The Method Of Quality Assessment Of Stereoscopic Videos

The activity of content production, transmission and processing of stereoscopic videos require systematic, effective and robust methods to assessment the quality of stereoscopic videos. By the help of a standard we can describe the quality of stereoscopic video, or represent the injury of stereoscopic videos introduced by any processing techniques. The objective metric to judge the quality of stereoscopic videos is very important.Because we can use it to assess the performance of processing system or transmit channel in real-time or non-real-time.In addition, we can also use it to valuate a video codec or optimize and design a stereoscopic display system. It can be seen that an effective objective metric of stereoscopic videos will make a great promotion to the industry of stereoscopic videos.Because the stereoscopic videos are the extension of the traditional two-dimensional videos, any factors that affect the quality of two-dimensional videos will also work in stereoscopic case. In addition, the stereoscopic videos are different to the traditional case. We have to consider the depth perception and visual comfort and so on in addition to the quality of different views that consisted in the stereoscopic videos.Objective stereoscopic videos quality assessment is the essential evidence to evaluate the performance of stereoscopic video system. It is very important to develop an effective quality assessment metric for stereoscopic videos because two-dimensional metrics directly extended to the stereoscopic case do not work well. This paper proposed a new perceptual quality assessment metric for stereoscopic videos with multiview videos plus depth format which considering the spatial, temporal and depth structural information of stereoscopic videos. Specifically, our method extracts five types of feature, including luminance feature, color feature, texture feature, motion vector feature and depth feature,to express the spatial, temporal and depth structural properties of visual scene. By calculating the similarity of all the features contrast from both original and distorted stereoscopic videos, we design a full-reference stereoscopic video quality assessment model. Experiment results on stereoscopic videotest databases show that compared with the relevant existing metrics, the proposed algorithm can achieve better correlative with subjective scores. Our algorithm is consistent with the characteristics of human visual system, and can effectively predict human eyes’subjective perception of stereoscopic videos.