A fuzzy rule-based system for the subjective quality evaluation of distorted images

摘要

A quality metric for distorted images is presented which is adapted to human perception. For this purpose, a model of the human visual system is used, considering various physiological phenomena like the sensitivity to background illumination levels and to spatial frequencies. The image is divided into characteristic parts, in which the errors are assessed separately. For these values a fuzzy rule-based system calculates a grade which is better adapted to human perception than other common error measures. The output value of the rule-based system is given on an absolute classification scale between the grades 1 (excellent) and 6 (unusable).