一种自由视点视频系统中立体图像质量的评价方法

A quality assessment metric for stereoscopic image in free viewpoint video system

提出了一种自由视点视频系统中立体图像的质量评价方法。分别对参考图像与失真图像进行小波分解后用对比敏感度函数加权,计算两者之间的平均结构相似度得到基于内容的质量分数;在双目匹配区域中找到中心凹区域,在该区域上计算图像之间的平均相位相似度,得到基于深度感知的质量分数;利用阈值提取匹配区域中为人眼所感知的失真区域,与利用视差映射过来的另一视点图像的相对应区域计算平均结构相似度得到基于双目特性的质量分数。融合各部分质量分数得到最终图像质量。实验结果表明,本算法评价结果与主观感知一致性较好,更符合人眼视觉特性。

A quality assessment metric of stereoscopic image in free viewpoint video system was proposed. After the reference and distorted images being decomposed by wavelet then weighted by contrast sensitivity function respec- tively, the content score was obtained by calculating the average structural similarity. After the center region being found in the binocular matching region, the average phase similarity was calculated on the region to obtain depth perception. The distortion region perceived by the human eye in the matching region by using the threshold, the av- erage structural similarity was calculated between this region and the corresponding region of the other view image by disparity mapping, then the binocular characteristic score was obtained. The final image quality can be obtained by fusing score of each part. The experimental results show that the proposed algorithm has better consistency with sub- jective perception and it is more in line with human visual characteristics.