Hi Ross,
I just wanted to keep you in touch about my issue, and I'm glad to say
it's now solved.
In fact, ffmpeg libraries already contain some useful parsers, and I
especially use the 'av_parser_parse' method, which simply reads an input
buffer and rebuilds a proper output frame.
Once a frame is complete, you just have to decode it using traditional
methods, and everything rolls on.
Don't know if you're interested in those news though, but it might be
useful if some liveMedia users want to do the same as me :)
Best regards,
Guillaume.
Le 07/09/2010 08:47, Ross Finlayson a écrit :
This video has (on purpose) no audio track, so I can focus on video
stream only.
Payload of value 32 is an MPEG1 or MPEG2 video
That's correct.
An important thing to note about the RTP payload format for MPEG-1 or
MPEG-2 video (defined in RFC 2250) is that the data units delivered by
RTP are not (necessarily) complete video frames (even though our code
refers to them as 'frames' - which is admittedly a bit confusing).
For MPEG-1 or 2 video, the data units delivered by RTP - i.e.,
delivered one-at-a-time by a "MPEG1or2VideoRTPSource" object - can be
MPEG headers, or one or more video 'slices', but need not be a
complete video frame. (See RFC 2250 section 3.1 and Appendix 1.) This
allows a decoder to handle data loss without necessarily discarding an
entire video frame, but also means that it has to be a little smarter
about properly detecting when it has received enough data to make up
an entire frame.
--
Guillaume FERRY
Bertin Technologies
Département Bertin Conseil
Activité Traitement de l'Information et du Contenu
/Tél/ 01.39.30.62.09
/Fax/ 01.39.30.62.45
/Mail/ [email protected]
/Web/ www.bertin.fr <http://www.bertin.fr>
_______________________________________________
live-devel mailing list
[email protected]
http://lists.live555.com/mailman/listinfo/live-devel
