SMIM: Superpixel mutual information measurement for image quality assessment

Abstract

The image quality assessment (IQA) is a fundamental problem in signal processing that aims to measure the objective quality of an image by designing a mathematical model. Most full-reference (FR) IQA methods use fixed sliding windows to obtain structure information but ignore the variable spatial configuration information. In this paper, we propose a novel full-reference IQA method, named “superpixel normalized mutual information (SMIM)” based on the perspective of variable receptive field and information entropy. First, we find that consistence relationship exists between the information fidelity and human visual of individuals. Thus, we reproduce the human visual system (HVS) to semantically divide the image into multiple patches via superpixel segmentation. Then the weights of each image patches are adaptively calculated via its information volume. We verified the effectiveness of SMIM by applying it to data from the TID2008 database and data generated using some real application scenarios. Experiments show that SMIM outperforms some state-of-the-art FR IQA algorithms, including visual information fidelity (VIF).