Stereoscopic Video Quality Objective Evaluation Based On Visual Saliency

3D video has the characteristic of large data,witch bring in a huge challenge to 3D video transfer.To solve this problem,people introduce video compressing methods witch usually have high compressibility.Compressing the 3D videos can often lead to the degradation of overall 3D video quality.The degradation can affect viewer experience,sometimes can also impair the visual system of viewers.So,stereo video quality assessment is very impotent.In stereo video the motion information is a fatal feature.When we view the 3D video which has two different viewpoints,the binocular fusion which appear in human vision system make us feel the stereo sense.We considered this characteristics,and proposed three new stereo video quality assessment methods based on motion saliency and binocular fusion.First method is named MS_VQM,second method is named MSFW_VQM and third method is named MSFWMR_VQM.Last two methods are come from improving first method,which more conform to human visual cognitive habits.In the method of MS_VQM,we first segment the pictures into different segmentary areas in 3D video according to their motion significant.Then we gave them different weights according to their intensity of motion significant.Secondly,we assessed the quality of segmentary areas with the method of SSIM.In order to get a global quality of everyone picture,we weighted the segmentary area quality with their motion significant weights.Finally,we constructed the quality of 3D video according to binocular fusion which appear in human vision system.MSFW_VQM which based on the MS_VQM uses adaptive weight adjustment algorithm to determine different frame weights of stereo video.The application of adaptive weight adjustment algorithm can not only improve the performance of assessment method,but also find the key frames that can decision the quality of stereo video.In the end of this paper,I establish a visual multichannel model with the wavelet.MSFWMR_VQM is built through combining visual multichannel model and MSFW_VQM.In this paper,150 different contents and different degree of distortion stereo videos are used.The spearman correlation coefficient between objective assessment results and subjective assessment results are 94% for MS_VQM,96% MSFW_VQM and 98% for MSFWMR_VQM respectively.The experimental results shown that these methods can not only reflect human visual characteristics,but also assessment the quality of 3D video accurately.