Medical Ultrasound Image Segmentation Based on Improved Watershed Scheme

With the characteristics of precision edge and closed contour, watershed segmentation algorithm is used widely in image segmentation. However, over-segmentation always occurs when it is applied to ultrasound images associated with speckle noise. In this paper, an improved watershed segmentation scheme is proposed based on comprehensive consideration of gray scale information, edge information and the relationship between neighboring regions of ultrasound images. A novel criteria deciding the relationship between neighboring regions was defined, and edge information was also added to the merging guidelines. Similar region merging was then applied to the initial segmentation results. The proposed scheme was tested using representative ultrasound image. The experimental results show that the proposed scheme can produce accurate contours and overcome the over-segmentation phenomenon availably.