Information-based reduced reference image quality assessment incorporating non-tensor product wavelet filter banks

Abstract

To achieve the better performance of image quality assessment, this paper develops an informationbased scheme. First, visual maps that contain visual information are extracted by non-tensor product wavelet filter banks at different scales. Subsequently, to model visual difference (VD) caused by image degradations, the variation of information entropy between distorted and reference images is employed. Then the final visual quality is calculated with the VDs at each scale. In experiments, evaluation results show that it performs very well in terms of high accuracy of image quality measurement as well as low data rate.