Reduced-Reference Image Quality Assessment via Intra- and Inter-Subband Statistical Characteristics in Reorganized DCT Domain

In this paper, a novel reduced-reference (RR) image quality assessment (IQA) is proposed by depicting the intraand inter-subband statistical characteristics in the reorganized discrete cosine transform (DCT) domain. Firstly, the block-based DCT coefficients are reorganized into a threelevel tree. Generalized Gaussian density (GGD) function is employed to capture the intra-subband characteristics. The difference between the actual coefficient distribution and GGD is depicted by city-block distance (CBD). For the inter-subband characteristics, the mutual information (MI) between adjacent reorganized DCT subbands is utilized to depict the corresponding relationships. By combining the CBD of intrasubband and the MI of inter-subband together, the proposed RR IQA is developed. Experimental results demonstrate that compared to existing methods a smaller number of RR features is required for representing the image perceptual quality. The proposed method outperforms the RR WNISM and FR PSNR, and performs comparably with the FR SSIM.