Pattern-based side match vector quantization for image compression

The contiguous pixels are often similar in images. Hence side match vector quantization (SMVQ) can use the contiguous blocks to predict the other bloc ks. SMVQ has, however, a derailment problem that the error of the current s tate may result in errors in the next states. To remedy this problem, the p attern-based side match vector quantization (PSMVQ) is used in this paper. Unlike SMVQ, PSMVQ partitions an image into many patterns first. A pattern is a group of n x n blocks. Next, in each pattern, PSMVQ seeds one or more seed blocks, whose corresponding codewords are found by the full search met hod and employs the side match method to predict the residual blocks. The e xperimental results show that PSMVQ has three advantages, namely (a) that i t is a faster method to encode an image than SMVQ, (b) that it has the abil ity to reduce the bit rate of an image effectively and (c) that the reconst ructed image has a high image quality.