[FFmpeg-devel] [PATCH] Set color_space instead of?deprecated?YUVJ* pixel formats
michaelni at gmx.at
Thu Sep 1 16:29:42 CEST 2011
On Thu, Sep 01, 2011 at 09:24:34AM +0100, Gavin Kinsey wrote:
> On Thursday 01 September 2011 08:20:44 Gavin Kinsey wrote:
> > On Wednesday 31 August 2011 21:13:01 Michael Niedermayer wrote:
> > > On Wed, Aug 31, 2011 at 02:02:40PM +0100, Gavin Kinsey wrote:
> > > > On Wednesday 31 August 2011 13:52:35 Gavin Kinsey wrote:
> > > > > On Wednesday 31 August 2011 13:27:56 Michael Niedermayer wrote:
> > > > > > On Wed, Aug 31, 2011 at 08:20:47AM +0100, Gavin Kinsey wrote:
> > > > > > > Set color_space instead of deprecated YUVJ* pixel formats in
> > > > > > > MJPEG decoder.
> > > > > >
> > > > > > i think the pixel format should be left for compatibility for
> > > > > > now, just setting color_range in addition seems better
> > > > >
> > > > > The colour conversion is different in that case. Something
> > > > > different happens in swscale if the pixel format is YUVJ*, with or
> > > > > without color_range.
> > > > >
> > > > > Now perhaps this altered colour palette is correct, it wouldn't be
> > > > > the first time our servers send the wrong data, but it just looks
> > > > > washed-out to me.
> > >
> > > as you say "our servers" could you maybe check what they send and
> > > elaborate on that a bit ?
> Okay, I've talked with one of the server guys and got some details on what
> is being sent. The server just passes through the colour space that is used
> by the camera, so it sends CCIR 601 format for both JPEG and MPEG. It
> doesn't appear to set the "CS=ITU601" in the JPEG comment data that ffmpeg is
> looking for though, so as a result ffmpeg picks the wrong pixel format.
> I'll have to adjust this in my application as ffmpeg is doing the right thing
> as far as it knows.
Is there some other way than "CS=ITU601" to detect this ?
i think it might make sense if ffmpeg had a workaround and detected
> New patch attached that just sets color_range and leaves pixel format
> > I understand keeping YUVJ* in for application compatibility, but YUV420p
> > + AVCOL_RANGE_JPEG should be functionally equivalent to YUVJ420p inside
> > ffmpeg. If it isn't, then that's a bug surely? Or am I misunderstanding
> > something?
> Still think this is an issue though.
absolutley agree, and patch is welcome :)
and patch applied and
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
DNS cache poisoning attacks, popular search engine, Google internet authority
dont be evil, please
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 198 bytes
Desc: Digital signature
More information about the ffmpeg-devel