[x264-devel] Theora Ahead of H.264 In Objective PSNR Quality

Jason Garrett-Glaser darkshikari at gmail.com
Mon May 11 11:12:34 CEST 2009


On Mon, May 11, 2009 at 2:02 AM, Panagiotis Issaris
<takis.issaris at uhasselt.be> wrote:
> Hi,
>
> On Thu, 2009-05-07 at 20:57 -0700, Jason Garrett-Glaser wrote:
>> On Thu, May 7, 2009 at 7:25 PM, Jason Garrett-Glaser
>> <darkshikari at gmail.com> wrote:
>> > On Thu, May 7, 2009 at 6:30 PM, Jason Martens <cacepi at gmail.com> wrote:
>> >> From Slashdot:
>> >>
>> >> http://tech.slashdot.org/article.pl?sid=09/05/07/2352203
>> >>
>> >> Let the debunking begin!
>> >
>> > Apparently they seem to have used catastrophically bad settings for
>> > encoding.  I'm going to guess they just encoded idiotically with
>> > ffmpeg without using any presets, thus using ffmpeg's quite broken
>> > defaults.  I suspect this is entirely intentional.
>> >
>> > It isn't worth spending time "debunking" though; anyone dumb enough to
>> > actually believe such a thing deserves the godawful quality Theora
>> > will give them.  And when they get that godawful quality, if they are
>> > still dumb enough to not realize they've been scammed by Xiph, they
>> > deserve every last bit of suffering they will experience.
>> >
>> > Dark Shikari
>> >
>>
>> And it turns out the problem was a bug in ffmpeg when writing the y4m
>> file prior to PSNR measurement, cutting a nice 4db off of x264's
>> score.
>
> Just out of curiosity, do you have a link to any info regarding this
> bug? I've been looking at ffmpeg-cvslog and am on the regular
> ffmpeg-devel list and haven't found it. Was it discussed on IRC or
> somewhere else?

IRC and ffmpeg-devel ML.

12:35 < CIA-33> ffmpeg: conrad * r18795 /trunk/libavcodec/ (8 files): Add a
                chroma_sample_location field to define positioning of chroma
                samples
12:36 < CIA-33> ffmpeg: conrad * r18796 /trunk/libavformat/yuv4mpeg.c: Use
                chroma_sample_location in reading/writing yuv4mpeg

Committed a few hours ago.

It isn't really an ffmpeg bug, since no PSNR tool should ever do
interpolation for any reason whatsoever.

Dark Shikari


More information about the x264-devel mailing list