Weighted feature similarity - A nonlinear combination of gradient and phase congruency for full-reference image quality assessment

Abstract

In the paper the modified Feature Similarity metric has been discussed which is based on the nonlinear combination of two elements being the basics of the recently developed Feature Similarity metric for full-reference image quality assessment. Nevertheless, the influence of the gradient magnitude and phase congruency, used as two main elements of the metric, on the perceived quality is not necessarily equal. For this reason some experiments have been conducted in order to propose the weighting coefficients, applied as the local exponents, increasing the rank order correlation coefficients with subjective quality evaluations. The verification of the obtained results has been conducted using 5 "state-of-the-art" benchmark databases and the obtained weighted FSIM metric's performance results are better for all of them. © 2013 Springer-Verlag.