THE DIVISIVE NORMALIZATION TRANSFORM BASED REDUCED-REFERENCE IMAGE QUALITY ASSESSMENT IN THE SHEARLET DOMAIN

Reduced-reference (RR) image quality assessment (IQA) metric aims to employ less partial information about the original reference image to achieve higher evaluation accuracy. In this paper, we propose a novel RRIQA metric based on the divisive normalization transform (DNT) in the discrete nonseparable shearlet transform (DNST) domain In this metric, the coefficients in the DNST domain are normalized employing the Gaussian scale mixture statistical model, and then the marginal distribution of the coefficients changes into approximate Gaussian distribution. A set of statistical features is extracted from DNT-domain representations of the reference and distorted images, respectively. The weighting of these features is performed based on the characteristics of the human visual system. Structural similarity comparison of these features is conducted as an objective quality score of the distorted image. The proposed metric is evaluated on the public LIVE database and demonstrates fairly good performance across a wide range of image distortions.