rom Chroma | Spline36 Luma Anti-ringing
18ms Catmull-Rom Chroma Anti-ringing | Spline36 Luma Anti-ringing

[madshi]

Here's a new test build which will hopefully get rid of the blurring. This comes at a slight performance cost:

http://madshi.net/madVRantiringing5.rar

Is that mean that you're (madshi) using the older algorithm or the new had speeded up?
Older algorithm + anti-ringing.

Maybe you could borrow some ideas from Didée's DeHalo_alpha (http://avisynth.org/mediawiki/DeHalo_alpha) avisynth script to compensate for this.
f3kdb filter in my setup eliminates most ringing without blurring
madVR's algorithm is totally different to what DeHalo_alpha and f3kdb do. The AviSynth DeHalo scripts try to find ringing artifacts in an image and then try to remove them. madVR does not do that *at all*. madVR modifies the resampling algorithm to not ring in the first place. Which IMHO is a much better approach for clean sources. Of course the AviSynth scripts are great for sources that already have ringing in them. You need to use the right tool for the right job. DeHalo_alpha has the purpose of removing *source* ringing. madVR tries to avoid adding new ringing artifacts to the source while resampling. Totally different things. And that means there's zero chance for me to get any ideas from the AviSynth scripts because technically they work very differently.

Of course you could use a normal ringing Lanczos implementation and then follow it up with those AviSynth scripts to remove the ringing again. But that's IMHO much less effective than what madVR is doing. For your interest, here's a 4x zoom comparison (Monsters Inc. Blu-Ray):

http://madshi.net/ringing.png

I've used the default parameters for the AviSynth scripts. Maybe some tweaking of the parameters would have improved the results. But honestly, for real time playback (which is what madVR is all about), you DON'T want to contantly tweak a dozen of complicated parameters.

Anti-ringing on chroma causes colors to have increased blur and desaturation compared to normal.
Does that still happen with the new build? If so, can I have a sample and see some screenshots? I hope you're comparing the saturation of the *original image* as the reference?

Edit: PM sent with the two samples used in the screenshot comparisons above.
Thanks.

There's no problem, so to speak, it does what it's intended to do
Blurring the image is certainly *not* intended.

Also do you still plan on back reading the old posts (much offtopic) to find requests/issues or do you want us to post them again?
I commented on that when I released the first anti-ringing test build.

[cyberbeing]

The test5 build is ~17% slower, but still has issues reducing fine contrast and blurring fine details in the samples I provided you.

It would seem the anti-ringing modifications essentially remove the beneficial effects of these sharper algorithms, as a trade-off for no ringing?

http://screenshotcomparison.com/comparison/140217

Edit: There does seem to be a very minor improvement on the 1080p sample, but it wasn't noticeable until I made a 400% crop of the 1920x1080 -> 2560x1440 resize.

Spline36 Test4 anti-ringing vs Test5 anti-ringing 400% crop (http://screenshotcomparison.com/comparison/140221/picture:0)
Spline36 Test4 anti-ringing vs Normal 400% crop (http://screenshotcomparison.com/comparison/140221/picture:1)
Spline36 Test5 anti-ringing vs Normal 400% crop (http://screenshotcomparison.com/comparison/140221/picture:2)

[chros]

Thanks for the new build (v5)! I can't say anything about image quality, but performance on my 9600m GT (compared to v4):
- using Luma Lanczos3 + antiringing on 720p@24Hz is indeed slower
- but (!) using Chroma Lanczos3 + antiringing on 1080p@24Hz is much faster than v4!!! Is antiringing used in this case?

Edit: "Is antiringing used in this case?" Yes, it's used.

[madshi]

Next try:

http://madshi.net/madVRantiringing6.rar

I've reduced the anti-ringing strength a bit, which means that (especially with Lanczos8) some hints of ringing might not be totally eliminated. But it seems to improve the blurring problem. So basically anti-ringing is slightly less effective, but blurring should be noticeably reduced. Besides, speed is the same as test4 again.

[ryrynz]

6 is a hell of a lot better than 5. Still wiping out a bit too much for my tastes, especially the darker colors.

Comparison between having no anti-ringing and build 6.

http://screenshotcomparison.com/comparison/140233

All the lighter stuff is being maintained now though unlike in b5.

[cyberbeing]

Overall I'm satisfied with madVRantiringing6. You seem to be in the right ballpark, considering the 1080p sample I provided now has excellent detail retention with anti-ringing enabled, while still eliminating 99%+ of Spline36 ringing. With a few minor tweaks it will probably be perfect. As ryrynz said, there is still a bit of weakness on darkish lines, which can been seen on the below 400% crop of the 720p sample when scaled to 1440p:

http://screenshotcomparison.com/comparison/140238

This is somewhat nitpicking at this point though.

[madshi]

@ryrynz, can I have a video sample, matching your screenshot?

@cyberbeing, how about anti-ringing chroma desaturation? Is that solved, too?

