关键词: 图像分割, 二维直方图, 标记分水岭, 形态学

Abstract: 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.

Key words: image segmentation, two-dimensional histogram, marked watershed, morphology