[x264-devel] rate control for static macroblocks

[Anne Aaron]

I'm running in single-threaded mode.

The code base that I tested is actually a few months old so I don't have adaptive quantization.
After reading more I realized that the commit last week on VBV improvements should help the issue I'm encountering:
http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=7331ea4e381898421a1a908ed57f73126faacd36
The new check in x264_ratecontrol_mb( x264_t *h, int bits ) for updating the QP after a row will suppress the QP increase.


Anne




-----Original Message-----
From: x264-devel-bounces at videolan.org [mailto:x264-devel-bounces at videolan.org] On Behalf Of Alex Giladi
Sent: Tuesday, June 10, 2008 2:59 PM
To: Mailing list for x264 developers
Subject: Re: [x264-devel] rate control for static macroblocks

Anne,
Which revision of x264 are you using? Are you using the multithreaded mode?
Alex.

2008/6/10 Anne Aaron <anne at dyyno.com>:
> I'm running x264 with CBR rate control under a few seconds of VBV buffer
> constraint,  on content where some rows of macroblocks are perfectly static
> (zero distortion from the previous frame co-located macroblocks). For these
> macroblocks,  the satd is zero so their statistics are not included when
> update_predictor(…) is called. Aside from divide by zero, any reason for
> when (var <  10), update_predictor() returns right away?
>
>
>
> This may be the cause of one artifact that I am observing --  After a row of
> static macroblocks, the rate control tends to think that it will overproduce
> for the remaining rows and therefore increase the QP for the remaining rows.
> In fact, when the entire frame is static, the QP of the frame is very high.
> Since the frame is static, this has no detrimental effect on the frame
> itself. However, since the QP of the  I frames are based on the average of
> the previous P frames, the I frames tend to have low quality and beating is
> observed . Anyone has had similar observations or explanations?
>
>
>
> Anne
>
> _______________________________________________
> x264-devel mailing list
> x264-devel at videolan.org
> http://mailman.videolan.org/listinfo/x264-devel
>
>
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel