Research On Image Quality Assessment Based On LBP And Fractal Analysis

In recent years, with the rapid developed Internet highly used and smart phones abundantly available, image and video information network has become the main data flow and become required and necessary in everyone’s daily life. As a primary way to describe and exchange of information, image data has carried with a wealth of information which brings excellent experience and makes it more convenient for people. During the process of the image acquisition, compression, transmission, storage and even recovery, the quality of image will be badly influenced with the distortion, which greatly affect the accuracy of getting access to the original information. Therefore, how to assess the quality of given image, and to design a model which can measure the degree of image distortion has become an important topic of concern in the field of image processing.The objective image quality assessment method is the main measure of image quality assessment, which consists of three methods according to dependence on original image information in the evaluation process: full reference, reduced reference and non-reference type. Reduced reference model only requires part of the original image information to evaluate the image quality, the aim of the thesis is focusing on image quality assessment algorithm based on Local Binary Pattern(LBP) and fractal analysis.To deal with the basic problem of image quality assessment, this paper use LBP method to extract the local features of distorted images, separating the images into blocks makes the local LBP features more localized decentralized, then obtain the geometrical characteristics of LBP image using fractal analysis method, highlighting the different characteristics of located information, we use these features to capture visual perception changes which caused by the distortion, then comes out with the framework for reduced reference image quality assessment based on LBP and fractal analysis. We implement the algorithm which described above on the international mainstream image quality assessment dataset TID2008, LIVE, CSIQ, IVC, MICT, A57 and WIQ to verified the proposed algorithm, the experimental results show that the objective quality scores of from the algorithm can fit with the subjective evaluation values in good way, the algorithm can accurately reflect the image quality of the human eye’s subjective feelings.