Perceptually Optimal JPEG2000 Coding of Color Images

Abstract

Human eyes are ultimate receivers of color images, while the human visual system (HVS) has limited sensitivity in discriminating color signals of small differences. To the human vision, there exists in color images a certain amount of perceptual redundancy. By accurately measuring the perceptual redundancy inherent in color images and shaping coding distortion into the perceptual redundancy, color images are expected to be represented more efficiently. This paper presents a visual model for measuring the perceptual redundancy inherent in color images, and an approach to perceptually optimize the JPEG2000 coders for color images by making coding errors the perceptual redundancy. The visual model employs Euclidean metric in the uniform color space to find for each pixel the range of colors which are perceptually indistinguishable from the color of the pixel in the color space it is represented. The JPEG2000 coder is refined by minimizing the perceptible distortion involved in the rate control of the compressed image in the YCbCr color space. Simulation results show that, in both cases, the performance of the perceptually tuned coder is superior to that of the un-tuned coder in terms of the bit rate required for achieving the same visual quality