A visual model for estimating the perceptual redundancy inherent in color images

An efficient compression algorithm that is transparent to human visual perception is required for representing high-quality color images as the storage or transmission bandwidth resources are limited. Since the human eyes are not perfect sensors for discriminating small differences of color signals, there exist spaces of perceptual redundancy (PR) in color images. This paper presents a visual model for estimating PR of color images in terms of a triple of values for each color channel as a visibility threshold of distortion, or just noticeable difference (JND). The visual model is built on defining a perceptually indistinguishable region for each color in a color space by mapping colors, which are barely distinguishable from the target color in a perceptually uniform color space (PUCS), to the target color space. To justify the proposed color visual model, the estimated PR is utilized to improve the performance of JPEG-LS. Simulation results show that the perceptual performance in terms of the inspected visual quality and the amount of perceivable errors of the perceptually tuned JPEG-I-S coder is superior to that of the un-tuned coder.