An image quality assessment algorithm based on feature selection

Image Quality Assessment(IQA) is of fundamental importance to numerous imaging and video processing applications. For most of the applications, the perceptual meaningful measure is the one which can automatically assess the quality of images or videos in a perceptually consistent manner. However, most commonly used IQA metrics are not consistent well with the human judgments of image quality. Recently, the SSIM metric which takes people's visual characteristics into consideration performs much better than the traditional PSNR/MSE. But the defects of it still exit on some specific kinds of distortions. A new algorithm of IQA based on feature selection is proposed in this paper. Local gradient entropy and phase congruency are added to the SSIM framework. Through in-depth feature selection and definition plus better pooling strategy, this algorithm performs much better in LIVE datasets.