摘要 |
Encoding or decoding images using a palette coding mode in which each block of pixels has an associated palette which gives a correspondence between pixel values and indexes. The blocks of pixels are encoded as a block of indexes using the palette. At the decoder, the method comprises decoding encoded data of two self-contained coding structures having the same hierarchical level, using a palette coding mode. The palette associated with each processed pixel block is predicted from a palette predictor 1215 using prediction information from the bitstream 1200. A palette predictor initializer 1221 associated with a coding structure hierarchically above the two self-contained coding structures is obtained from the bitstream and used to initialize the palette predictor for each self-contained coding structure using the same obtained palette predictor initializer. The palette predictor initializer may be dependent on the whether or not the image is monochrome as signalled by a flag in the bitstream. The initializer may be defined for a picture, sequence or tile and signalled in the picture parameter set (PPS) or sequence parameter set (SPS). The self contained coding structures should be independently encoded and may be slices. Preferably, a palette predictor is built 1214 from the union of the previous palette and the previous palette predictor. |