Abstract: Ultrasound image segmentation is very difficult problem because of low resolution, low contrast between tissues and organs, weak borders and serious speckle noise.This paper proposes a graph cut based interactive image segmentation method according to the character of ultrasound images.First, remove speckle noise using anisotropic diffusion method;Then construct the energy functional, increase curve length constraint in energy model, creates an association between source point, sink point and image pixels, and defines data items and regularization term of functional model; Finally gives the proof that energy functional can use graph cuts to minimize, derive solution of the energy functional model in detail, construct the corresponding s-t diagram based on energy model characteristics and give steps of graph cut method.Experimental results show that this method can accurately and quickly segment the target in the image and solve small areas and no smooth problems. The segmentation effect is better than that of traditional graph model.