Structural Similarity Image Quality Assessment Method

Nowadays, with the rapid development of information technology,3G, Sensor, and Multimedia have widely attracted people’s attention and interest. Digital image, as one kind of the Multimedia technologies, is frequently used in many fields in people’s daily life. However, distortions and errors inevitably occur during the transmission, collection, processing and restoration of the image, which most probably causes the change of the image before and after operation, thus results in the decline of the image quality. Since poor image quality has brought about a lot of inconvenience to people’s life, research on image quality assessment methods becomes a hot topic within these a couple of years.Image quality assessment methods can be divided into two branches:subject assessment and object assessment. Then according to the reference degree of the initial image, object assessment can be further divided into three types:integral reference, partial reference and zero reference. This paper highly focuses on integral reference object assessment of image quality. Main work is as follows:(1) Introduce image quality assessment methods, including its research background, research value and research status.(2) Present related knowledge of image quality assessment, such as human visual system, image statistics parameters, performance indicators of image quality assessment method, and etc, to further realize the analysis and modification of the algorithms of current image quality assessment methods.(3) Through the study of image structural similarity, make analysis and modification on the two assessment methods of structural similarity algorithm. One of the methods is based on structural information extraction to realize image quality assessment, which means it first breaks down image, and then use structural similarity algorithm to make assessment. The other method is based on structure direction information, taking advantage of the partial structural direction similarity instead of structural similarity to complete assessment. This paper conducts modification on the operator model of the partial structural direction similarity, with experiments proving that testing results are optimized.(4) Since structural similarity algorithm can’t achieve satisfactory assessing result on faint image, this paper introduces two image quality assessment algorithms using edge structural similarity, and one of them has been improved. The improved performance of the algorithm has been optimized.Whereas, the running speed of this method is slow because of the complexity. Therefore, in this paper,8-direction Sobel operator is applied to replace4-direction Sobel operator on image edge extraction. Then use structural similarity algorithm to make assessment on the gradient image obtained. Finally, the algorithm achieved satisfactory result both in performance improvement and processing speed.