Quality assessment of despeckling filters based on ratio images and visibility graphs

Abstract

We present a new approach to assess the quality of despeckling filters using ratio images. Under the multiplicative model, the ratio of the original and perfectly despeckled image is a collection of independent identically distributed Gamma variates. We map such ratio images with the Image Horizontal Visibility Graph transformation and analyze their structure in this new domain. We obtain a powerful assessment by combining this approach with traditional metrics like MSE and SSIM. The proposal also supports filter parameter optimization.