二维直方图分水岭分割方法的研究

Research of watershed segmentation method based on two-dimensional histogram

提出了一种二维直方图分水岭分割方法。该方法选取遥感影像总体分辨能力较强的两个特征构建二维直方图,直接将二维直方图看作一幅灰度影像,然后用标记分水岭算法对二维直方图进行分割以确定不同类别的地物在特征空间中的位置,最后对遥感影像中的每个像素进行类别标记。与传统二维直方图分割方法相比,新方法可以筛选出区分能力较强的两个特征构建二维直方图,因而能取得更为理想的分割效果。

A watershed image segmentation method based on two-dimensional histogram is proposed. It selects two features of remote sensing image with stronger overall resolution capability to build a two-dimensional histogram which can be considered as a grayscale image directly. Then in order to locate diverse surface features in the eigenspace, the marked watershed algorithm is adopted to segment the grayscale image. Finally, label each pixel of the image according to its category. Compared with the traditional two-dimensional histogram segmentation method, the new method can filter two features which own better distinguishing ability to construct a two-dimensional histogram, and therefore shows better segmentation result.