Extended mean-distance-ordered search using multiple l1 and l2 inequalities for fast vector quantization

Mean-distance-ordered search (MOS) algorithms were proposed as a fast and efficient method for vector quantization (VQ) encoding. However, the efficiency of MOS algorithms is limited because they use only one inequality in reducing the search region. Therefore, we propose an extended MOS algorithm using multiple anchor vectors which exploits multiple l₁ and l₂ inequalities to reduce the search region further. We also describe a greedy algorithm for selecting an anchor vector set that reduces the computational cost of the extended MOS algorithm. For eight test images, the number of operations required in the extended MOS algorithm was 66.2% of that in the previous MOS algorithm on the average when the codebook size is 256, while producing the same encoding quality to that of the full-search VQ