kilg0r3
7th June 2003, 11:20
one day, scrolling through the stream i was about to compress i noticed that ther was something to mpeg to keyframes which renders them a bad foil for the xvid k-frame that would have to be constructed from it. many of the source-k-frames differed very strongly from the succeeding p-frames. Looking at it i saw that it was full of film grain and some other very fine details. some 5 frames later all these tiny bits were gone due to the underlying principle of mpeg compression. the fact that it is this way is not surprising but ...
Today, this leads me to the following question. Are mpeg2 keyframes really good keyframes for the destination stream? I don't think so because it cannot really be used as reference for the succeding frames. Yet, as always, I only have a very vague idea of how to compensate for this. One possibility would be to force an additional keyframe about 5 frames after the actual scene change. Another to simply compress kframes stronger (!, this is crazy) than the succeding p-frames (could be done by appropriate qmatrices.
The more intelligent solutions and thought, i am sure will follow below.
Today, this leads me to the following question. Are mpeg2 keyframes really good keyframes for the destination stream? I don't think so because it cannot really be used as reference for the succeding frames. Yet, as always, I only have a very vague idea of how to compensate for this. One possibility would be to force an additional keyframe about 5 frames after the actual scene change. Another to simply compress kframes stronger (!, this is crazy) than the succeding p-frames (could be done by appropriate qmatrices.
The more intelligent solutions and thought, i am sure will follow below.