Digital Signal Processing Reference
In-Depth Information
becomes the motion-compensated prediction block for the
current macroblock.
5. The difference block and the motion-compensated predic-
tion block are summed together to produce the reconstructed
macroblock. Macroblock by macroblock, the video frame is
reconstructed.
6. The reconstructed video frame is buffered, and can be used
to form prediction blocks for subsequent frames in the GOP.
The buffered frames are output from the decoder in sequential
order for viewing.
The decoder is, in essence, a subset of the encoder functions,
and fortunately requires much fewer computations. This
supports the often asymmetric nature of video compression. A
video source can be compressed once, using higher-performance
broadcast equipment, and then distributed to many users, who
can decode using much less costly consumer-type equipment.
