Novel codeword search algorithm based on Hadamard transform

Abstract

With the aim of overcoming the encoding complexity, a novel and fast neighbor codeword search algorithm for vector quantization in the Handamard transform domain was presented. In the proposed algorithm, firstly the Hadamard transform was applied to all the codewords in the codebook and the input vector. Then the initial match codeword was selected from the codeword whose norm was nearest to the norm of input vector on Hadamard transform. Furthermore, the triangle inequalities with multiple control vectors and the two elimination criteria were utilized to reject mismatch codewords. Finally, the best-match codeword to the input vector was found. Experimental results show that the proposed algorithm has greatly reduced codeword search time and computational complexity under the precondition of good restored image quality.