REMOTE SENSINGIMAGE TEXTURE CLASSIFICATION BASED ON BEST-RESOLUTION WAVELET FEATURES
具有更佳分辨率小波分解的遥感影像纹理分类

The wavelet transform is an applied mathematical theory which rose in the middle 1980s. And it has been applied widely in many fields such as image processing. The authors have been studying the features of cross wavelet transform of images. And some good results have been obtained in tex- ture classification of geomorphologic image . In this paper , a kind of wavelet features with best-resolution was put out. We know , the cross wavelet transform is to resolve the image at varied levels of a framework , that is different resolu- tions of the image. Each sub-section of the framework has a unique feature of frequency and spatial orientation. However , most of the important information of the texture image is located in the range of medium frequency. So if the cross wavelet transform is resolved in this range , more texture infor- mation will be obtained. This is a kind of wavelet features with best-resolution. In the view of the above , the classification experiment was carried out for 25 geomorphologic images under the condi-tions of different features , different filters and different resolutions , and achieved highly accurate classificatory results.