An improvement of fast search algorithm for vector quantization

One of the major issues of vector quantization encoding is to reduce the computational complexity from searching the best-matched codevector, so as to shorten the search time. In this paper, we extended the partial-sum algorithm to exclude much more codevectors from doing the complex Euclidean distance computation. An optimal trade-off has been found between the computational complexity and the search space. According to the experimental results, our algorithms can significantly reduce the number of redundant codevectors, which do not need to do the distance computation, while obtaining the same encoding quality as that of full search algorithm.