Improvement of VQ Index Compression with Relative Index Tables

Image vector quantization (VQ) has many current applications, like speech and image compression, and envisioned applications, such as digital watermarking, data hiding and speaker identification. In this paper, we propose a novel lossless compression algorithm, called ISAIAL, to improve the coding efficiency of image VQ. Given an image, VQ produce an index map by quantizing the image block by block, and the main idea of the ISAIAL is to exploit the inter-block correlation in the index domain. We introduce new coding structures, relative index tables, for encoding the index map of VQ. For each code word of the code book, there is a corresponding relative index table that records the indices, which will probably appear in the next block. Two valid methods are introduced for establishing relative index tables.We also show that our ISAIAL is lossless, i.e., the index map can be reconstructed without any distortion. By the experiments, the proposed methods did decrease the bitrate apparently as compared to the conventional VQ andsome commercial lossless compressors, such as rar, zip andarj.