Parameterized Morphological Watershed Segmentation
Yangping Wang · Journal of the China Railway Society · 2013
Morphological watershed transform is an efficient way to segment an image,which can produce closed object contours.However,it often results in over-segmentation when it is directly applied to original images.Usually,over-segmentation can be avoided to some degree by pre-filtering,whereas the position of the segmented region contours may be offset.For the purpose to segment an image without over-segmentation and region contours position bias,the method of parameterized morphological watershed segmentation was proposed.Firstly,the function relationship between the gradient level and the size of structure element was established by simulating the process of viscous fluid flooding the image relief.Secondly,the morphological closing operator with larger sized structure elements was used to modify the lower relief surface of the gradient image,whereas the smaller-sized to the higher relief.By this way,most local minimums caused by irregular details and noises were removed,and region contours positions were not changed or slightly changed due to less modification of high gradient levels.Finally,standard watershed transform was employed to implement segmentation on the basis of modified gradient images.Simulations show that this method can effectively avoid watershed over-segmentation and precisely preserve the region contours positions.