Watershed Segmentation Based on Morphological Scale-Space and Gradient Modification
基于形态学尺度空间和梯度修正的分水岭分割

A method for watershed image segmentation based on morphological scale-space and gradient modification is proposed to avoid over-segmentation and the drawbacks of some improved watershed segmentations. Firstly, morphological hybrid opening and closing by reconstruction scale-space is employed to smooth the original image, after smoothing, the essential region contours are preserved and unimportant details and noise which are often the causes of over-segmentation are removed, and the problem of the traditional morphological opening and closing scale-space, including the lost of partial essential region contours and not satisfying scale causality, are both avoided. Secondly, in order to eliminate over-segmentation and to keep the scale causality from the extreme to the segmented regions, gradient modification is used before the standard watershed transform, to remove the regional minimum in the gradient image caused by the regional maximum in the smoothed image. Simulations show that this method can efficiently not only avoid over-segmentation, but also satisfy scale causality, and the localization of region contours is precise.