几种图像相似性度量的匹配性能比较

Comparison of matching capabilities in similarity measurements

针对景象匹配中匹配性能和匹配实时性会受相似性度量选择影响的问题,从常用的相似性度量(归一化积相关、相位相关、均平方差和去均值均平方差)入手,对图像施加噪声和进行畸变(如图像旋转变化、图像比例变化、光照强度变化和云层遮挡等),通过相似性度量匹配性能的仿真试验,从匹配性、适应度和实时性等方面对各相似性度量进行比较,并对结果进行了归纳总结与证明。

Image matching performance and real-time capability mostly depend on how to choose similarity measurement method of image matching algorithm. In order to analyze this topic in-depth, the paper carried out a series of simulation experiments for the matching performance evaluation of four similarity measurements （ NProd, PC, MSD and equalization MSD） through noise addition and image distortions. The image distortions include picture orientation, image scale change, illumination intensity change, clouds cover, and so on. The experimental results were compared with those from the matching performance, adaptive capacity and real-time capability. Analysis and theoretical demonstration of the results were also given.