基于分形维数的纹理图像分割

纹理图发割过程一般分为特征抽取和特征划分，文中提出一种新的基于分形维数的纹理图像分割方法，在特征抽取上，以分形作为纹理特征，运用图像变换的思想，结合差分盒计数和基于分形布朗自相似模型的分形估计方法。

TEXTURE IMAGE SEGMENTATION USING FRACTAL DIMENSION

There are two important technique in image segmentation, feature extraction and feature segmentation. In this paper, fractal dimension is used as a feature for image segmentation. A new method combining the image transform with two kinds of fractal dimension estimator is investigated. In this method, a feature vector is assembled, and an edge preserving filter is used for getting rid of the estimating error, feature space is firstly classified based on K means clustering, the final segmentation is accomplished by using probabilistic relaxation techniques. The testing experiment is made out on synthesized texture mosaics and real texture images, the result is better than other algorithms.