基于IGM和深度感知的立体图像质量评价

Reduced-reference quality assessment of stereoscopic images based on IGM and depth perception

根据人类视觉系统的特点,提出了一种基于内在推理机制(internal generative mechanism,IGM)和深度感知的半参考立体图像质量评价(stereoscopic image quality assessment,SIQA)方法,用图像质量和深度感知质量2个因素来评估立体图像的体验质量(quality of experience,QoE).首先,对于图像质量,根据大脑的内在推理机制将左右视点分别分解成可预测部分和不确定部分,用基于灰度共生矩阵(gray level co-occurrence matrices,GLCM)和基于视觉信息量的质量评价方法计算这2个部分的质量;然后,对于深度感知质量,采用一种改进的自然场景统计(natural scene statistics,NSS)模型来预测这部分质量;最后,将图像质量和深度感知质量融合为立体图像体验质量.实验结果表明,该算法在常用视频库上的结果优于现有的评价方法,且与主观感知具有较高一致性.

In this paper, a reduced reference stereoscopic image quality assessment (SIQA) method is proposed based on the internal generative mechanism (IGM) and depth perception. The method determines 3D quality of experience (QoE) by focusing on image quality and depth perception quality. For image quality, a stereoscopic image is decomposed into a predicted portion and an uncertain portion according to IGM of the brain, handled with the metric based on gray level co-occurrence matrices (GLCM) and visual information separately. The depth perceptual quality is measured with an improved natural scene statistics (NSS) model. The image quality and depth perception quality are then integrated to obtain the overall quality. Experimental results show that the proposed metric outperforms the state-of-the-art metrics and is consistent to subjective ratings over widely used databases.