基于边缘及特征点匹配的立体图像质量评价

Quality evaluation of stereo image based on edge and characteristic point matching

为了实现对立体图像质量的精确预测,提出了一种基于边缘和特征点匹配的立体图像客观质量评价方法。首先,对图像的质量进行评价,在基于结构相似度的质量评价方法的基础上,考虑了边缘信息对人眼视觉特性的重要性,加入了边缘结构相似度指标;然后,对图像的立体感进行评价,采用特征点匹配的方法提取立体感指标。最后,根据总体视差失真法将图像质量和立体感指标拟合为一个综合指标。实验结果表明,采用本文提出的方法对立体图像测试库进行评价,总体评价的PLCC(Pearson Linear Correlation Coefficient)均在0.94以上;与其他方法相比,本方法具有较高的预测精确性。

In order to predict the quality of stereoscopic images accurately,an objective stereoscopic image quality assessment method based on edge and characteristic point matching is proposed.Firstly,for the quality assessment,a method that takes the edge information into account is presented,which is based on the plane image quality assessment method;secondly,for the stereo sense assessment,a stereo matching method is used for extracting stereo sense index.Finally,according to quality and stereo sense,a total evaluation index is proposed based on an appropriate weighted method.The experimental results indicate that compared with other methods,the overall Pearson linear correlation coefficient indicators are all above 0.94,which indicates that the proposed method can achieve higher prediction accuracy.