[x264-devel] [PATCH]adaptive lowpass

[Gabriel Bouvigne]

Loren Merritt a écrit :
> There shouldn't be any zigzag involved in your code. The input of 
> x264_lowpass_dct comes straight from sub*_dct without reordering.

My bad, thank you for pointing at this.
Strange thing is that correct ordering increases psnr, but doesn't 
change ssim much.

> Skip the whole process if QP is sufficiently low that it does nothing. 
> This would both improve speed and eliminate half of the table.

Done in this updated version.

> Aside from being adaptive, does lowpass provide any benefit over a CQM 
> which quantizes the high frequencies more?

 From a theoretical point of view, it provides a transition band, which 
is supposed to be better than a brickwall filter.
Practically, there is treillis quant used after it, so the transition 
band is preserved...sometimes.

 From a practical point of view, the main benefit is adaptability. Using 
a CQM with very high quantization of high freqs (or using an insane 
deadzone), you can also do a lowpass (but only a brickwall one). 
However, CQM does not allows for adaptability, and both intra and inter 
MB would have the same filtering.

--
Gabriel
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: dct_lowpass_2.patch
Url: http://mailman.videolan.org/pipermail/x264-devel/attachments/20070627/7d6f10f9/attachment.txt 
-------------- next part --------------
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel