Color quantization of images based on human vision perception

Abstract

In many color-imaging applications, it is desirable to represent the image with as few different colors as possible without loss of visual quality. In this paper, we address the problem of color quantization of images for display on an 8-bit color monitor. We have developed a subjective distortion measure which accounts for the higher sensitivity of the human visual system to errors in low activity areas compared to errors in high activity areas. We propose a color quantization algorithm based on this distortion measure applied to a cubic prequantizer and to a weighted tree structured vector quantizer (TSVQ) to obtain the final 8-bit color palette. The resulting images displayed with our 8-bit color quantization algorithm show higher visual quality than the images displayed using dithering, error-diffusion and other existing color image display software. >