A Mpeg2 stream is compressed both spatially and temporally. Each frame of video is compressed using a method similar to JPEG and MPEG1 compression. Additionally, if two successive frames of video data display essentially the same picture, only the differences between the two frames need be described. The process of reconstructing frames of video from these descriptions is known as motion compensation.
Motion compensation is done on the level of the Macroblock. Motion compensation refers to the fact that the locations of the macroblock-sized regions in the reference frame can be offset to account for local motions. The macroblock offsets are known as motion vectors.
A Mpeg2 stream consists of the following frame types
- Intra (I) Frames, which are competely described,
- Predicted (P) Frames, which are described in reference to the previous I or P Frame,
- Bidirectionally predicted (B) frames, which are described in reference to two I or P frames -- one in the future, and one in the past.
Many video cards support motion compensation in hardware. In many cases, however, the proper documentation has not yet been released.
See InverseDiscreteCosineTransform
EditText
of this page
(last modified 2001-02-21 21:16:16)
FindPage
by browsing, searching, or an index
Or try one of these actions:
LikePages,
SpellCheck
MoinMoin 1.108, Copyright © 2000-2001 by Jürgen Hermann
config = 0.000
total = 0.250
imports = 0.170
send_page = 0.070
