[Libav-user] Sampling parts of a video on packet level

Jesper Taxbøl jesper at taxboel.dk
Tue Sep 6 20:21:43 EEST 2016


I am working on a remote application that is using OpenCV to detect a ball
in a video stream. I have a need to monitor how its doing from home, and
therefore need access to source video and the detections found.

The system is unfortunately behind limited bandwith connection so I am
forced to only transfer samples of the video and their detections home. I
want to be able to reproduce the detections at home.

To do that I have setup an encoder that copies frames from the opencv video
stream and create a small mp4 file of a range of images. e.g: 1000 frames
somewhere inside the middle of the video.

My problem is that I am currently encoding the frame again, introducing a
small change in the video, so I cant expect the same results at home as the
remote system is seeing.

I was therefore wondering if its possible to pickup data frames before the
video decoding and assemble them into a small sample video that would give
me the exact same video data as the remote system is seeing?

I am aware that I probably need to sync with existing keyframes in the
sourcevideo, but I assume there also will be issues with presentation times
etc, as the sample is shorter.

Do you think its possible at all?

kind regards

Jesper Taxbøl
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ffmpeg.org/pipermail/libav-user/attachments/20160906/8511f556/attachment.html>

More information about the Libav-user mailing list