Abstract: A new textural images segmentation method was proposed based on textural statistical concept. By using this method to describe texture, an improved c-means fuzzy clustering algorithm was adopted to realize non-supervised textural images segmentation.

Key words: texture primitive, texture segmentation, fuzzy C-mean clustering, K-L transform

摘要： 提出了一种采用灰度向量描述纹理基元的结构性统计方法,该方法可以较好地提取物体表面的结构特征。在用该方法对纹理进行描述的基础上,采用了改进的模糊C均值聚类算法对提取的纹理特征进行分割。将此方法应用到Brodatz标准纹理分类实验中,得到很好的效果。

关键词: 纹理元, 纹理分割, 模糊C均值聚类, KL变换