A complete statistical evaluation of state-of-the-art image quality measures

Abstract

A quantitative predictive performance evaluation of seventeen full-reference image quality assessment metrics has been conducted in the present work. The process has been made over six publicly available subjectively rated image quality databases for four degradation types namely JPEG and JPEG2000 compression, Gaussian blur and Gaussian noise. Results show that there is no significant impact of the choice of image quality databases on the evaluation of the different full reference image quality metrics. These latter are found to fall into the same range of performance in terms of their correlation to the subjective human ratings, accuracy and monotonicity.