Classified Vector Quantization Using Reversible Integer Time Domain Lapped Transform for Image Coding

A serious problem in ordinary vector quantization is edge degradation, it can not accurately preserve the edge information. To tackle this problem, a novel classified vector quantization based on Reversible integer Time Domain Lapped Transform (RTDLT) is proposed. Firstly, the image is divided to several blocks and RTDLT is performed on the original image. Secondly, the image block is classified, according to the gradient magnitude within each image block and RTDLT coefficient. Finally, the RTDLT coefficients of different classified block are coded using fuzzy c-means vector quantization. Simulation results indicate that the proposed approach can compress images at lower bit rate and reconstruct images with higher peak signal-to-noise ratio than other approaches such as JPEG2000.