一种新的基于Gabor小波的非监督纹理分割方法

基金项目: 国家863计划资助项目(2003AA135010)

详细信息

A New Unsupervised Texture Segmentation Method Using Gabor Wavelet

¹Institute of Remote Sensing Applications,CAS,3 Datun Road,Beijing 100101,China; The Second Artillery Petty Officer School,Tuoshan South Road,Qingzhou 262500,China

Funds: 国家863计划资助项目(2003AA135010)

摘要: 提出了一种新的基于Gabor小波的非监督纹理分割方法,与传统Gabor方法相比,该方法最大的特点在于分割的过程中利用了纹理尺度之间的依存关系和像素之间的空间约束关系。为了克服增加特征所带来的“维数灾难”问题,用独立分量分析(ICA)进行特征的整合。采用Brodatz测试集的实验结果验证了方法的有效性。
- Gabor小波 /
- 非监督纹理分割 /
- 尺度变化 /
- 空间约束 /
- 独立分量分析
Abstract: A new unsupervised texture segmentation method is proposed by using Gabor wavelet.Compared to the classic one,we add Gabor scale relationships and space constraints features to the multi-scale features.In order to find prominent features for segmentation,texture features are integrated by independent component analysis(ICA).The effectiveness of the proposed method is demonstrated by using a set of images from the Brodatz texture album.
- Gabor wavelet /
- unsupervised texture segmentation /
- scale relationship /
- space constrains /
- independent component analysis