In this paper, an objective assessment model based on three-component weighted structural similarity is presented for stereoscopic image. Since different distortions affect details of stereoscopic image differently, the image is classified into three kinds of regions including edge, smooth and texture regions so that different weights are associated to different regions to assess distorted stereoscopic image. The experimental results show that the assessment scores are consistent with subjective stereoscopic perception when stereoscopic image is degraded by blur, white noise, JPEG and JPEG 2000 process. Thus the objective assessment results can well reflect the image quality and stereoscopic perception.
Stereoscopic imaging technology has aroused a great concern recently due to the increasingly wide range of stereoscopic applications. Stereoscopic image quality assessment is of vital importance to evaluate the performance of three-dimensional video (3DV) systems. However, stereoscopic image quality assessment methods are very scarce presently. In the paper, according to human visual sensitivities to different regions with different scales, an objective stereoscopic image quality assessment (OSIQA) model is proposed to predict visual quality of a stereoscopic image. The structural differences among scales and regions are firstly calculated to obtain the left-right image quality assessment (LR-IQA) index after multi-scale decomposition. Then, a concept of absolute disparity image is defined to describe absolute difference between the left and the right images, and the changes referring to the brightness, structure and pixel statistical information between the original and the distorted absolute disparity images are regarded to gain the stereoscopic perception image quality assessment (SP-IQA) index. Finally, according to the weights of LR-IQA and SP-IQA indices, LR-IQA and SP-IQA are combined into OSIQA. Experimental results show that the proposed method closely approximates subjective evaluation with the correlation coefficient of 0.938, the Spearman rank-order correlation coefficient of 0.946, and the root mean square error of 5.952 for all the five distortions, including Gaussian blurring, Gaussian white noise, JP2K, JPEG and H264 compressions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.