Video Post-processing Help

[Nagato]

I couldn't find thread for this issue, so I guess anyone can post video post-processing problems, suggestions, etc here.

Lately, I often use post-processing for almost anime I watch; because, there's always artefact, especially banding, blocking, and flickering whether in the raws or fansubs I download, and I feel that sharpening like in the Haali's video renderer is not enough anymore.

So far, I use ffdshow and add simple avisynth filters there. But with HDTV (720p) it can't play live. Since I have quad core on my system, then, I tried writing simple avisynth script using DSS2 (using coreavc) and load it to mpc, and it play smoothly now. But writing script for each video is just kind of monotonous and lame; anyone now how to do this automatically like in ffdshow?

Then about flickering; I still cant solve this issue. I tried using the filter in ffdshow also changing my graphic/display configuration, but no luck. (my monitor is a LG LCD)

[martino]

Originally Posted by Nagato

Lately, I often use post-processing for almost anime I watch; because, there's always artefact, especially banding, blocking, and flickering whether in the raws or fansubs I download, and I feel that sharpening like in the Haali's video renderer is not enough anymore.

Sharpening option and artefacting/banding/blocking... What do those have to do with each other? Mind elaborating?

Originally Posted by Nagato

So far, I use ffdshow and add simple avisynth filters there. But with HDTV (720p) it can't play live. Since I have quad core on my system, then, I tried writing simple avisynth script using DSS2 (using coreavc) and load it to mpc, and it play smoothly now. But writing script for each video is just kind of monotonous and lame; anyone now how to do this automatically like in ffdshow?

Please give us the script that you used in ffdshow's AviSynth tab and any other filtering options that you used. Obviously you wouldn't be able to run something like FFT3DFilter in realtime, and many other filters. So you need to stick to fast solutions. I have a C2D @ 3.2GHz, and one of my profiles, named "heavy", has mplayer's accurate deblocking enabled, aWarpSharp, and hqdn3d in the AviSynth tab. It runs perfectly fine with 720p videos. You might want to take a look at the AviSynth plugin called "MT". It should speed things up a bit.

[Dark Shikari]

If there's artifacting, perhaps you should find better subbing groups to download from?

For banding: gradfun2db(), also known as the Deband option in FFDshow.
For blocking: SPP deblocking is incredibly effective, under the Postprocessing tab. I personally like the Soft Threshold option.
For flicking: DeFlicker or something maybe?

[Nagato]

Originally Posted by martino

Sharpening option and artefacting/banding/blocking... What do those have to do with each other? Mind elaborating?

Please give us the script that you used in ffdshow's AviSynth tab and any other filtering options that you used. Obviously you wouldn't be able to run something like FFT3DFilter in realtime, and many other filters. So you need to stick to fast solutions. I have a C2D @ 3.2GHz, and one of my profiles, named "heavy", has mplayer's accurate deblocking enabled, aWarpSharp, and hqdn3d in the AviSynth tab. It runs perfectly fine with 720p videos. You might want to take a look at the AviSynth plugin called "MT". It should speed things up a bit.

My bad, I worded it wrong, maybe. There's artefacting and the image is also blur.
So, besides fixing the artefacts, I also do a bit of sharpening.

Anyway, the script is look like this:

SetMTmode(2,4)
#SetMemoryMax(1024)
DSS2("F:\Path~\D.C.II S.S. ～ダ・カーポII セカンドシーズン～ 第03話 「正義の魔法使い」 (x264 1280x720).mp4")
aud=ffaudiosource("F:\Path~\D.C.II S.S. ～ダ・カーポII セカンドシーズン～ 第03話 「正義の魔法使い」 (x264 1280x720).mp4")
deblock_QED()
FastlinedarkenMOD(strength=0, thinning=80)# I know this filter is not supposed to be used for live playback, I guess. But somehow it runs smoothly on quad.
degrainmedian(mode=3, LimitY=4, LimitUV=5)#can't afford more than this with this filter chain when using ffdshow.
gradfunk(1.8)
Limitedsharpenfaster(strength=30, undershoot=3, Smode=4)
vid=last
audiodub(vid,aud)#dunno how to make the audio play, so I just add something like this. Any better method?
#delayaudio(0.10)

Edit:
@Dark_Shikari
ah, deflicker, OK, I'll try and report the result.
Results:
The flickering is still there.
My LCD is on 75Hz.

