Partial distortion search algorithm based on two-step d _min updating

Abstract

The encoding process of vector quantisation is computationally very expensive. To reduce distortion computation between the input vector and candidate codewords, the well-known partial distortion search (PDS) algorithm has been proposed. To reduce further its computational cost, a new PDS algorithm is proposed by making current d min decrease as quickly as possible. Experimental results have demonstrated the effectiveness of the proposed PDS algorithm in terms of both the computing dimensions and arithmetical operations per input vector.