Abstract:

This paper proposes an improved watershed segmentation algorithm to improve the quality of medical image segmentation.Firstly,the medical image is pre-processed before using this algorithm,including the application of minimum threshold method to the region of interest,the separation of foreground background area,and the determination of the candidate regions by operations of dilation and erosion.Then in the process of watershed transform,the connected clusters of the same characteristics with the primary cluster are added to the segmentation result by the pixel clustering combination rule to give the merge area.Experiments prove that this improved watershed algorithm can effectively extract the lesion region from the medical image while diminishing segmentation.

Key words: watershed segmentation algorithm;medical image segmentation;clustering merger