Image compression using adaptive vector quantizer with segmentation

Abstract

An iterative image compression scheme based on variable block size segmentation and transform vector quantization is proposed. The proposed algorithm efficiently allocates coding bits among regions with different activity characters. The bit allocation involves two steps. First, regions with different levels of activities are allocated different number of bits with more bits being assigned to highly active regions. Second, for those highly active regions that have been segmented into the same categories, bits are reallocated according to the local characteristics. This variable rate coding scheme achieves high quality at low bit rate by more uniformly distributing the distortion and by more efficiently coding the highly active regions. More importantly, implementation of the proposed algorithm is less complex than other comparable schemes.