A Codebook Modification Method of Vector Quantization to Enhance Compression Ratio

Abstract

AbstractThe performance of an image compression algorithm is based on the amount of compression ratio achieved keeping the visual quality of the decompress image up to the mark. In an image compression algorithm, two performance measurement parameters, compression ratio, and visual quality of the decompress image are inversely proportional. So, improving the compression ratio of the compression algorithm, keeping visual quality of the decompress image as close to the original is a major challenging task. Vector quantization is one of the widely used lossy image compression techniques found in literature. The compression ratio of this algorithm basically depends on the size of the index matrix and codebook generated during the process. In this present work, a new technique is proposed which represents each and every value of the codebook by 5 bits instead of 8 bits that means it reduces the amount of memory required to store the codebook by 37.50% and which increases the compression ratio of the algorithm significantly. The proposed method is applied on many standard color images found in literature and images from UCIDv.2 database. Experimental results show that the proposed method increases the compression ratio significantly, keeping the visual quality of the decompressed image almost same or slightly lower.KeywordsCodebookCodevectorCompression ratioImage compressionPSNRSSIMVector quantization