Hi Everybody,
Can some one advice me any methods to convert H.264 motion vectors to MPEG-2 Motion vectors.
Also I desperately need H.264 baseline encoder with source code. Actually I need to transcode H.264 video into MPEG-2 video. I tried working with JM9.x. I have broken my head but no result in case of Inter Prediction. Can some one give me ideas or point me proper resource to achive this.
Thank you in advance
----
Kalls
Full Version: How to convert H.264 motion vectors to MPEG-2 mvs
Another open source encoder would be x.264.
you should contact 'akupenguin'. find him on doom9 forum.
i can imagine this is hard or even impossible to do.
i can imagine this is hard or even impossible to do.
QUOTE(i4004 @ Nov 7 2005, 08:45 AM)
you should contact 'akupenguin'. find him on doom9 forum.
i can imagine this is hard or even impossible to do.
i can imagine this is hard or even impossible to do.
Thank you Latexxx and i4004.
there could be some problems though...
H.264 uses 3 different block sizes where MPEG-2 uses 1. there will be more vectors and they'll be in different places because of this.
you can interpolate an mpeg-2 vector-field from the H.264 vectors, but this interpolated field will most likely be of minimal use, as the motion search in both codecs is not based on actual motion, but rather on a search for a nearby block that has the lowest comparison metric (usually sum of absolute differences, or sum of squared errors which is a tad more noise-tolerant). this means you could end up with a very fugly video indeed, and most likely with more effort required than just a straight decode -> re-encode.
of course, you can go ahead and solve the above problems - this would be extremely helpful to the video community
H.264 uses 3 different block sizes where MPEG-2 uses 1. there will be more vectors and they'll be in different places because of this.
you can interpolate an mpeg-2 vector-field from the H.264 vectors, but this interpolated field will most likely be of minimal use, as the motion search in both codecs is not based on actual motion, but rather on a search for a nearby block that has the lowest comparison metric (usually sum of absolute differences, or sum of squared errors which is a tad more noise-tolerant). this means you could end up with a very fugly video indeed, and most likely with more effort required than just a straight decode -> re-encode.
of course, you can go ahead and solve the above problems - this would be extremely helpful to the video community
QUOTE(MugFunky @ Nov 9 2005, 06:25 AM)
there could be some problems though...
H.264 uses 3 different block sizes where MPEG-2 uses 1.
H.264 uses 3 different block sizes where MPEG-2 uses 1.
3 different block size? 16x16 , 8x8, 4x4 right?
And what about 8x16 , 16x8, 4x8, 8x4?
would you believe "a bunch of different block sizes" then?
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.