Enhancing the performance of vector quantizers for data compression and pattern matching using sorted codebooks

Abstract

Vector quantization (VQ) has found application in very low-rate speech and image encoding, together with content-based multimedia retrieval. In order to reduce the average distortion, large codebooks are required to store sufficient representative source vectors. The fundamental drawback with large codebooks is the search time required at the encoder. In order to reduce this search time, several methods have been proposed in the literature. This note describes an improvement on the so-called triangle-inequality based search. Experimental results are presented to demonstrate the usefulness of the method, which is able to reduce search times to as little as 10-20% of the full-search equivalent method under certain circumstances.