[FFmpeg-devel] new open source h.264 source

Michael Niedermayer michaelni
Sat Mar 1 01:58:18 CET 2008


On Fri, Feb 29, 2008 at 02:18:12PM -0500, Francois Oligny-Lemieux wrote:
> On 2/20/08, Michael Niedermayer <michaelni at gmx.at> wrote:
> >
> >
> > Speaking of deblocking, our filter_mb_fast() is buggy it butchers some of
> > the reference streams. Iam just mentioning that before someone wastes time
> > optimizing buggy code ... he should fix it first :)
> >
> > [...]
> 
> 
> 
> Last week I found out when I decode a stream with high gop size (300) +
> deblocking filter on + MBAFF, I get some blocky artefacts on the decoded
> picture (when using ffmpeg to decode). If I turn off MBAFF it works fine. If
> I turn off deblocking it works fine. Using JM12.2 as a reference decoder
> shows that the original stream can be decoded successfully without
> artefacts.
> 
> Do you think it's the same problem as in the filter_mb_fast() ? Most of the
> code in filter_mb_fast is about non-MBAFF.
> 
> Any hint how to debug this? 

The standard way to debug decoder differences between ff and JM is to have
both output raw YUV and then search for the first differing byte, after that
find out in what MB that is and add printf() av_log() as needed until you
find out where JM and ff diverge for that MB. (it should be somewhere in
the loop filter according to your description) Last you then look in the
spec to see how the specific part should be done, fix it and send a patch.


> I'm trying to see if the lastest ffmpeg svn has
> this problem but I can't decode my h.264 file to .yuv (see below)
> 
> I can supply a sample .ts file if needed or .264 annex b.
> 
> BTW I know this is user related question, but when building ffmpeg as-is
> from svn with cygwin, I cannot get to take h264 .ts nor .264 files and
> output .yuv files, it responds with:
> 
> ...
> [rawvideo @ 0x7606f0]dimensions not set.
> Could not write header for output file #0
> 
> Is it my build that's corrupted or this is expected ?

what command line did you use and what was the complete and uncut output
from ffmpe?

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The worst form of inequality is to try to make unequal things equal.
-- Aristotle
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20080301/cf5335ed/attachment.pgp>



More information about the ffmpeg-devel mailing list