Log in

View Full Version : OGM - error correction container extension? or Detecting defective data.


Ram
31st March 2005, 19:59
I'm wondering if this is needed.

I've noticed on ocassion that some OGM encapsulated files have 'troubles' namely there length is wrong and it looks like some of the video and audio data was corrupted.
The question really is
1) Is it needed? or desireable?
Xvid Divx variants of MPEG4 I believe all have error detection schemes for decoding. Basically throwing out garbage data. The issue here key frame timing, as I've noticed that if the key frame isn't interposed within every 60 frames (some people like to be sure that there are almost NO key frames in the encoding). Corrupted data makes for rough viewing (I call it the block smear syndrom). This appears to happen with Divx a lot.
MPEG4 has it's own error correction scheme right then?
Ogg Vorbis, MPEG4 audio and MP2 layer 3 audio have there own schemes for correcting errors as well. So is it necessary to correct bad data at all is my question.

2) If it is something that would be useful what streams would it be useful for?


I've seen XCD discusion else where but it appears to be a dead in the water developement now (as no updates, developement or discusion is active). XCD is just a format, this format requires a CD that supports XA Mode 2 Form 2 extraction (in layman's term think Playstation FMV data). This is NOT supported by all CD-ROM units for PC's and not supported by all CD writter units as well. Unless of course I missed something (somewhere).

So anyhow I thought I would ask if this would be an idea for an improvement on container data. Now I know error correction at the container level seems redundant, however it can be supported for DVD and standard ISO9660 format CD's (Mode 1 in other words). It does reduce how much you can cram into the CD, but doesn't change DVD's at all. This means you wouldn't have to differentiate between CD's and DVD's in how there file system appeared to your decoding sub system The data would just get feed as a stream of data weather it's from a CD or DVD (or 2 layer DVD or blue RAY or what have you).

The real 'issue' is again is this something useful or is it blowing smoke? LOL.

Ram

bond
6th April 2005, 17:59
first of all, all of the "new containers" we normally discuss here might be able to store error detection/correction code, but still there exist no tools which would be able to use such an info on the container level when reconstructing the data of a borked file, so this is more a theoretical discussion imho

more practically possible would be to use of error resilience inside the a/v streams themselves, eg the ffmpeg mpeg-4 encoder can set these during encoding
still there would be the question which decoder can make really use of such error resilience data when decoding corrupted streams