New no-reference stereo image quality method for image communication
Ying Wang, Yu Mei, Fen Chen, Gangyi Jiang · 2016
According to the effect of human vision on stereo image, a new no-reference stereo image quality assessment method with binocular perception is proposed for image communication. Firstly, convergent cyclopean images are constructed by using the energy gain control model to simulate binocular perception. Then, curvelet domain scale coefficient and energy distribution features are extracted from the convergent cyclopean image, meanwhile, generalized Gaussian distribution features and logarithm normal distribution features are also generated from disparity map and uncertainty map. Finally, the objective score of stereo images is obtained by establishing the relationship between the features and the subjective scores. Experimental results show that for the proposed method, its Pearson linear correlation coefficient reaches 0.94, and its Spearman rank correlation coefficient is close to 0.94, better than some classical image quality assessment methods. It indicates that the proposed method can accord with the human visual perception.