[Libav-user] Accessing decoding information in MPEG-2/4 video streams?

[Sebastian Wahl]

Hi all,

As a hobby I have been looking into and experimenting with image restoration techniques. As of late I have been trying out multi-frame super resolution and I'm thinking about looking into MPEG compression.

To give a bit of context, in image restoration you try to recover a degraded image. Usually this is done by creating a model for how the image was degraded and then try to estimate the image which when passed through that model will give the degrade image. Thus the better you understand how the image was degraded, the better estimation you can make.
With MPEG compression you want to know as much as possible about the decoding process to better estimate the parameters in the degradation model.

So I want to dig deeper than just the resulting YCrCb frames but I don't know how this is most easily done, reading a media container and decoding the video from scratch is just too much work. I know you can get the motion vectors from libav, but what else does it expose? (I'm primarily interested in H.262 and H.264.) If I need to go deeper, for example to decode the video stream from scratch, what parts of libav should I look into to avoid having worry about containers, etc.? Or is there another library which might be a better way to approach this?

Just to avoid misconceptions, I want to do this more because of technical curiosity than trying to create a solution for a problem. It would be a fun way to learn more about video compression and image restoration but avoiding to reimplement everything would make it a tad more realistic...

Sebastian Wahl
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ffmpeg.org/pipermail/libav-user/attachments/20150804/d44ba946/attachment.html>