科研之友
微信
新浪微博
Facebook
分享链接
摘要
Nowadays, active contour model and level set method have been widely used in the image segmentation, but these methods also have drawbacks: the huge amount of calculation and sensitiveness to the initial contour. To solve these problems, the paper proposes a new segmentation method called two-stage segmentation method (TSSM) which has two steps: firstly, obtain a rough segmentation result which is close to the real target boundary through watershed algorithm. Secondly, split the image by the improved Chan-Vese model which uses the rough result obtained at the first step as its initial contour. To improve the Chan-Vese model, the paper also completes the following jobs: adding local characteristic of the image to constrain the process of the curve evolution and proposing new iteration terminal condition that can stop the curve evolution automatically. To demonstrate the effectiveness of TSSM, we compare this method with the traditional Chan-Vese algorithm through processing the synthetic image, rice image and SAR images. The result of the experiment proves that TSSM has better numerical accuracy and faster division speed.
