[x264-devel] Re: x264: svn commit r266 (pengvado)

[Loren Merritt]

On Mon, 20 Jun 2005, Mans Rullgard wrote:

>> custom quant matrices
>
> Are these working now?  From the irc discussions I got the impression there
> were some chroma issues.

As far as I can tell, the reconstruction is consistent with JM, and the 
encoder has no problems.

The only remaining issue is:
I use the default 4x4 flat matrix for fast skip detection. If a custom 
matrix has a significantly different average coeff value (and thus 
different bitrate per qp), then the skip threshold may be wrong.
I tried using the custom 4x4PY matrix, but at least for (--cqm jvt)
that produces way too few skips.
RD skip decision is possible, but slow, and not consistently beneficial 
without CQMs.

--Loren Merritt

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