[ryrynz]

@Madshi

http://www.mediafire.com/?5hlgg6e3bms8ahc

If you want to use the same frame, shift and right arrow a couple of times in MPC-HC (frame 307)

[cyberbeing]

I did a quick test and it seems improved, but not eliminated. Previously it was ~2-3% change in RGB values, now it seems to be ~0.5-1% change with Catmull-Rom. I don't really have a good sample to provide for this though. Just find something with a saturated Red and it should be noticeable. It's early morning where I am, so that's about all the testing I can right now as I need to get some sleep.

[chros]

On performance wise using Chroma Lanczos3 + antiringing on 1080p@24Hz:
- v6 is slower than v5, but a little faster than v4

Can you make faster the chroma antiringing function as it was with v5?
Thanks!

[turbojet]

Here's a new test build which will hopefully get rid of the blurring. This comes at a slight performance cost:

A lot less blurring and performance difference is insignificant here, maybe 1-2%. Still a chroma only feature for me but my priorities in video are a bit different then most in this thread. Max detail with a tolerable amount of artifacts.

madVR's algorithm is totally different to what DeHalo_alpha and f3kdb do. The AviSynth DeHalo scripts try to find ringing artifacts in an image and then try to remove them. madVR does not do that *at all*. madVR modifies the resampling algorithm to not ring in the first place. Which IMHO is a much better approach for clean sources. Of course the AviSynth scripts are great for sources that already have ringing in them. You need to use the right tool for the right job. DeHalo_alpha has the purpose of removing *source* ringing. madVR tries to avoid adding new ringing artifacts to the source while resampling. Totally different things. And that means there's zero chance for me to get any ideas from the AviSynth scripts because technically they work very differently.

Of course you could use a normal ringing Lanczos implementation and then follow it up with those AviSynth scripts to remove the ringing again. But that's IMHO much less effective than what madVR is doing. For your interest, here's a 4x zoom comparison (Monsters Inc. Blu-Ray):

I've used the default parameters for the AviSynth scripts. Maybe some tweaking of the parameters would have improved the results. But honestly, for real time playback (which is what madVR is all about), you DON'T want to contantly tweak a dozen of complicated parameters.


f3kdb's intention is to deband which is does a fantastic job, prevention of ringing is a positive side effect. I think it detects mosquito noise near edges and adds grain to fool the resizer into either blurring or treating it as grain instead of trying to get a bunch of details from it. In my tests f3kdb is more effective pre-resize then post-resize. In your pics I prefer the f3kdb to the others. MadVR has less ringing but f3kdb has the most detail.

Agree with default settings, I rarely tweak them unless it's suggested.

Blurring the image is certainly *not* intended.

Thus the side effect which is something unexpected/unintentional. Intention is to anti-ring and it does a good job of it.

I commented on that when I released the first anti-ringing test build.

Just trying to save you some hours/days. There may have been 10 issues/requests in 100+(?) pages.

[BetA13]

Hy madshi,

Thanks for ur tips, the presentation glitches are mostly gone now with dx11 path and limit rendering times.
Only one or 2 glitches in 1 - 1hour and 10min.

its totaly ok for me :)

Also the new anti ringing is really nice..rendering times stay around 5ms..(+-0.20ms)

I just have one Problem now..Before i used the DX11 render path my movies matched my HZ, i mean run @ 60FPS/hz, but now when enabled DX11 Render Patch, it doesnt work anymore, it stays at the Movies fps, no matter what i try.. Is there a reason for this? Or is it just not possible to match the hz and fps in dx11 mode?

i really kinda need this, cause it makes a good difference here for me if i play teh movies @ 23 - 30 fps or @ 60 fps...


Thanks..

Greetz BetA

[l0rdraiden]

@madshi Taking into account all the testing done with the builds would be nice to release the final version already with the best scaling settings possible for quality, or at least put them in the first post.

[sexus]

hell im all for a final version , if madshi is that far in on dev , best scaling settings included please ;) , max quality and minimum artifacts ,minimum colour change and minimum ringing and minimum aliasing, is where its at as others have mentioned, keep it up madshi you going far with madvr now , together we can reach video playback heaven wohoo!!!

p.s: after reading through the last pages i guess then it comes down to these settings? with madvr testbuild6 :confused::

chroma upscaling - lanczos3 +AR
luma upscaling - the same
luma downscaling - bicubic 75 + scale in linear light

[pie1394]

http://madshi.net/madVRantiringing6.rar

(2) How much worse is the performance with anti-ringing turned on vs. off?


C2D P8600, Ion-LE chipset , Dual-ch DDR3-1333 8GB, 1280x720p24 at 1280x720 window mode. It shows the GPU loading like this:

Chroma + Lanczos3 --> 40 ~ 41%
Chroma + Lanzcos3 + AR --> 60 ~ 61%
Luma + HW Bilinear (effectively no scaling)