| 摘要 |
A method for encoding (and, conversely, decoding) an image comprising blocks of pixels, comprises: transforming (e.g. using DCTs) pixel data blocks into transformed block coefficients, and quantizing the transformed (DCT) coefficients into quantized symbols; obtaining alphabets Ai of symbols, each alphabet defining possible symbols of an associated quantized transformed block coefficient; for each alphabet, obtaining a probabilistic distribution P of the possible symbols; using the probabilistic distributions to group 1010 at least two alphabets into a new alphabet, the new alphabet defining, as symbols, the combinations of possible symbols of the (at least) two transformed block coefficients associated with the grouped alphabets; obtaining binary entropy (e.g. Huffman, Shannon) codes for the possible symbols of each remaining alphabet; and encoding the quantized symbols using obtained binary entropy codes. Alphabet grouping may comprise grouping alphabets having the smallest and largest entropy; alternatively, alphabets may be grouped to maximize a reduction in distance between the entropy of the alphabets and that of the binary codes constructed from the alphabets. The method may be applied to base and enhancement layer processing. Huffman code entropy is improved to be closer to theoretical probabilistic distribution entropy. |
