Image quality assessment using Histograms of Oriented Gradients

Since it is commonly believed that human visual perception is highly adapted for extracting structural information from the scene, many gradient-based image quality assessment (IQA) metrics were proposed. The main research focus in this theme is about designing the computational models of gradient similarity to measure the changes of image quality. In this paper, we turn our attention to a different question: how to estimate the visual importance of different regions in one image using the gradient changes to improve the performance of existing IQA metrics. A novel gradient-based full reference IQA is proposed based on combining Histograms of Oriented Gradients (HOG) with the structural similarity (SSIM) index. Extensive experiments conducted on the LIVE image database show that the proposed HOGM approach achieves much higher consistency with the subjective evaluations than a number of competitive IQA algorithms.