Image Quality Assessment Complemented with Visual Regions of Interest

Abstract

Measurement of visual quality is of fundamental importance to numerous image and video processing applications. This paper introduces an image quality index, visual region of interest weighted quality index (VroiWQI). The index integrates the notion of visual regions of interest with the measurement of structural distortion between the original image and the distorted image to effectively match with human visual system (HVS). The method evaluates the structural distortion between the original image and the distorted image to arrive at the quality index in a local region. These local indices are then weighted based on the visual interest of the corresponding region, characterized by entropy in the local region that emphasizes the texture variance. VroiWQI of the image is calculated by averaging these weighted indices, considering different subsets of visual regions with degree of interest greater than a threshold value. Test results show that VroiWQI correlates well with subjective scores