Multimedia compression is a general term referring to the compression of any type of multimedia, most notably graphics, audio, and video.
Because multimedia typically derives from data sampled by a device such as a camera or a microphone, and because such data contains large amounts of random noise, traditional lossless compression algorithms tend to do a poor job compressing multimedia. Multimedia compression algorithms, traditionally known as codecs, work in a lossy fashion:
- Transform the data according to a model designed to reduce sample-to-sample correlation, concentrating the important signal in a few data values.
- Quantize the data, most of which has become noise. Some codecs use a scalar quantizer followed by run-length encoding; others use vector quantization.
- Use entropy coding such as Huffman coding to reduce the number of bits that the most common values use.
Multimedia compression has become the primary focus of compression research, primarily in a search for more efficient models. It is the most important part in video coding formats.