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