形态梯度重构的标记分水岭高光谱影像分割

Segmentation of Hyperspectral Remote Sensing Image Based on Morphological Gradient Reconstruction and Marker Extraction

传统分水岭算法通常对梯度图像做无标记分割,其结果是容易造成过度分割。为了克服过分割的缺陷,进而应用于复杂的高光谱遥感图像分割,结合形态学预处理方法,在对图像实施平滑处理的同时,利用形态学开闭重构技术对梯度图像进行重建,在此基础上对高光谱遥感梯度重建图像进行标记分水岭分割。实验证明,这种处理技术对高光谱遥感图像的分割效果良好,能够满足高光谱遥感图像分类与信息提取的需要。

Gradient images are usually implemented by traditional watershed algorithm with no-signal segmentation, which can easily cause over-segmentation. To overcome the defects and apply watershed algorithm to the complex hyperspec- tral remote sensing image segmentation, it combines with morphological preprocessing method in the text. An image is smooth processed, its gradient image is reconstructed by open and close reconstruction technology of morphology at the same time. Based on that typerspectral remote sensing gradient and reconstruction images are segmented by marked watershed algorthm. It is proved by the experiment that this processing technology has a good effect on the segmentation of the hyperspectral remote sensing images, and can satisfy the classification and information extraction needs of hyperspectral remote sensing images.