基于Gabor滤波器和模糊支持向量机的纹理分割

Texture Segmentation Based on Gabor Filter and FSVM

文章提出基于Gabor多通道滤波器和模糊支持向量机的纹理分割.采用Gabor滤波器提取特征向量,送入FSVM进行分类,并与SVM、RBF神经网络的分割结果做比较.结果表明:FSVM和SVM比RBF神经网络具有较好的泛化性能,训练时间也大大减少.此外,FSVM比SVM分类错误率低,有更强的抵抗噪声能力.

In this paper we introduced the theory of FSVM briefly and application in texture segmentation, and discussed in detail the core techniques and algorithms which determine the fuzzy membership based on kernel methods, and comparing its classifying ability with RBF network and SVM. During simulation experiment, Gabor wavelet analysis technique is adopted to extract feature vectors of texture. The results show that FSVM had higher correct recognition rate and shorter training time. Furthermore, FSVM is proved to have stronger ability to resist noise.