A Fast Partial Distance Search Algorithm for VQ Codebook

Vector quantization (VQ) is an efficient technique for image compression. However, one of the most serious problems of VQ is the heavily computational load for searching the closest codeword to the input vector in the codebook. To overcome this problem, we propose a simple and fast partial distance search algorithm for VQ codebook by using the sum-norm pyramid structure of codewords. The proposed scheme can eliminate many codewords that are impossible to be closest to the input vector in the codebook. Therefore, it can save a great deal of computational time for the image encoding process of VQ. Experimental results show that the proposed scheme significantly reduces the encoding time of VQ and obtains approximately the same image quality as the full search algorithm (FSA). It also shows that our scheme outperforms the recent fast codeword search algorithms in terms of the coding time.