Robust approach for color image quality assessment

This paper presents a visual color image quality metric assessment with full reference image. The metric is highly based on human visual system properties in order to get the best correspondence with human judgements. Contrary to some others objective criteria, it doesn't use any a priori knowledge on the type of introduced degradations. So the main interest of the metric is on its ability to produce robust results independently of the distortions. The metric can be decomposed into two steps. The first one projects each images, the reference one and the distorted one, in a perceptual space. The second step achieves the pooling of errors between perceptual representation of two images in order to get a score for the overall quality. Since we have shown that these two steps have equivalent importance regarding metric performance, we have particularly paid attention in correct balancing when designing the two steps. Especially, for the second one, that is generally limited to poor consideration in literature, we have developed some new original approaches . We compare results of the metric with human judgments on images distorted with different compression schemes. High performances are obtained leading to assure that the metric is robust, so this approach constitutes an alternative useful tool to PSNR for image quality assessment.