[x264-devel] Re: [x264 SVN-623] rate-control miserably fails with -m6 or -m7

Loren Merritt lorenm at u.washington.edu
Tue Feb 27 07:14:49 CET 2007


On Mon, 26 Feb 2007, Jirka Klaue wrote:

> The sequence
>  <http://www.tkn.tu-berlin.de/~jklaue/flower_cif_10fps.yuv>
> failes to be correctly encoded with the command line
>  "x264 -B 56 -m6 --fps 10 -o flower_cif_10_56.264 flower_cif_10fps.yuv 352x288"
> .
>
> However, it is encoded as expected with the command line
>  "x264 -B 56 -m5 --fps 10 -o flower_cif_10_56.264 flower_cif_10fps.yuv 352x288"

cause: the -m6 encode erroneously detects a scenecut, so then it has to 
raise the qp because I-frames are big, so inter-prediction is even less 
usable, so the next frame gets even more bias towards a scenecut, repeat.

quick fix: --pre-scenecut (uses the input frames instead of the compressed 
frames for scenecut detection, so high qp doesn't influence the decision)

BTW, don't ever encode a bitrate/resolution combination that needs 
average QP>40. It will look better if you downscale. e.g. your clip 
looks better at 224x192 QP=39 than at 352x288 QP=45 (both 56 kbps).

--Loren Merritt

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html



More information about the x264-devel mailing list