An improved C-V model for breast MR images segmentation
Hong Fan, Chengcheng Zhang, Kezhen Zhang, Ruoxia Yao · 2018
This paper proposes a novel improved Chan-Vese (ICV) model to segment breast MR images which have large amount of information, intensity inhomogeneities and weak boundary. The method integrates the local information of image into the global information by introducing a local fitting term and a self-adaptive scale parameter in Chan-Vese (CV) model to deal with the effect of bias field on the global images and improve the segmentation results of MR images. A penalty term is added into the improved model to avoid re-initialization procedure and ensure the model stableness. The performance of our model is confirmed by testing its robustness, anti-noise and accuracy. Experimental results for both synthetic and real breast MR images are compared with that of the CV model and Region-Scalable Fitting (RSF) model, thus verify the effectiveness and efficiency of the our model further.