Image coding based on visual vector quantization

The authors present a novel image coding technique which explicitly aims to retain perceptually important features of the image. The system exploits the psychovisual properties of the human visual system, in that it identifies perceptually important vector patterns (referred to as visual vectors) within small blocks of the image, and codes these using a scheme similar to vector quantization (VQ). Thus the system is referred to as visual vector quantization (VVQ). One of the main advantages of the new VVQ scheme over traditional VQ is that it is computationally much simpler, and yet its performance is comparable to that of traditional VQ. Experimental results are presented which demonstrate that the new technique can achieve visually satisfactory image reconstruction at a compression ratio as high as 17:1. Comparison of peak signal to noise ratio (PSNR) values shows that the performance of the new technique is comparable to that of other VQ-based image coders at a similar bit rate.<
>