基于深度敏感特性的立体质量评价模型

Stereoscopic Quality Evaluation Model Based on Depth Sensitive Characteristics

提出了基于深度敏感特性的立体质量评价方法.引入人眼深度最小可觉察误差模型构建了深度敏感性函数,根据原始深度级确定敏感度因子,进而结合可觉察阈值和敏感度因子计算深度的相对层次性,最后统计得出立体感客观质量值.实验结果表明,该方法结果与主观质量值之间的相关系数高于0.9,均方误差低于7.6,两者存在较好的一致性,并且优于传统的利用深度图信息评价立体图像质量方法的结果.

The objective quality evaluation of stereoscopic video is more convenient compared to subjective evaluation, and can be treated as target constraint criteria of the stereoscopic video compression and processing algorithms. Frontier methods obtain the depth map of stereoscopic images by extracting binocular parallax information of related objects, and the pixel values are mapped into depth perception quality values using a linear process.But the sensitivity characteristic of the human eyes to different depths is not the same,and the ：relationship between pixel value and depth perception of the human eye is nonlinear.This causes that the evaluation val- ue of stereoscopic sense is inconsistent with the actual human perception.This paper proposes a stereoscopic quality evaluation method based on the depth-sensitive characteristics, which introduces the just nodicable differencies for depth model to builde depth sensi- tivity function,and determines the original depth level based on the sensitivity factor, and then combines the sensitivity threshold and the sensitivity factor to obtain relative depth hierarchy.Finally stereoscopic objective quality values are obtained statistically.Experi- mental results show that the correlation coefficient between the results of this method and subjective quality values is higher than 0.9 ,and the mean square error is less than 7.6.There is good agreement between them,and the results are better than the traditional quality evaluation methods of stereoscopic images by depth map information.