基于DT-CWT和SVM的纹理分类算法  

Texture classification algorithm based on DT-CWT and SVM

作　　者：练秋生[1] 尚燕[1] 陈书贞[1] 王林[1] 

机构地区：[1]燕山大学电子与通信工程系,河北秦皇岛066004

出　　处：《光电工程》2007年第4期109-113,共5页Opto-Electronic Engineering

基　　金：河北省教育厅自然科学项目资助(2004124)

摘　　要：提出了一种基于双树复数小波变换(DT-CWT)和支持向量机(SVM)的纹理分类算法。双树复数小波变换不仅具有实数小波的诸多优点,而且还具有近似平移不变性、良好的方向选择性和低冗余度,并且能对图像进行完全重构,能够更好地刻画纹理的特性;支持向量机算法是近年发展起来的性能优越的分类算法,比传统分类器有很大的优越性:避免了局部最优解和“维数灾”问题,其最优分类超平面的思想能够提高分类准确度。该方法用双树复数小波对纹理图像进行滤波并在各方向子带上进行重构,再计算其局部能量函数得到每个像素的特征向量,最后利用支持向量机算法实现对纹理图像像素的分类。将本方法与其它的分类算法进行比较,实验结果表明,提出的算法能有效地提高正确分类率。A texture classification algorithm based on Dual-tree Complex Wavelet Transform （DT-CWT） and Support Vector Machines （SVM） is proposed. The DT-CWT has the same advantages as the discrete wavelet transform （DWT） and some important additional properties： approximately shift invariance, better directional selection, lower redundancy and perfect reconstruction, so it can characterize the textures more precisely. SVM developed recently can achieve better classification performance, which offers several typical advantages that are not found in traditional classifiers. For example, it avoids the problems of local extremums and ‘dimension disaster＇, and the principle of optimal separating hyperplane can increase the correct classification rate. The texture image is filtered by DT-CWT and then reconstructed the image in each direction. The feature vector of each pixel is obtained by calculating the local energy function to the reconstructed image. At last we classified the pixels of texture image using SVM algorithm. Compared with other classification algorithms, the experiment results show that our method can improve the correct classification rate effectively.

关 键 词：双树复数小波变换 支持向量机 特征提取 纹理分类 

分 类 号：TN911.7[电子电信—通信与信息系统]