在研究人类立体视觉特性及现有立体图像/视频质量评价算法的基础上,提出了一种基于运动信息和视差信息的立体视频质量的客观评价方法。方法包括视频质量评价(VQA)和视频立体感评价(VSSA)两个指标,其中VQA的估计基于梯度的结构相似度(GsSIM)算法,并充分考虑了帧内的亮度信息和结构信息、帧间运动信息以及人眼的感知特性对视频质量的影响,特别是根据人类的视觉特性,对左右视点的质量赋予了不同的权重;VSSA的估计是通过计算参考视频的绝对差值图和降质视频的绝对差值图之间的峰值信噪比(PSNR)而得到。实验结果表明,本文方法对基于H.264编码的失真视频的评价结果与主观测试有较高的一致性,很好地体现人眼的视觉特性。
Through analyzing the characteristics of the human visual system and the existing algorithms of stereo image or video quality assessment, an objective quality evaluation method of stereo video based on motion and disparity information is proposed in tiffs paper. The stereo video is assessed from the two as- pects of video quality and stereo sense. We follow the method of gradient-based structural similarity (GSSIM) and give full consideration to the effects of the intra-frames luminance and structure informa- tion, the inter-frames motion information as well as the characteristics of the human visual system on the quality of stereo video, when dealing with objective quality assessment of the stereo video. Especially, we give different weights to the quality of left and right view,according to the characteristics of the human visual system. In addition,we compute the peak signal-to-noise ratio (PSNR) between the absolute dis- parity maps of original video sequence and distorted video sequence when assessing the stereo sense of the stereo video. Experiment results demonstrate that the objective assessment method presented in this paper gets similar results with the general subjective assessment method and is related to the human vis- ual system when assessing the encoded stereo video based on H. 264.