1. Another name for data compaction, although, strictly, data compaction is lossless while data compression need not be (see def. 2).
2. Removal from a file or data stream of information that may be redundant either in the sense of information theory, or in the sense that the retention of precision, definition, or some similar measure of quality is less important than the necessity to abbreviate the data. In the former sense, the abbreviation is lossless, while in the latter sense it is lossy. Compression permits either or both kinds, and so its effects are not always exactly reversible.
Generally, in the context of discrete and continuous systems, the output from continuous systems, if it is to be abbreviated, can often be lossily compressed. This is notably the case with sound, and with halftone and coloured images. See also image compression.