Modified Watershed Algorithm for Medical Image Segmentation

Two methods for modifying watershed algorithm are proposed to resolve the over-segmentation problem of watershed algorithm in region-based segmentation process. Firstly, the morphological filtering is carried on the high-hat and low-hat shape of the original images to filter out the noise and the non-perceptional information in order to preserve the structural information of the original images. Secondly, the gradient-based watershed segmentation is adopted by using mended label symbol. The entire segmentation process does not need region merging so that the complexity of segmentation is reduced. The improvement can not only suppress the over-segmentation phenomena properly, but also segment the pathological change regions in medical images effectively. Moreover, the methods can be used in arbitrary shape ROI and is compatible with the international standard of JPEG2000.