A new efficient image compression technique with index-matching vector quantization

Abstract

A new efficient image compression technique is presented for low cost applications, such as multimedia and video-conferencing. Since address vector quantization (A-VQ) proposed by Nasrabadi and Feng (1990) for image coding, has the main disadvantage of high computational complexity of reordering the address codebook at the transmitter and the receiver during the encoding of each block, we propose a new efficient approach to overcome this disadvantage. The proposed algorithm is based on tree search vector quantization via, multi-path search and index matching in the index codebook, and may achieve a better performance as well as low computational complexity. We theoretically prove that the proposed algorithm is superior to the A-VQ algorithm and experimentally show that a lower bit rate than that of the A-VQ is obtained.