基于多尺度小波的城市遥感图像纹理特征提取

遥感技术为人们提供了大量的图像数据，如何从中提取所需信息，是遥感图像处理技术研究的热点。基于多尺度小波变换，可提取城市图像的纹理特征。根据城市影像中地物特征和小波技术的特点，确定合适的窗口大小，采用逐一滑动窗口的方法，对每一窗口提取相应的小波统计特征；应用小波变换的多尺度特性，在不同尺度上提取出子图像特征，构造出纹理特征矢量，为遥感图像的分类提供纹理特征。

City Texture Feature Extraction of Remote Sensing Image Based on Multi-resolution Wavelet Transformation

Texture analysis plays an important role in remote sensing, especially in city remote sensing. This paper describes a method of the feature-vector extracting of texture. Considering the character of city remote sensing and the wavelet transform, an appropriate size of sub-image window is introduced. Moving the window pixel by pixel, multi-resolution wavelet transform is processed. Then, feature vectors are constructed by computing entropy of each scale image of wavelet transformation. Further study, such as classification on feature space, is required to validate the feature vector of city image texture.