结合支持向量机与C均值聚类的图像分割

针对支持向量机进行图像分割时需要用户设定训练样本问题，提出一种根据图像特征使用C均值聚类算法自动获取支持向量机训练样本的方法。首先将图像分成几个区域，对每个区域利用小波分解去掉含有图像边缘的区域，然后对剩余的平滑区域计算能量均值作为特征值，使用C均值聚类算法对平滑区域分类，将特征值与类别标记作为支持向量机的训练样本，最后用训练后的分类器对图像进行分割。实验结果表明提出的方法取得了很好的分割结果，同时用一幅有代表性的图像进行支持向量机训练，所产生的分类器可以应用于所有该类图像，因此可以很容易应用到体数据的分割中。

Image segmentation combining support vector machines with C-means

Image segmentation based on support vector machines(SVM) requires the user to provide the training data.The proposed method in this paper used C-means to obtain feature vectors and labels for training SVM.Firstly,image was divided into several regions and discrete wavelet transform was performed on each region in order to remove edged region.Secondly,after applying C-means to smooth region classification,the energy of region and labels were taken as training data of SVM(Support Vector Machine).Finally,image segmentation was performed using SVM classifier.Experimental results show that the method has good performance in image segmentation.Meanwhile,using one representative image for the training of SVM,the produced classifier can be applied to the set of similar images and 3D volume data.