No-Reference Image Quality Assessment Based on Natural Scene Statistics in NSCT Domain and Spatial Domain

No-reference / blind image quality assessment (NR-IQA) methods aim to predict the quality of distorted images with respect to human perception automatically without prior knowledge of reference images. We present an efficient general-purpose NR-IQA method in this research using natural scene statistics (NSS) in nonsubsampled contourlet transform (NSCT) domain and spatial domain, which we name TDSDQA. Firstly, we analyze the strong correlation of parent-child coefficients and relative coefficients in NSCT domain, and calculate the mutual information (MI) between those coefficients to describe their cor-relation. Second, the structure similarity of those coefficients is determined and utilised to represent the picture structure information statistics. In addition, when used in conjunction with the arithmetic perfect of close by normalised brightness constants in the three-dimen-sional area, we extract 84 statistics features which are sensitive to the presence and severity of image distortion. At last, these features are used to predict image quality scores in a support vector regression (SVR) approach, and the assessment approach is tested on LIVE and TID2008 IQA database. The experimental results show that this method is suitable for many common image distortion types and correlates well with the human judgments of image quality. And that, it has highly competitive performance to other state-of-the-art NR-IQA algorithms in many respects, such as database independence, classification accu-racy, computational complexity and so on.