On 8/28/06, James Richard Tyrer <[EMAIL PROTECTED]> wrote:
Stephen Pollei wrote: > On 8/28/06, Lourens Veen <[EMAIL PROTECTED]> wrote: >> On Monday 28 August 2006 16:38, Timothy Miller wrote: >> > On 8/27/06, Dieter <[EMAIL PROTECTED]> wrote: >> Without having looked at it in detail, some thoughts. We need to do two >> things: motion compensation, and iDCT. >> http://www.xfree86.org/~mvojkovi/XvMC_API.txt it
> Also Maybe Not For This Version of the Card but later might look into > http://dirac.sourceforge.net/index.html > > It uses Wavelets and not Inverse Discrete Cosine Transforms(iDCT) AFAIK. > They also use "overlapped block motion compensation".
> http://dirac.sourceforge.net/documentation/algorithm/upconversion/index.htm > http://dirac.sourceforge.net/documentation/algorithm/global_motion/index.htm
Wavelets (small 2D FIR filters) are a superior technology for compression. This is what JPEG-2K uses.
Yes I've heard plenty of raves about wavelets. no block artifacts, but blurs more IIRC.
However, MPEG is based on block oriented DCT and that is what we need to decode.
Yes I know something of the math behind DCTs thats why I spelled it out in my prior email. Actualy I mispelt discrete. http://en.wikipedia.org/wiki/Discrete_cosine_transform Yes I know that wavelets and DCT are different. I was mostly mentioning the motion compensation aspects of dirac. I did so because it was mentioned that a more generic universal solution might be wanted. Someone mentioned theora. I mentioned dirac, however I haven't looked into whether dirac has any special MC desires above that of theora or mpeg4 . That's why I mentioned those two uri's at their site. Also XvMC can be used without invoking IDCT, if I'm reading the spec correctly. [[XvMC provides video acceleration starting at one of two places in the video pipeline. Acceleration starting at the first point, which we shall call the "Motion Compensation" level, begins after the the inverse quantization and IDCT at the place where motion compensation is to be applied. The second point, which we shall call the "IDCT" level, begins before the IDCT just after the inverse quantization.]] I imagine if one wanted to provide the most generic interface you could make it so you could mix wavelets,dct, and raw all in the same frame to be outputed. Heck how close is it to a composting engine? I also explicitly said "Maybe Not For This Version Of The Card". Maybe wavelet support would be nice but I think MC and DCT is more important. -- http://dmoz.org/profiles/pollei.html http://sourceforge.net/users/stephen_pollei/ http://www.orkut.com/Profile.aspx?uid=2455954990164098214 http://stephen_pollei.home.comcast.net/ _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
