摘要
纹理图像的分类是目前一个非常活跃的研究课题。针对现有纹理图像分类算法的局限性,本文提出了一种基于Contourlet变换和仿生模式识别方法的纹理图像识别算法。首先应用Contourlet变换获得能量特征的方法提取能量特征,进而利用仿生模式识别算法实现对纹理图像的识别。采用Vistex纹理库数据进行仿真实验,结果表明:与传统的分类方法相比,利用Contourlet变换和仿生模式识别结合进行纹理图像的识别能获得更高的正确率和速度,最佳正确率可达100%。
Texture image classification is a very hot research topic. A texture image classification method based on the Contourlet Transform （CF） and Biomimetic Pattern Recognition （BPR） is proposed while introducing some prevailing classification methods. In this method, the Contourlet Tranform that obtains the inner characteristics of images is applied to extract the energy feature, and then the texture images are recognized using the BPR algorithm. The Vistex image database is used in the simulation experiment. The experimental results show that the combination of CF and BPR improves the texture image recognition rate effectively, and the best recognition rate attains 100 %.
出处
《计算机工程与科学》
CSCD
北大核心
2010年第1期60-63,共4页
Computer Engineering & Science
基金
湖南省自然科学基金资助项目(06JJ50109)
