为进一步进行纹理特征分析，从纹理的方向性入手，给出了纹理方向的数学定义式，合理选择差异函数，构造了具有物理意义的纹理方向描述特征向量，数据处理方面，运用模糊贴近度的概念，结合改进后的属性均值聚类算法，对一类具有方向性的纹理图象进行分类与分割实验，取得了较好的结果，试验表明，该方法对纹理的方向性有很好的描述能力。

For the further analysis of texture characteristics, considering from the texture direction, a definition formula of texture direction is presented, followed by some useful explanation and discussion. To describe the texture images' direction, a discriminating function is selected properly and an eigenvector, with physical significance, is constructed successfully. About the data processing, the theory of fuzzy information, attribute means clustering and new distance measurement formula are employed, associated with an improved fuzzy clustering method, to directional texture images, satisfied results in the clustering &segmentation experiment had been obtained. Experimental outcomes show that the method has good capability to describe the directionality of texture image. Benefit from this method, stripe analyses, relatively movement analyses and other complexity image analyses regions could give an easy description to the directional objects.