Quality assessment measure based on image structural properties
David G. Asatryan, Karen Egiazarian · 2009
In this paper, a new objective quality assessment measure for images is proposed based on statistical structural image analysis using Weibull model and Cramer-von Mises statistics. It is estimated via proximity of parameters of empirical distributions of a gradient magnitude of pixel intensities. Results of numerical experiments demonstrate that the proposed measure is more adequate to perception by human visual system than the usual pixel-by-pixel measures. Unlike other quality assessment measures, a new one can be used on not well-aligned images or on images having different sizes.