x264. macrblocks in dark areas

[daimroc]

I am trying to code a sample of a movie and y notice that I get macroblocks in dark areas. This movie is 1920x1080 from a BD.

My command line is:

x264 --scenecut 40 --bframes 16 --b-adapt 2 --b-pyramid normal --ref 8 --deblock 0:0 --crf 16 --rc-lookahead 100 --ipratio 1.1 --pbratio 1.1 --keyint 240 --min-keyint 24 --qpmin 5 --qpmax 24 --aq-mode 2 --aq-strength 0.5 --mbtree --qcomp 1.0 --partitions all --direct auto --weightp 2 --me tesa --merange 32 --mvrange -1 --mvrange-thread -1 --subme 9 --psy-rd 1.0:0.0 --mixed-refs --8x8dct --trellis 2 --no-fast-pskip --no-dct-decimate --deadzone-inter 0 --deadzone-intra 6 --nr 6 --threads auto

I tried with differents variations, for example subme 10, or aq strength 1.5, 2.0, deadzone 0,0... but in this scene the results are always the same, a cloud of macroblocks and loose of details.

In not dark scenes the quality is very very near to the original.

I would like to know how to reduce or elimnate this effect, because I think that I am using a suficient quality (crf 16) to encode the video.



Thanks.
Daimroc.

[Dark Shikari]

I am trying to code a sample of a movie and y notice that I get macroblocks in dark areas. This movie is 1920x1080 from a BD.

My command line is:

x264 --scenecut 40 --bframes 16 --b-adapt 2 --b-pyramid normal --ref 8 --deblock 0:0 --crf 16 --rc-lookahead 100 --ipratio 1.1 --pbratio 1.1 --keyint 240 --min-keyint 24 --qpmin 5 --qpmax 24 --aq-mode 2 --aq-strength 0.5 --mbtree --qcomp 1.0 --partitions all --direct auto --weightp 2 --me tesa --merange 32 --mvrange -1 --mvrange-thread -1 --subme 9 --psy-rd 1.0:0.0 --mixed-refs --8x8dct --trellis 2 --no-fast-pskip --no-dct-decimate --deadzone-inter 0 --deadzone-intra 6 --nr 6 --threads autoYour commandline is an insane mess. I would start by removing basically all the options.

If you're getting blocking in dark areas at CRF 16, it's more likely a monitor calibration issue than anything else.

[daimroc]

When you say that could be a calibration problem, you mean in the contras and bright configuration?

But if it is a calibration problem, with the original video I would see the same problem, but it is not. am I wrong?



Thanks.
Daimroc.

[Dark Shikari]

When you say that could be a calibration problem, you mean in the contras and bright configuration?

But if it is a calibration problem, with the original video I would see the same problem, but it is not. am I wrong?This has been posted about many times before. Use search for more detailed information.

Dither is extremely difficult to retain--the best method is a lot of AQ, lots of psy trellis, no-dct-decimate, and low qpmin. However, in most cases, the loss of dither in the darkest areas is not very visible. A badly calibrated monitor can make it much more visible than it should be. TV levels being displayed incorrectly (without convert to PC) can have a similar effect.

[MasterNobody]

daimroc
Do NOT use --aq-mode 2 (a specially in conjunction with MBTree). I tell you this as the author of this mode.

[nurbs]

What about --aq-mode 4 and MB-Tree? That works, right?

[MasterNobody]

--aq-mode 4 is OK. But it exist only in patched builds.

[daimroc]

Masternobody, then you suggest to use aqmode 1 or disble it?

What is a better option, use mb tree and not aq mode 2 or use aqmode 2 and disable mb tree?




Thanks.
Daimroc.

[MasterNobody]

Better to use --aq-mode 1

[daimroc]

A do some other tests and yes, with the aq mode 1 I the results are much better.

Also I use psy-rd 1.0:1.0, and I get a slight better quality.



Thanks.
Daimroc.

[jpsdr]

Have you tested with the new aq-mode 2, wich is the aq-mode 4 of the old patch commited ?
I'm interested in the result, to see if now the new aq-mode 2 is Ok, and maybe even better...