A Study On Subjective And Objective Assessment Of Stereoscopic Video Quality Based On H.264

With the rapid development of media communication, stereo video technology has become the focus research. However, due to the limits of transmission resources and insufficient bandwidth, it often needs certain video compression algorithms to solve these problems, which could cause varying degrees of mass damage to the reconstructed videos. Therefore, it is particularly important to develop an efficient and comprehensive stereo video quality assessment system for evaluating the degradation degree of stereoscopic video, and at the same time, the improved quality evaluation system can provide important guarantee for conducting stereoscopic video coding and transmission services reasonably.Stereo video quality is explored from both subjective and objective aspects in this paper. In the subjective evaluation, accurate experimental data are the prerequisite and basis for exploring the impact of compression transmission on quality degradation damage of stereoscopic video, so evaluation data are obtained by a large number of subjective experiments, a rational and effective subjective evaluation data processing method is designed and the processed results are analyzed. This treatment includes elimination of abnormalities, removal of outliers and average processing, and the experimental results show that the data processing ability of the method proposed in this paper is better than that of traditional Grubbs test. By the processing method adopted in the above experimental data, the experimental data of stereo video processed by H.264 compression algorithm are analyzed, from which the conclusions of four areas are drawn: the size of the quantization parameter has influence on the damage degree of a stereo video quality; the information carried by reference viewpoint takes more dominance; high-resolution video contributes to degradation compensation; the ability of women to accept visual impairment is higher than that of men. The conclusions drawn from this paper provide technical support for the optimization of the stereo video communication technology.As for objective evaluation, this paper presents a new full-reference stereo video quality objective assessment method, which is to study mainly from two aspects, namely, quality and stereo sensation. For quality, full consideration is taken to the temporal characteristics of stereo video. Firstly, three-dimensional Sobel operator is used to determine visual points and partial three-dimensional structure tensor of the obtain points is subject to characteristic decomposition, according to the eigenvalues and eigenvectors, configuration descriptor matching the pixel is constructed around viewpoint, and finally the scores Q_q obtained by calculating the similarity of descriptors between reference video and degraded video, in stereo aspect, for each frame of disparity map of resulted reference video and degraded video, the algorithm GSSIM is applied to calculate the structure information, and ultimately all the frame similarity comparison results are weighed to obtain the average score Q_d,combined with subjective evaluation results, BP neural network is adopted for training the weigh ratio of these scores above Q_q and Q_d to get the final objective evaluation score Q.In order to verify the correctness of the algorithm, the correlation of both subjective and objective evaluation results, and the mean square root error are consideration as two aspects for measurement in this paper. Experimental results show that the objective algorithm proposed is highly related to subjective evaluation and the performance is superior, consistent with human visual characteristics.