Novel hybrid classified vector quantization using discrete cosine transform for image compression

摘要

We present a novel image compression technique usingna classified vector Quantizer and singular value decomposition fornthe efficient representation of still images. The proposed method isncalled hybrid classified vector quantization. It involves a simple butnefficient classifier-based gradient method in the spatial domain,nwhich employs only one threshold to determine the class of the inputnimage block, and uses three AC coefficients of discrete cosinentransform coefficients to determine the orientation of the block with-nout employing any threshold. The proposed technique is bench-nmarked with each of the standard vector quantizers generated usingnthe k-means algorithm, standard classified vector quantizernschemes, and JPEG-2000. Simulation results indicate that the pro-nposed approach alleviates edge degradation and can reconstructngood visual quality images with higher peak signal-to-noise rationthan the benchmarked techniques, or be competitive with them.