Stereoscopic Image Quality Assessment Method Based On Binocular Fusion

Stereo images will inevitably be degraded in the process of collection,transmission and display,lowering visual effect.Stereo image quality objective assessment technology has thus become one of the hotspots of research due to its conciseness and efficiency.In this paper,a proper stereo image quality objective assessment model has been proposed,based on a full understanding of the human visual system.The model embraces two aspects: stereo image quality distortion measure and stereo sense.In terms of distortion measure,first,monocular vision model has been establish and reference and distortion images are computed through simulating Masking effect,Multichannel effect and Contrast sensitivity.Next,frequency domain has been introduced into structure similarity function,with the consideration that existing structure similarity cannot properly assessing blurred image quality.Finally,the frequency domain structure similarity of the whole stereo image has been calculated with Binocular Fusion,instead of calculating monocular image quality.In terms of stereo sense,a stereoscopic sensation assessment method based on high frequency information has been put forward in the paper,because stereoscopic vision is mainly caused by the different visual effect of the right and left eyes and these differences,most of which are high frequency information,are mainly obtained from the edge of the image.First,high frequency information are extracted from stereo image through discrete cosine transform,and Absolute disparity map are calculated.Then image manipulation enhanced algorithm is used to make the high frequency information distinguishable.Last,the frequency domain structure similarity of the Absolute disparity map is calculated as the stereoscopic sensation index.540 different degradation test images have been assessed by the stereo image quality assessment model to verify the correlation between subjective and objective evaluation results.The experimental result shows that in terms of distortion measure,the method based on Binocular Fusion proposed in this paper has higher correlation between subjective and objective result,compared with the method using weighting left and right image quality.In terms of stereoscopic sensation,the Pearson Correlation Coefficient of the assessment method is 0.892155,thus this method can accurately calculate the stereoscopic sensation.In a word,the method proposed in this paper has high consistency with the subjective score and therefore has certain reference value for the development of stereo image quality objective assessment technology.