[x264-devel] [PATCH] improved scene cut detection

[Gabriel Bouvigne]

Loren Merritt a écrit :
> And what content was that (or any other sample that the patch helped)?

It was a musical content (RHCP)

Some other tests:

*T3 trailer
before:
x264 [info]: SSIM Mean Y:0.9286776
x264 [info]: PSNR Mean Y:37.972 U:41.085 V:41.541 Avg:38.698 Global:35.166

after:
x264 [info]: SSIM Mean Y:0.9282334
x264 [info]: PSNR Mean Y:38.133 U:41.241 V:41.753 Avg:38.861 Global:35.137
(does this count as a positive or a negative result?)

*CSI trailer
before:
x264 [info]: SSIM Mean Y:0.9096916
x264 [info]: PSNR Mean Y:34.967 U:39.457 V:39.828 Avg:35.941 Global:34.465

after:
x264 [info]: SSIM Mean Y:0.9100573
x264 [info]: PSNR Mean Y:35.142 U:39.614 V:40.032 Avg:36.114 Global:34.463

*Final Fantasy trailer
before:
x264 [info]: SSIM Mean Y:0.9479361
x264 [info]: PSNR Mean Y:42.540 U:46.955 V:48.034 Avg:43.486 Global:39.432

after:
x264 [info]: SSIM Mean Y:0.9477924
x264 [info]: PSNR Mean Y:42.584 U:47.030 V:48.109 Avg:43.530 Global:39.415

*M&M's advert
before:
x264 [info]: SSIM Mean Y:0.8675424
x264 [info]: PSNR Mean Y:29.512 U:34.251 V:33.833 Avg:30.425 Global:29.224

after:
x264 [info]: SSIM Mean Y:0.8671178
x264 [info]: PSNR Mean Y:29.801 U:34.435 V:33.921 Avg:30.669 Global:29.218


> All of my test videos lose quality-per-bitrate with the patch. Not much 
> (-.01 dB), but very consistent.

I'm going to have another look at it, as I noticed that it goes wrong in 
fades to/from black/white.


--
Gabriel
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel