Abstract:

An adaptive fast search algorithm for codewords is presented based on the wavelet transform ,with the aim of overcoming the encoding complexity in the vector quantization(VQ)process.In the proposed algorithm,the squared Euclidean distance is used as the distortion measure of the quantization,and a rational initial matched codeword is chosen for the input vector.The triangle inequalities with multiple control vectors and the characteris-tics 0f the vectors in the transform domain are then utilized to reject non-matched codewords．The best-matched codeword is finally obtained by gradually reducing the search space.Simulated results show that,with little pre-cal-culation and memory cost,the proposed algorithm can significantly reduce the computational complexity of VQ en-coding,and the quality of the encoded image approaches to that obtained by the full search algorithm.

Key words: wavelet transform, vector quan tization, codeword search