Research On Image Quality Assessment Based On Structural Information

Researches on image quality assessment are meaningful and basic for the image processing engineering。Traditional image quality measures: distortion oriented and human visual system (HVS) modeling oriented cannot correlate well with human perception or have very complex implementations .Recently, Wang proposed a novel image quality assessment method based on structural similarity (SSIM), which is easy to implement and correlates well with human perception. SSIM attracts a lot of attention and has been used in some image processing systems.With the developing of digital image processing technology, Contourlet-domain and WBCT-domain image processing have become an important component of image processing. It is very important to find an image quality assessment model that can be used in image processing algorithms based on Contourlet and WBCT transform. In this paper we introduce the SSIM into the Contourlet-domain and WBCT-domain, and propose three image quality assessment methods that are called the CT-SSIM-ABS ,CT-SSIM-DIR and WBCT-SSIM.SSIM is proved to have some unreasonable elements for some image blocks through theoretical analysis and experimental proof and JND can get a good performance for these blocks.Through deeply analysis, a new method of image quality assessment based on SSIM and JND—(JND-SSIM) is proposed in this paper.The main contents of this paper are as bellows:1. Based on the research of image structural properties in Contourlet domain, two image quality assessments methods that are called the Contourlet Transform based Structural Similarity CT-SSIM (CT-SSIM-ABS and CT-SSIM-DIR.) are proposed in this paper The result shows that the proposed CT-SSIM (CT-SSIM-ABS and CT-SSIM-DIR.) obtain a very good evaluation result and performs better than the SSIM in every respect.2. Based on the research of image structural properties in WBCT domain, an image quality assessments method that is called the WBCT Transform based Structural Similarity (WBCT-SSIM) is proposed in this paper. The result shows that the proposed WBCT-SSIM correlates well with the judgment of human observers. 3. Based on the research of SSIM and JND, an image quality assessments method that is called JND-SSIM is proposed in this paper. The result shows that the proposed JND-SSIM is more accurate and has better correlation with the subjective assessment of human beings for JPEG and JPEG2000 distorted images.