At 6:54 PM -0400 9/21/05, Manfredi, Albert E wrote: >Craig Birkmaier wrote: > >> Ultimately, the greatest potential for further >> gains in compression efficiency will come from >> improvements in the ability to predict B-frames. >> The good news is that what we are doing today is >> still VERY CRUDE - as more MIPs become available >> it will be possible to further refine the >> prediction routines. > >This seems logical. > >In my mind, the even better news, from a standpoint >of compatibility, is that the improvements made in >predicting B frames should also apply to MPEG-2. >After all, the cruder they are today, the more they >have a potential for improvement. > >For systems that won't jump to AVC quite so quickly, >such as Aussie and US DTT, and digital cable, this >could breathe some extra life in the existing >standard. To a very large extent, this has already been done with MPEG-2. This year at NAB, Snell And Wilcox offered an MPEG-2 pre-processor based on their standards convertor technology. There is some room for application of these techniques, but the gains are probably going to be a few percent at best. The main reason is that there is no way to improve prediction accuracy beyond what is possible in an MPEG-2 decoder. Issue like the macroblock size, the impact of the DCT on small area distortions, and the accuracy of motion vectors limit what is possible. For example, H.264 allows for 1/4 pel accuracy for motion vectors while MPEG-2 is limited to 1/2 pel. The H.264 algorithm allows for 4 x 4 coding blocks, which further isolates the energy in the block limiting the negative impact on small area distortions that causes so many problems in MPEG-2; this works hand-in-hand with the more accurate motion vectors, allowing the decoder prediction to be more accurate. With MPEG-2 you may know where the pixels are supposed to be, but if you can't put them there due to the limits on the encoding tools, you cannot do a better job. Bottom line, you can only work around these limitations through more clever encoder designs - in the end, you must live with the limits of existing decoders. This is one case where periodically replacing hardware based decoders can make a huge difference. Regards Craig ---------------------------------------------------------------------- You can UNSUBSCRIBE from the OpenDTV list in two ways: - Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org - By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word unsubscribe in the subject line.