| 一种基于高阶统计量的纹理图像识别新方法 |
| |
| 引用本文： | 毕晓君,静广宇. 一种基于高阶统计量的纹理图像识别新方法[J]. 哈尔滨工程大学学报, 2004, 25(3): 363-366 |
| |
| 作者姓名： | 毕晓君 静广宇 |
| |
| 作者单位： | 哈尔滨工程大学,信息与通信工程学院,黑龙江,哈尔滨,150001;哈尔滨工程大学,信息与通信工程学院,黑龙江,哈尔滨,150001 |
| |
| 基金项目： | 哈尔滨市自然科学研究基金资助项目(2002AFXXJ033). |
| |
| 摘 要： | 提出一种基于高阶统计量的纹理图像识别方法,可以识别多种纹理图像.在二阶统计特征基础上,引入高阶统计量作为纹理图像的特征参数,并与人工神经网络相结合,建立起基于15个纹理特征参数的自组织神经网络,进行自学习 训练,从而实现图像识别;对于未经训练的纹理图像,网络可自动进行学习,并存储网络权值.实验结果表明,该识别方法能有效提高纹理图像的正确识别率.
|
| 关 键 词： | 高阶统计量 纹理特征 自组织特征映射网络 |
| 文章编号： | 1006-7043(2004)03-0363-04 |
| 修稿时间： | 2003-12-31 |
A texture image recognizing method based on high order statistics |
| |
| BI Xiao-jun,JING Guang-yu. A texture image recognizing method based on high order statistics[J]. Journal of Harbin Engineering University, 2004, 25(3): 363-366 |
| |
| Authors: | BI Xiao-jun JING Guang-yu |
| |
| Abstract: | A method based on high order statistics was proposed for recognizing many kinds of texture images. At first, on the basis of the traditional two-order statistics, high order statistics was imported as the feature function of the images. Then, according to the neural network theory, a self-organization neural network with 15 inputs was built,which can recognize a large of images. Especially, to strange texture, it can study and train by itself. The experiment results show that the recognizing method in this paper can improve the recognizing rate of texture image effectively. |
| |
| Keywords: | high order statistics texture feature self-organization neural network |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
