From: Michel L. <wa...@zo...> - 2001-10-06 06:46:49
|
On Fri, Oct 05, 2001 at 11:29:02PM -0700, Lu Xiaoxiang wrote: > Hi Michel, > Thank you for your answer. Yes, I am interested in the error > concealment stuff. Now I am working on mpeg2 video over IP network. I use > libmpeg2 package as the decoder. I concern the video quality degradation > after transmission across the network.So actually I will implement some > error concealment algorithms.I think the first step is how libmpeg2 > detects errors. Are you thinking about LAN or WAN applications ? I suppose for IP networks the first layer of error correction should be on the coded stream itself - using some kind or redundancy in the packet transmission. I had sent an email to the videolan mailing list not long ago about that subject, I also recommended people to have a look at http://www.csl.sony.co.jp/person/morelos/ecc/codes.html and http://www.iet.unipi.it/~luigi/fec.html You probably knew some of that already but I figure extra information never hurts :) As for the error detection in libmpeg2. The most common cause of problems (segfaults) right now is when we get an incorrect motion vector and we try to use pixels outside of the prediction frame and boom! I've started some work to detect that but I need to finish it. The second step would be to detect parsing errors like invalid huffman codes. I suppose after that we'll have a quite reliable detection of fscked up slices. -- Michel "Walken" LESPINASSE Of course I think I'm right. If I thought I was wrong, I'd change my mind. |