Need advice on deblocking

[K0zi]

I can't get rid of blocking in some clips.
Samples:
1. http://rapidshare.com/files/77977763/SAMPLE.rar.html (RestoreFPS(25,0).yadif().nnedi() removes blending and fix screwed-up edges)
2. http://rapidshare.com/files/78644142/SAMPLE2.rar.html

I'm trying to encode these clips to x264. No matter what matrix, deblock settings or bitrate (well, within reasonable boundaries :p ) I use, I end up with the same (at least) ugly pixels in resulting file. These are especially visible during the flickering (2nd clip).
Could you suggest any x264 settings that would improve the quality? Or shall I rather try to work on this via Avisynth? (already tried gradfunkmirror without satisfying results)

[ToS_Maverick]

are you experiencing blocks like this:
http://forum.doom9.org/showthread.php?p=996829#post996829
(2nd row, 1st right)

if so, you have to try out AQ to get rid of it. Dark Shikari's new AQ is very effective!

[Dark Shikari]

Do you mean your original clips have blocking you want to remove--or x264 is adding blocking?

[K0zi]

Blocking is in original clips.
Screens from the mpeg2 source:

1st clip, processed as posted above:
http://img214.imageshack.us/img214/5904/blocking1qh5.th.png (http://img214.imageshack.us/my.php?image=blocking1qh5.png)
2nd clip, unprocessed, blocks visible especially during flickering:
http://img89.imageshack.us/img89/7747/blocking2sa6.th.png (http://img89.imageshack.us/my.php?image=blocking2sa6.png)

[Dark Shikari]

The purpose of x264's deblocking filter isn't to deblock input video--its to prevent encoding artifacts.

That video is very low quality--you're going to need some serious deringing/deblocking.

[Sagekilla]

You can try any of the various deblocking filters: Deblock, Deblock_QED, etc. Check the avisynth filter page and try them out. Also, you should try to use identical images and identical clips when comparing processed vs unprocessed. It's hard to look for what you're trying to point out with 2 different pictures at different resolutions.

[Ranguvar]

Also, you should try to use identical images and identical clips when comparing processed vs unprocessed. It's hard to look for what you're trying to point out with 2 different pictures at different resolutions.Seconded :)

ffdshow also has a nice selection of deblockers.

[K0zi]

Also, you should try to use identical images and identical clips when comparing processed vs unprocessed.

All images have been taken from AvsP.
"Processed" - I meant that the source (1st clip) was treated with this script:
RestoreFPS(25,0.00)
AssumeTFF()
yadif()
nnedi()


I think Deblock or Deblock_QED remove too much details. SmoothDeblock3 seems quite fine, I'll test it tonight.

[Sagekilla]

All images have been taken from AvsP.
"Processed" - I meant that the source (1st clip) was treated with this script:
RestoreFPS(25,0.00)
AssumeTFF()
yadif()
nnedi()


I think Deblock or Deblock_QED remove too much details. SmoothDeblock3 seems quite fine, I'll test it tonight.

Understood, but the images you used for comparison in your initial post were from two different points in the clip and of two different resolutions. Comparing that way can be very difficult unless the problem is blatantly obvious. Just a "so you know now," is all :)

Also if you're finding Deblock or Deblock_QED too intense, you can try lowering the quants for one of them. I don't remember which does that, but both should let you decrease deblocking parameters to make it at least less visible and easier for x264 to work with.

[foxyshadis]

Try also the new Unblock. I've had good results that don't require any tweaking with it, although I'm sure Didée can still come up with some magic to improve it further.