An image quality index based on coefficients of spatial association with an application to image fusion

In the last decade, image quality indices have received considerable attention to quantify the dissimilarity between two images. The codispersion coefficient, commonly used in spatial statistics to address the association between two processes has also been used for this aim. Here we introduce an image quality index (CQmax) that is based on codispersion. This new coefficient is a directional evaluation of the spatial association, and consists of computing the maximum codispersion for a finite set of spatial lags on the plane, which also allows to obtain the direction associated with the maximum codispersion. From the CQmax index, a pseudo-metric that can be used as a cost functional for related optimization problems is defined. We carry out Monte Carlo simulations to explore the performance of the proposed index and its capability to detect directional contaminations. Additionally, we introduce a novel algorithm to restore directionally contaminated images and present an application with real data in the context of image fusion.