Perceptual color image quality metric using adequate error pooling for coding scheme evaluation

We propose a visual quality bivariant criterion for the evaluation of coding scheme. This criterion is based on human visual system properties to get the best correspondence with human judgments. Contrary to some others objective criteria, it doesn't use any information on the type of degradations introduced by coding schemes. We use two main stage. The first one computes the visual representation of errors distributed over color, spatial and frequency dimensions between two image. This stage is entirely based on results from psychophysics experiments conducted in the laboratory. The second stage computes the error pooling over color, frequency and space to get the overall visual quality between two images. Since we have previously showed importance of this stage, we propose an original approach extended here to color images. In particular, we point out in this paper how to take into account color information in a visual quality criterion. We compare results of the criterion with human judgments on a database of images distorted with 3 types of compression schemes (JPEG, JPEG2000 and a ROI-based algorithm using metrics defined by the Video Quality Expert Group. Results indicate that criterion provides good prediction accuracy, monotonicity and consistency. Proposed approach is so a useful alternative tool for coding image searchers.