Huffman coding is a lossless data compression algorithm. Huffman coding uses variable length code to represent a symbol of the source (eg a character in a file). The code is determined from an estimate of the probabilities of appearance of the source symbols, a short code being associated with the most frequent source symbols.

A Huffman code is optimal in the sense of the shortest length for encoding by symbol, and a known probability distribution. More complex methods of probabilistic modeling of the source allow for better compression ratios.