摘要
在分形维数的基础上研究了将其用于纹理分割的方法。采用差分盒维数(DBC)方法和一种改进的边缘保持算法计算象素点的分形维数FD,基于原始图象的方向性差分和多重分形的概念提取出一组特征,并将Kohonen的SOFM网用于对得到的图象特征矢量进行分类,得到了较好的纹理图象分割效果。
Deals with the problem of recognizing and segmenting textures in images based on the fractal dimension (FD). Differential box counting (DBC) method and a modified edge preserving approach are used to estimate the FD. Six features are based on the original image, the horizontally differential image, the vertically differential image, two diagonally differential image, and the multi fractal dimension of order two. Kohonen′s SOFM net is adopted to cluster the feature vectors, and the segmentation results show the efficiency of the technique. The result is compared with feature smoothing and K means approach.
出处
《数据采集与处理》
CSCD
1996年第3期163-164,共2页
Journal of Data Acquisition and Processing
基金
国家自然科学基金
关键词
分形维数
纹理图象分割
神经网络
图象处理
feature extraction
neural networks
texture
fractal dimension
edge preserving