Objective Quality Evaluation Method Of Stereo Image Based On Edge And Local Matching

Recently, with the rapid development of image coding and stereo imagingtechnology, the stereo image processing technology has been the point of focus, andhas become a popular research point at home and abroad. Compared with subjectivefeeling of2D images,3D images are better at giving people a feeling of reality andshocked telepresence. However, the amount of data raised substantially in the processof acquisition, transmission and storage of stereo images. How to ensure the quality ofthe stereo images under the precondition that the amount of data decreased willinfluence the development of stereo image processing technology directly. Therefore,it is of great importance to build an accurate and effective mechanism to evaluate thequality of stereo images.Through deep research on plane images quality assessment methods and fullconsideration about human visual characteristics, this paper proposes a method toevaluate the stereoscopic images from two aspects: quality and stereo sense. As toquality assessment, a method that takes the edge information into account is presented,which is based on the plane image quality assessment method. As for stereo sense, astereo matching method is used for disparity map extraction, then calculate thestructural similarity of each disparity map. Finally, we proposed a reasonable totalevaluation metric by appropriate weighted method.In experiments, quality of116images with common distortion types, such as addnoise, compression, fuzzy and translation distortion, was evaluated. The results showthat the proposed stereo image quality method get a better result than the methodbased on absolute difference graph PSNR and SSIM, the method proposed in thispaper can reflect the human visual perception, and is able to reflect the stereoscopicimage quality and stereo sense, what is more, it can make a contribute to thedevelopment of stereo image quality evaluation technology.