AN ADAPTIVE HYBRID DPCM/DCT CODING METHOD FOR HIGH QUALITY IMAGE COMPRESSION

Discrete cosine transform（DCT） is one of the key techniques in data compression. DCT can compact the most image energy into only a few of the DCT coefficients. This method is, however, not suitable for the blocks with some edges contained. Transformation of such an edge block will not lead to an efficient energy compaction. An adaptive hybrid DPCM/DCT coding method is proposed for solving this problem. For each incoming block, edge detector switches to the DPCM coder if this block contains sharp edges,and if not it switches to the DCT coder. The DPCM coder adaptively selects from the given set a predictor that results in a minimum prediction error, and then the residuals obtained is encoded with Rice coder. For non\|edge blocks, DCT, run\|length and variable length coding are applied. Experiments show that the proposed method is better than JPEG lossless mode and baseline JPEG performance both on compression ratio and on decoding run\|time when bpp=1—4 approximately.