Research On Objective Quality Assessment Of Stereoscopic Vision Information

Stereo technology has become national hottest issues in visual and infrmation processing fields, because of its wide application prospect. In the processing of frontier applications,e.g., the design, manufacture , transmission and consumer market, stereo image and video processing business inevitably need image quality assessment. How to evaluate stereo image quality becomes a fundamental and challenging job in the field of image processing.In this paper, we explore which consists of key technologies about the human visual system's forming mechanism, processing elements, effects on image quality of three-dimensional visual characteristics. This paper has proposed three different methods to evaluate stereo image quality.Firstly, based on the study of physiological and medical stereo visual experiments'consequences and existing visual theory, we analyze image information processing mechanism of human visual system (HVS). The affect of luminance, contrast, resolution and chroma to two-view stereo imaging system has been studied by lots of experiments in order to find image attribute's effect on stereoscopic imaging system.Secondly, further through experiments we validate the importance of stereo image's special characteristics different from two-dimensional image visual in the visual image quality assessment, and propose stereo image objective quality evaluation model which adopts contrast sensitivity, masking and multi-channel decomposition characteristics. This model is applicable in stereo compressed image's quality evaluation, and further gives distortion deviation degree index compared with the original stereo image. This paper also figures out the correlation between image quality and stereo perception,the future research can consider combining the two evaluation of estimates into one or declining the correlation.Thirdly, because of high complexity and difficult modeling of human vision accurately .The method further explores binocular visual difference in stereo imaging, extract two classes feature information closely related to stereo image quality, that is absolute difference and parallax. Model build use the most efficient and wide application peak signal-to-noise ratio and structural similarity measures. The method can apply for kinds of distortions. Experimental results are consistent with the subjective analysis result.Finally ,the paper proposes joint psychological and physiological image quality assessment,which take clues of the psychological stereoscopic vision. into account. Two models with different complexity are presented comprising physical clues of psychological stereoscopic vision. The experimental results testify that the method can work well in stereo image quality evaluation. The objective evaluation results can keep step with the subjective evaluation results.