Log in

View Full Version : Blending after encoding. How to fix it?


Bill Ein
23rd March 2014, 03:29
Hi everyone!
I have a problem: blending after encoding on the dark scenes fades.
CRF 15, no-mbtree and deadzone 6,4 doesn't help to solve the problem. This can be fixed if I set bframes=0 but maby you have another way?

Video (bframes=0):
http://www.mediafire.com/view/atr7kql58j8b40a/Sequence%2007_alt10bit0.mp4
If you reencoded this video with bframes=8 - you have blending on 10th, 16th, 18th and 19th seconds or only on the one of them like on this screenshots:
10th sec. 1st frame (without blending)
i57.fastpic.ru/big/2014/0320/44/4b2ae7c338b2133408d58e675a01bd44.png
2nd frame (light blending on the mouth, nose, ear and hair)
i58.fastpic.ru/big/2014/0320/fc/c5aa63e453b3262d3318edd34f9811fc.png
3rd frame (worse blending)
i59.fastpic.ru/big/2014/0320/fb/925a066a8832f617689e731860893bfb.png
4th frame (without blending)
i59.fastpic.ru/big/2014/0320/64/3b70d6906a10981ab76d5caf7acdf064.png

P.S. Sorry for my English.

foxyshadis
26th March 2014, 00:03
mbtree and deadzone wouldn't affect this. weightp and weightb might; try one at a time --weightp=0, --no-weightb, and both together to see if that affects the result.

But first, you're absolutely 100% certain it didn't show up in the final filtered source you're feeding x264? Even if that is the case, a mild one-frame blend during a fast fade isn't going to affect actual viewing at all.

Bill Ein
1st April 2014, 15:44
mbtree and deadzone wouldn't affect this. weightp and weightb might; try one at a time --weightp=0, --no-weightb, and both together to see if that affects the result.
I find the best settings for AMV encoder and AMV have to many action scanes, but weightp=0 - realy bad for action.
--no-weightb - bad for low bitrate, but for normal and hight not fixed this problem.
But first, you're absolutely 100% certain it didn't show up in the final filtered source you're feeding x264?
Absolutely

Even if that is the case, a mild one-frame blend during a fast fade isn't going to affect actual viewing at all.
But It's no good.

Bill Ein
24th May 2014, 08:22
Now I've found the best settings for my video and fixed blending artefacts:

8bit 240p (2 pass; without psy-rd; keyint=fps, b-adapt=0)
--bitrate 544 --direct auto --deblock +0:+1 --ref 2 --bframes 2 --b-adapt 0 --trellis 2 --me tesa --subme 11 --partitions all
--no-mbtree --aq-mode 2 --no-dct-decimate --no-psy --no-fast-pskip --b-pyramid strict --keyint 24 --ratetol 100

8bit (without psy-rd) ver1
--crf 18.0 --direct spatial --deblock +0:+1 --ref 2 --bframes 2 --b-adapt 2 --trellis 2 --me tesa --subme 11 --partitions all
--no-mbtree --aq-mode 2 --no-dct-decimate --no-psy
8bit (without psy-rd) ver2
--crf 17.0 --direct spatial --deblock +0:+0 --ref 3 --bframes 3 --b-adapt 2 --trellis 2 --me tesa --subme 11 --partitions all
--no-mbtree --aq-mode 2 --no-dct-decimate --no-psy

10bit (without psy-rd; b-adapt=1!)
--crf 17.0 --direct spatial --deblock +0:+0 --ref 5 --bframes 5 --b-adapt 1 --trellis 2 --me tesa --subme 11 --partitions all
--no-mbtree --aq-mode 2 --no-dct-decimate --no-psy

8bit
--crf 16.0 --direct spatial --deblock -1:-1 --ref 6 --bframes 6 --b-adapt 2 --trellis 2 --me tesa --subme 11 --partitions all
--no-mbtree --aq-mode 2 --no-dct-decimate --psy-rd 0.70:0.00

10bit (b-adapt=1!)
--crf 15.0 --direct spatial --deblock -2:-2 --ref 9 --bframes 9 --b-adapt 1 --trellis 2 --me tesa --subme 11 --partitions all
--no-mbtree --aq-mode 2 --no-dct-decimate --psy-rd 1.00:0.00