[GLH]

I have a post-processing related question/problem.

I use the CCCP and fast spp deblocking under FFDShow's post-processing options. This cleans up my .AVI files nicely. However, it doesn't seem to affect my .MKV files. Is there an option I need to change to fix that? Or do I need to use a different tool to apply deblocking to my .MKV files? Thanks!

[martino]

Originally Posted by Nagato

My bad, I worded it wrong, maybe. There's artefacting and the image is also blur.
So, besides fixing the artefacts, I also do a bit of sharpening.

Anyway, the script is look like this:

SetMTmode(2,4)
#SetMemoryMax(1024)
DSS2("F:\Path~\D.C.II S.S. ～ダ・カーポII セカンドシーズン～ 第03話 「正義の魔法使い」 (x264 1280x720).mp4")
aud=ffaudiosource("F:\Path~\D.C.II S.S. ～ダ・カーポII セカンドシーズン～ 第03話 「正義の魔法使い」 (x264 1280x720).mp4")
deblock_QED()
FastlinedarkenMOD(strength=0, thinning=80)# I know this filter is not supposed to be used for live playback, I guess. But somehow it runs smoothly on quad.
degrainmedian(mode=3, LimitY=4, LimitUV=5)#can't afford more than this with this filter chain when using ffdshow.
gradfunk(1.8)
Limitedsharpenfaster(strength=30, undershoot=3, Smode=4)
vid=last
audiodub(vid,aud)#dunno how to make the audio play, so I just add something like this. Any better method?
#delayaudio(0.10)

Edit:
@Dark_Shikari
ah, deflicker, OK, I'll try and report the result.
Results:
The flickering is still there.
My LCD is on 75Hz.

1. While you are setting the MT mode and the number of threads you are not really telling any of the filters to be used multithreaded. You have to do that with MT("filter"). Some may not be safe for that use, check MT's pages for more info.

2. Why use DSS2? You can use DirectShowSource and it will input audio too so no need for the aud definition and audiodub.

3. And yes, that is probably bound to be slow. Use filters which can run in realtime, like you could replace the denoising with hdn3d and line thinning with (a)warpsharp for example.

4. After you have it cleaned up you can create a profile and dump into ffdshow's configuration.

[Nagato]

Ah, thanks, I'll try clean-up the filters.
But I don't understand your first point. Why there's need to use MT after apply SetMTmode? This two filters use different slicing method, right? And afaik SetMTmode is easier to use because there's also filter that demand specified mod (like mod8, 16 etc). Well, when there's too many filters can't run multi-threaded usually MT is easier.

Edit:
with directshowsource I get too many haali's media splitter and ffdshowaudiodecoder tray icons. Well, caused by the SetMTmode filter maybe. Besides, directshowsource runs much slower, hmm. Also tried it with audio=false, but still cant run live.

[martino]

SetMTMode only sets the options to be used for MT. It doesn't specifically tell it to be used on filter x and y or z.

SetMTMode(2,2)
RemoveGrain()

is not multithreaded, whereas

SetMTMode(2,2)
MT("RemoveGrain()") #will now multithread RemoveGrain, on mode 2 and with 2 threads

is.

[Nagato]

Originally Posted by GLH

I have a post-processing related question/problem.

I use the CCCP and fast spp deblocking under FFDShow's post-processing options. This cleans up my .AVI files nicely. However, it doesn't seem to affect my .MKV files. Is there an option I need to change to fix that? Or do I need to use a different tool to apply deblocking to my .MKV files? Thanks!

What format used in the mkv? well, most likely h.264.
make sure ffdshow is used for h.264 decoding.

[Nagato]

Originally Posted by martino

SetMTMode only sets the options to be used for MT. It doesn't specifically tell it to be used on filter x and y or z.

SetMTMode(2,2)
RemoveGrain()

is not multithreaded, whereas

SetMTMode(2,2)
MT("RemoveGrain()") #will now multithread RemoveGrain, on mode 2 and with 2 threads

is.

OK, now I'm lost. When using MT I don't apply setMTmode bacause MT will automatically use mode 3 (?).

[martino]

Once again... SetMTMode() sets global MT parameters, MT() applies multithreading to the filter in () with the global parameters. What is there not to get? Perhaps reading the docs would be a good idea, but given they have 2x as much text I dunno.

