Gradient vector flow and watershed transformation combined segmentation algorithm

Although watershed transformation is used extensively in image processing applications. Problems associated in sensitive to the noises and over-segmentation, however, have limited the utility. A new segmentation method combined gradient vector flow with watershed transformation which largely solves both problems is presented in this paper. This method diffuses the edge map and removes the noises using the one dimension gradient vector flow, and obtains a gradient image which is suitable for the watershed algorithm. And then the extended minima transformation will be chosen to preprocess the gradient image to further reduce the local minima in the image. Finally, the transformed gradient image is segmented by the watershed algorithm and the segmentation is merged by region merging. Experimental results show that watershed segmentation with the proposed algorithm effectively inhibits the over-segmentation phenomenon and raises the segmentation accuracy.