基于差别特征的纹理图像识别研究  

Texture Recognition with Discriminating Features

作　　者：石美红[1] 张军英[2] 李永刚[1] 

机构地区：[1]西安工程科技学院计算机学院,陕西西安710048 [2]西安电子科技大学计算机学院,陕西西安710071

出　　处：《计算机应用》2004年第1期66-69,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目 (60 0 71 0 2 6) ;陕西省教育厅专项科研计划资助项目 (0 2JK1 70 )

摘　　要：文中提出了一种基于PCNN的捕获率图的纹理图像差别特征提取的新方法。通过对PCNN的基本特性分析 ,指出PCNN的捕获率图不仅包含了全局图像灰度分布的统计特征 ,还包含了局部图像灰度的空间分布的相关统计特征 ,即相邻象素之间的几何信息。这恰是纹理图像的个性特征所在。实验结果证明 ,该方法具有良好的鲁棒性。获得这些纹理图像的特征旨在度量纹理图像的相似性或从图像中抽取纹理区域。A new method of discriminating features extraction of texture based on capture rate graph of PCNN is presented in this paper. By analyzing the basic characteristics of PCNN,it is indicated that the capture rate graph contains not only the statistical features of gray levels distribution in whole image but also the correlative statistical features of gray levels distribution in local image,which is the geometry information among adjacent pixels. It is just the individual feature of texture images. This method is well proved to be robust by our experiments. The aim of acquiring the features of texture images is to measure the comparability of texture images or to extract texture region from images.

关 键 词：脉冲耦合神经元 神经元点火 动态阈值 捕获率 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]