通过分析人类视觉系统的纹理方向特性和立体感知特性,并结合数字水印的半脆弱性和支持向量回归(SupportVectorRegression,SVR)的泛化学习能力,该文提出一种基于视觉感知和零水印的部分参考立体图像质量客观评价模型。该模型利用立体图像左右视点经小波分解后在同一空间频率的水平和垂直方向子带系数关系构造反映图像纹理方向特征的视点零水印,同时,利用左右视点视差值与自适应阈值的大小关系构造反映立体感质量的视差零水印,然后利用SVR来学习两类零水印恢复率(视觉加权视点零水印恢复率和视差零水印恢复率)与主观评价值的关系,最后用训练好的SVR完成立体图像质量预测。实验结果表明该模型符合人眼视觉特性,所得到的客观评价值与主观评价值具有较好的一致性。
Through analyzing the image structure direction characteristics and the stereoscopic perception of human visual system, and combining semi-fragile digital watermarking and Support Vector Regression (SVR), a reduced-reference stereoscopic image quality assessment model based on visual perception and zero-watermark is proposed. In this model, the view zero-watermark is constructed by judging the relation of the horizontal and vertical wavelet coefficients, which can reflect the image structure information. Meanwhile, the disparity zero-watermark that reflects the stereoscopic perception quality is constructed by using the disparity between the left and right views. And then the relativity of two watermark-recovering rates (watermark-recovering rates of the view and disparity zero-watermarks) and subjective quality scores can be learned by the training procedure of SVR. Finally, stereoscopic image quality is predicted by trained SVR. Experimental results show that the proposed reduced-reference model is in accordance with human visual characteristics, and consistent with the result of subjective evaluation value preferably.