基于灰度秩数向量的非监控纹理图像分割

纹理图像分割的首要问题是纹理特征的提取与描述.传统的统计纹理描述方法和结构纹理描述方法各有特点,但也各有不足.本文从结构与统计结合的角度,提出了一种基于灰度秩数的结构性统计方法,该方法可以较好地抽取物体表面的结构特征.在用本算法对纹理进行描述的基础上,本文采用了一种广义最小体积椭球聚类算法(GMVE)对所抽取的纹理特征进行聚类分割.算法有效性在实际纹理分割中得到了验证.

Non-supervised Textural Images Segmentation Based on Gray Level Rank Vector

In this paper,a new textural images segmentation method is proposed which is based on textural statistical concept.This method can well abstract the textural characteristics.By using this method to describe texture,a non supervised generalized minimum volume ellipsoid (GMVE) clustering algorithm is adopted to realize non supervised textural images segmentation.The effectivity and efficiency of the proposed algorithm are illustrated in the experiments for real textural images segmentation.