摘要： 提出一种卫星遥感图像中基于小波变换的形状描述新算法。对图像进行一级小波变换，用Hu矩、Zernike矩和Fourier描述子分别提取子波段特征向量，并根据各子波段的描述能力强弱对所得特征向量进行加权处理，并输入到分类器中检验其描述能力。实验结果表明，该算法能够对目标形状进行更准确的描述，较经典描述子性能更优。

关键词: 卫星遥感图像, 小波变换, Hu矩, Zernike矩, Fourier描述子

Abstract: A novel shape description algorithm in satellite remote sensing image is proposed based on the wavelet transformation. Images are decomposed by one-scale wavelet transform, the feature vectors are extracted by Hu moment, Zernike moment and the Fourier descriptor respectively, and the feature vectors are weighted according to each sub-wave band’s description ability. The feature vectors are put into the classifier to examine their description abilities. Experimental results demonstrate that the new algorithm has superior performance for more accurate description of the target shape. Compared with classical descriptors, its performance is much higher.

Key words: satellite remote sensing image, wavelet transform, Hu moment, Zernike moment, Fourier descriptor

中图分类号: 

• TP391