MP3 compression of sound attempts to reject sounds that could not be heard well or at all because they are close in frequency to louder sounds or follow a louder sound closely in time. It also reproduces louder sounds with more precision than smaller ones. All of these measures allow less information to be saved in the recorded/transmitted data and therefore result in fewer bits in the data. Conventional lossless compression is also appended to MP3 after the perceptual compression.
Perceptual compression of video typically encodes more information for luminescence ("brightness") than for color because the human eye is more sensitive to luminescence changes than color changes. Other peculiarities of human vision are taken advantage of as well in different ways by different compression schemes.
Because perceptual compression is ad hoc, every compressor is likely to produce somewhat different compressed data. In the case of MP3, the specification does not define the compression, just what a compressed file has to look like in order for a decompressor to work. Presumably this will be true of most (all?) perceptual compression schemes.
Different compression schema may work better with different material and all may have trouble with certain cases. Other problems may intrude such as difficulties in recompressing data that has previously been compressed then decoded, or difficulty in picking up streamed compressed data in midstream.
Perceptual compression is used with audio and video data. Presumably in the future it will be used with other types of sensory data.
Return To Index Copyright 1994-2002 by Donald Kenney.