Vector quantization using matrix decompositions of codebooks for image coding

Methods of using vector quantization combined with singular value decomposition (SVD) and simplified least square estimation (LSE) compensation are proposed. The complexity of computation is decreased by determining a simpler scheme to obtain the same optimal estimated singular values as the LSE. With the assistance of the matrix-decomposed codebooks, edge degradation and compression rate can be improved by using a few least-square-compensated singular values. Avoiding the high computational complexity of SVD during the encoding and decoding procedures and achieving good quality of image coding are the main contributions of the present work.<
>