Hi, On 7/23/2009 8:32 PM, Josh Allmann wrote: > On Thu, Jul 23, 2009 at 3:17 AM, Alessandro Ferrari > <[email protected]> wrote: >> Hello dear community, >> in your opinion, I can use ffmpeg library >> for parsing an encoded h.264 p-frame to extract Macroblocks and Motion >> prediction Vectors Information? My target is to transcoding an h264 stream >> for reducing bitrate by reduce granularity of the DCT coefficient, and for >> achieve this i want to first extract information by the original encoded >> stream, decode the stream, and then encoding the stream without recalculate >> the motion prediction information. >> Thank in advance for your attention. > > Take a look at the MV visualization code in ff_print_debug_info of > mpegvideo.c. > > I'm doing the same thing right now, and it involves a lot of hackery. > You can calculate MVs through motion_val field of avframe, but to be > useful you also need the MB info, which is tucked tucked away in the > mpegenccontext field of h264context, neither of which are available > via the public API (afaik). > > Right now I basically #include h264.h which is really messy. If there > is a better way of doing this I'd like to hear it.
I think this information should be available in AVFrame, no ? In any case, if that is useful to you guys, I believe exporting this value using API is ok and a patch is welcome. -- Baptiste COUDURIER GnuPG Key Id: 0x5C1ABAAA Key fingerprint 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA FFmpeg maintainer http://www.ffmpeg.org _______________________________________________ libav-user mailing list [email protected] https://lists.mplayerhq.hu/mailman/listinfo/libav-user
