Progressive image transmission using adaptive multistage vector quantization

A new technique based on adaptive multistage vector quantization (MSVQ) using variable block sizes is proposed for progressive image transmission. The authors employ larger block sizes in the low bit rate range and the block size is gradually decreased when the bit rate is increased. Variable block size is achieved by segmenting from the largest selected block size with a quadtree data structure. The adaptive encoding scheme is also applied to increase the compression efficiency by allocating variable bit rates according to the information detail of each vector block. Classification of blocks with different information detail is achieved by recursively identifying each of the blocks into edge or shade blocks in each stage in order to obtain different classes. Experimental results shows that adaptive MSVQ achieves an excellent reconstructed image quality at rates between 0.15 and 1.0 bpp (bits per pixel)