一种分形图像匹配方法

A Fractal Approch for Images Matching

通过对ＦＢＭ的Ｗｅｉｅｒｓｔｒａｓｓ－Ｍａｎｄｅｌｂｒｏｔ随机分形迈近函数的频诺分析，给出了多重分形符征的定义，并给出了判别图像的ＦＢＭ区域的方法。在ＦＢＭ区域内采用该特征进行匹配能克服复杂自然景物图像中灰度与边缘特征的不稳定性。实验表明该匹配方法能够获得比传统匹配方法，例如平均绝对差算法以及特征匹配法更高的匹配定位精度和匹配概率。

Owing to the unstable grey level and edge features of images in natural scenery,a matching method based on the grey level and edges of image cannot yield perfect results.Amultifractal-feature based on an FBM model and the random fractal approximation functionfor the FBM model by Weierstrass-Mandelbrot is presented.Image matching is carried outusing this feature for FBM regions. Experimental results show that the approach proposedgives higher matching probability and better positioning accuracy than those given by tradi-tional ones such as the MAD(mean absolute difference)algorithm and the feature matchingmethod.