基于Brushlet变换的SAR图像地物目标分类

Targets classification in SAR image based on Brushlet transform

Brushlet是一种新的图像方向信息分析工具,已被应用于纹理分割、分类以及去噪等领域.该文利用Brushlet变换为复函数这一特性,提取其能量及相位信息作为分类特征.通过对SAR图像中选出三类具有不同地貌特征的样本:道路平地、山脉和建筑物群的分类实验,仿真结果证明较之单一能量特征的分类方法,Brushlet能量及相位特征能有效地区分不同特征的地物目标.

Brushlet transform is a new tool for analyzing directional information in images,which has been widely used in texture segmentation,classification and denoising etc.The complex function characteristics of the Brushlet transform is utilized to extract,as the classification features,the power and phase information.Three kinds samples of SAR images with different physiognomy features,i.e.,roads,mountains and buildings,are chosen for classification experimentations,and the simulation results show that the classification method with features composed of the Brushlet power and phase can outperform the method with only the power feature.