[Nagato]

Originally Posted by martino

Once again... SetMTMode() sets global MT parameters, MT() applies multithreading to the filter in () with the global parameters. What is there not to get? Perhaps reading the docs would be a good idea, but given they have 2x as much text I dunno.

I read the documentation before and when only applying setMTmode without MT all cores are used, so still SetMTmode won't make used filter multithreaded. OK, I'll try add MT("filter") and check how speed change.

[martino]

Note: Your all cores will always be used. The system spreads out the load. It doesn't dump all of it one core and leave the rest sitting there doing nothing. AFAIK.

[GLH]

Originally Posted by Nagato

What format used in the mkv? well, most likely h.264.
make sure ffdshow is used for h.264 decoding.

It is indeed h.264. When I go to the "Settings" option under my "Combined Community Codec Pack" folder in the start menu I see the FFDShow video decoder choices. All the boxes, including h.264, are checked. Do I need to change an option anywhere else? To clarify a little bit, I have the same episode of a certain show in both its .avi format and its .mkv format. The .mkv version is a higher resolution but they both have blocking problems in the same areas. With fast SPP deblocking the .avi version is cleaned up and the .mkv version is not. Thanks!

[cyberbeing]

FFDShow disables postprocessing when it is decoding h.264. H.264 uses an in-loop deblocker which is done by the decoder so FFDshow disables postprocessing to prevent people from deblocking twice by mistake.

If you want to use FFDshow deblocking, go into MPC > External Filters > Select FFDshow Raw Video Filter > Set to Prefer > OK. By doing this, whenever you play a video file, FFDshow Raw Video Filter will also load and you can enable postprocessing there.

[GLH]

Originally Posted by cyberbeing

FFDShow disables postprocessing when it is decoding h.264. H.264 uses an in-loop deblocker which is done by the decoder so FFDshow disables postprocessing to prevent people from deblocking twice by mistake.

If you want to use FFDshow deblocking, go into MPC > External Filters > Select FFDshow Raw Video Filter > Set to Prefer > OK. By doing this, whenever you play a video file, FFDshow Raw Video Filter will also load and you can enable postprocessing there.

Thanks, but is there any way to do this with windows media player?

[cyberbeing]

If you set the merit of FFDshow Raw Video Filter to maximum then it would result in the same thing. I like using GSpot for setting codec merits.

Launch GSpot > System Menu > List Codecs and Other Filters > Find FFDshow Raw Video Filter in the list > Right Click > Set Filter Merit > Move the bar to the top (or type in 0xf00000) and click OK.

[GLH]

Originally Posted by cyberbeing

If you set the merit of FFDshow Raw Video Filter to maximum then it would result in the same thing. I like using GSpot for setting codec merits.

Launch GSpot > System Menu > List Codecs and Other Filters > Find FFDshow Raw Video Filter in the list > Right Click > Set Filter Merit > Move the bar to the top (or type in 0xf00000) and click OK.

I gave that a shot and am only met with "Failed to change merit."

[cyberbeing]

I'm assuming it's because you don't have administrative access on XP or you are running Vista. If on XP you'll have to find someone with an Administrator account and use that to make the change. If on Vista, by right clicking on the GSpot executable and clicking Run As Administrator and accepting the UAC prompt you should then be able to change the filter merit.

[Nagato]

@Martino
LOL, this MT thing forced me to reread the documentation, including this page http://avisynth.org/mediawiki/MT_support_page. It seems that mixing SetMTmode and MT is not a good idea. Well, see this and this

But, yes, in ffdshow SetMTmode doesn't seem to work ( or precisely, ffdshow just ignores it?), but MT works well.

Originally Posted by martino

Note: Your all cores will always be used. The system spreads out the load. It doesn't dump all of it one core and leave the rest sitting there doing nothing. AFAIK.

Yes, I think the system will always tries to allocate all of the available CPU, but it depends on the executed program's multi-threading compatibility. If it doesn't support multi-threading usually it'll ended with maxing one core only. In fact, there's still too few programs can run multi-threaded. It'd really great if ffdshow video decoder, ymc, etc can run multithreaded, coz they just max one core and leave the others idle orz.

I hope I didn't misunderstand you here.