A Singular Value Decomposition Based Objective Quality Assessment Model on Stereoscopic Images

Removing visual redundancy with binocular suppression in human visual system is an effective way to improve the encoding efficiency of stereoscopic video.Based on binocular suppression, an objective quality assessment model on stereoscopic images is proposed in this paper.By analyzing the visual quality of stereoscopic images, statistical features of binocular suppression are explored to establish and optimize the model.According to the singular values of the images with strong stability, singular value decomposition (SVD) is adopted to extract features of stereoscopic images in the model.Finally, a subjective study on the performance of the proposed model with VQEG standards is conducted, involving 370 images distorted by Gaussian blurring, white Gaussian noise, JPEG and JPEG2000 compressions, four performance indicators of the proposed model.Experimental results show that the Pearson's Correlation Coefficient (CC) value, the Root Mean Square Error (RMSE) value, the Spearman's Rank-Order Correlation Coefficient (SROCC) value, and the Outlier Ratio (OR) value are 0.955, 3.737, 0.906 and 0.811% respectively.The results indicate that the proposed model can predict human visual perception well.