摘要
剪切波变换(Shearlet)不仅继承了传统小波变换的多分辨率性、局域性、临界采样等特性,而且还具有多方向性和基函数各向异性的优点;同时其还可以描述纹理图像更多的方向信息和对图像具有更强稀疏表示的能力.本文提出一种基于剪切波变换和支持向量机(SVM)的纹理图像分割算法,首先对纹理图像进行Shearlet分解,获得不同尺度的方向子带系数;然后对各尺度不同方向子带系数的纹理特征进行提取;进一步,利用模糊C均值聚类算法(FCM)对纹理特征矩阵进行分类,获取训练样本;最后将训练样本输入支持向量机进行训练获得对特征图像的分割结果.实验结果验证了所提出算法的有效性.
The shearlet transform not only inherits the properties of the traditional wavelet transform like multiresolution,location and critical sampling etc, but also has merit of rnultidirection and basis functions that are anisotropic. The shearlet transform is optimally sparse in representing images and can describe more information of the directions for the texture images. In this paper, a texture image segmentation based on shearlet transform and support vector machine （SVM） is proposed. First, we obtain the coefficients at different directional subbands after the shearlet transform to the texture images, secondly, extract the texture feature of directional subbands at various scales and orientations, and then, classify the feature image by fuzzy c-means clustering and get the trained samples,finally, segment the whole feature image with the SVM which has been trained by the samples. The simulation results clear show that the proposed method is an effective algorithm.
出处
《吉林师范大学学报（自然科学版）》
2013年第2期1-5,共5页
Journal of Jilin Normal University:Natural Science Edition
基金
国家自然科学基金项目(41271422)
辽宁省自然基金项目(20102123)
计算机软件新技术国家重点实验室开放基金(KFKT2011B11)
南京邮电大学图像处理与图像通信江苏省重点实验室开放基金(LBEK2010003)
智能计算与信息处理教育部重点实验室(湘潭大学)开放课题(2011ICIP06)
