[Libav-user] Seeking by byte position in H.264 video

Дмитрий Дороговцев dorogovtsev.dmitriy at gmail.com
Sat Aug 6 15:31:36 CEST 2011

Hello. I am trying to implement some kind of seeking mechanism in H.264
without any container, just raw H.264 stream. I have some troubles with it.

1) Problem #1. I am using ffmpeg to cut let's say 143 frames from video in
this way:

ffmpeg -i source.h264 -vcodec libx264 -b 1200k -r 30 -vframes 143

It works without any problems. After that i'm getting all frames in pictures
from obtained file

ffmpeg -i result.h264 -vcodec mjpeg img%03d.jpg

It works also fine, but after this i get set of images with names img001.jpg
- ... - img142.jpg.
So i have 142 images instead of 143. Where is last( or first ) one?

2) Problem #2. I am trying to get amount of frames in the video using
while( av_read_frame( this->pFormatContext, &this->packet ) == 0 )
In this way i get amount_of_frames = 143 for the previous example. But i'm
not sure
about what should i use to get byte position in the file. Does AVPacket
packet.pos points
to the proper position in the file, so i can use it for calling
av_seek_frame? Will it properly
take me to the desired position of frame beginning? I'm curious because if i
avcodec_decode_video2() immediately after av_read_frame, when packet.pos
equals 0
( i think, it's start of file ), then 3rd parameter *got_picture_ptr stays
equal to zero meaning
that no frame could be decompressed.

I need byte positions to create frame index without actual decoding the
whole video. After that
i should be able to jump through it using av_seek_frame. If it is bad
approach for some reason,
please give me an advice to the better solution. Thank you.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ffmpeg.org/pipermail/libav-user/attachments/20110806/cd30cf1a/attachment.html>

More information about the Libav-user mailing list