Improved segmentation of remote sensing images based on watershed algorithm

This paper suggests an improved segmentation of remote sensing images based on watershed algorithm to reduce the over-segmentation of the watershed transform. Taking into the characteristics of remote sensing, remote sensing images were filtered using a fast peer group filtering (PGF). Then do watershed transform on the multi-scale morphological gradient image and merge regions by grey similarity. Experiments show that the proposed method was a suitable one for remote sensing images. The image was effectively segmented, and the over-segmentation was reduced as far as possible.