An Image Segmentation Method Based on Improved Watershed Algorithm

Abstract

Traditional watershed segmentation is sensitive to noise and can leads to serious over-segmentation. In order to overcome the shortcomings of traditional watershed segmentation, this paper presented an improved watershed image segmentation method. Firstly, the morphological opening/closing reconstruction filter is applied to remove the image noise. Secondly, multi-scale structure elements are used to calculate morphological gradient. Furthermore, the morphological gradient is modified by viscous morphological operators which can remove the most irregular local minimums. After the standard watershed transform, the region merging method based on neighbor regions edge value is employed to improve the segmentation result. Experiments show that this method can not only effectively avoid the over-segmentation of watershed, but also preserve the positions of regional contours.