A new paradigm for image quality assessment based on human abstract layers of quality perception

Abstract

Image Quality Assessment (IQA) plays a central role in many visual processing algorithms and systems. Most of the existing IQA methods try to detect the image distortions’ type and then evaluate its severity. But each image distortion type has its own characteristics, which can harden up the process of quality evaluation. Here, we propose a novel framework for image quality assessment, in which the image is evaluated based on the human judgment process, comprises the three abstract perception layers, namely, illumination sensation, whole content detectability, and details perception. What distinguishes the proposed framework from the existing ones is its independence from the image distortion type. Indeed, we try to assess the amount of success in visual perception, instead of measuring the amount of image distortions. The proposed method has been extensively tested on four benchmark datasets and shows highly competitive performance to state-of-the-art IQA methods.