At 11:27 AM -0400 9/14/05, Manfredi, Albert E wrote: >MPEG-2, just like AVC, depends on wide spacing of I frames >as one of its bandwidth compression techniques. That's why >MPEG and AVC are different from M-JPEG. Probably *the* >most important difference, in fact. This is not true. I frame spacing is primarily an issue of signal acquisition time. I agree that B frames do make major contributions to compression efficiency, but this is equally true for long and short GOPs. I frame spacing is ALSO a function of the content. You can get away with longer GOPs when the scenes do not change often, while content with many fast cuts and rapid motion will require higher I frame frequency. AVC improves on MPEG-2 by offering the ability to force new I frames in sub-regions of the raster when things are changing rapidly. > >The better the motion vectors, the further apart the I >frames can be spaced, for a given level of moving image >quality, and the greater the compression factor becomes. >Isn't that the whole point? Apples and oranges. The quality of predictions has NOTHING to do with I frame spacing. Better predictions mean less differences to encode, which improves compression efficiency. You still need new I frames when the scene content changes. I have seen many cases where poor predictions lead to the need for as much difference data as a new I-frame would require. When you see blocking artifacts, the algorithm is breaking down because it is overwhelmed by entropy... That's a fancy term for the lack of redundancy in the video stream Bert. >So if you are expecting to see lower bandwidth with AVC >streams than with equal quality MPEG-2, you will see >increased spacing between I frames as one of the techniques >used to achieve the lower bandwidth. If you don't see an >increase in I frame spacing, then you also won't see as >much of a bandwidth reduction. Most operators are not going to increase I frame spacing to save a few bits. You might get away with this for some scenes when encoding source for a DVD, but not with a continuous program feed from the typical cable, DBS or Broadcast channel. In these applications GOP length will be no more than 15 frames to keep channel acquisition times reasonable. AVC has plenty to offer in terms of improved compression efficiency when GOPs are held consistent with MPEG-2. I asked Bert how he comes up with these illogical conclusions. he responded: "By figuring out how this stuff works." Go figure... 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.