Picture turns to blocky after IVTC

[Great Dragon]

Hello,

After trying to do IVTC to a video I've figured out that some frames turn into a peace of crap.

So far I've reached my best result with
AnimeIVTC(2, bbob=5, extbob=MC_A_bob, overlap=6, pel=4)
http://t1.someimage.com/RWyPPjC.jpg (http://someimage.com/RWyPPjC)

Both mode1 and mode2 of AnimeIVTC produce worse results.

Here is 3 sec sample to those who wants to take chance with such kind of video.
https://dl.dropboxusercontent.com/u/30473146/tpodw.demuxed.m2v

I would be grateful for any kind of help from the community.

[captainadamo]

Those have nothing to do with the IVTC as the blocks are visible even when played back as is without any post-processing. Your source is heavily compressed with lots of visible compression artifacts. It also suffers from lots of analog noise and composite artifacts (dot crawl and rainbowing). Run a deblocker, some denoising and cleanup for the composite artifacts.

[Great Dragon]

Don't even know that it is possible that from 3 consecutive frames 1st and 3rd looks OK but 2nd looks blocky because of compression. I saw only sources where all frames have been blocked because of compression.

As for the dot crowl and rainbows that is what I can handle with ease.

[captainadamo]

If it weren't possible then I wouldn't be seeing the exact same blocks from your picture even without doing an IVTC.

[captainadamo]

Here is the exact frame as your screenshot straight from DGDecode with no IVTC or any filtering applied:

http://105.imagebam.com/download/CLLTktsZqTgi2Hfc37Xh1Q/40290/402890124/Untitled.png

As I said, the blocks are compression artifacts from your source. If all you've applied is AnimeIVTC to the source, then it is actually reducing the blocks as opposed to the untouched source. It is not causing them.

[Great Dragon]

Thanks, I've just saw it by myself after your first answer. It's just what I can't even imagine and encounter for the first time.

I've used DGDecode internal deblocker to made it not so ugly.
DGdecode_mpeg2source("index\01.d2v", idct=5,moderate_h=40, moderate_v=40, cpu2="xxoooo")

[fAy01]

Try dfttestMC and Simple MDegrain Mod
http://forum.doom9.org/showthread.php?t=147676 | http://forum.videohelp.com/threads/369142-Simple-MDegrain-Mod-v3-0d-A-Quality-Denoising-Solution

[Great Dragon]

Thanks for the advice. I'll give it a try. But isn't it's better to beblock before ITVC than after? And I'm taking just about beblocking and not overall picture denoising.

[Asmodian]

Yes, you have a good solution for deblocking. fAy01's suggestion is for after IVTC and is for denoising, not deblocking.

[kuchikirukia]

Try Edgecleaner.

[Great Dragon]

Edgecleaner stays for ringing, it doesn't have deblocking capability. It least in direct way.

So far I've tried deblock_QED and standalone deblock and none of them works as good as DGDecode's deblock.

[feisty2]

Edgecleaner stays for ringing, it doesn't have deblocking capability. It least in direct way.

So far I've tried deblock_QED and standalone deblock and none of them works as good as DGDecode's deblock.

an advanced denoise filter could be the one you are looking for

[feisty2]

http://i.imgur.com/OJc2JSA.png

padding (10,10,10,10)
blur=removegrain (17).blur (1)
detail=mt_makediff (last,nlmeanscl2 (a=10,b=0,s=4,aa=0.001,h=8))
nlm=nlmeanscl2 (a=10,b=0,s=4,aa=4.8,h=6.0).mt_adddiff (detail)
min=mt_lutxyz (nlm,blur,last,"x z - abs y z - abs > y x ?")
min.repair (min.removegrain (17),1)
crop (10,10,-10,-10)


EDIT: slight changes