基于水平集的局部自适应图像分割方法  

Local adaptive image segmentation using level set method

作　　者：王君伟[1] 刘利雄[1] 

机构地区：[1]北京理工大学计算机科学技术学院,北京100081

出　　处：《中国科技论文》2017年第8期895-899,共5页China Sciencepaper

基　　金：国家自然科学基金资助项目(61370133)

摘　　要：针对传统水平集图像分割模型无法准确分割灰度不均匀及多目标图像的问题,提出了1种改进的基于水平集的局部自适应图像分割模型。该模型在CV模型(Chan和Vese提出的模型)和LAW(local adaptive weighting)模型水平集演化方程的基础上,重新定义了1个局部自适应权重函数来表示像素点所在邻域的偏差信息,并约束该偏差信息与图像的局部灰度不均匀信息之间的差异为最小,以得到精确分割结果。将模型应用于多相位水平集中,实现了对多目标图像的分割。实验结果表明,该模型对灰度不均匀图像及多目标图像分割更准确,且对初始轮廓的位置更鲁棒。Aiming at the problem that the traditional level set models can not accurately segment the images with intensity inhomo-geneity and multi-target, this paper presents a local adaptive image segmentation method by combining a new variable-weight co-efficient matrix into the LAW model and extending to a three-phrase level set formulation Based on the level set evolution func-tion of LAW model, we first introduce a local adaptive weighting function, which represents the bias information in the neighbor-hood of pixels. Then, by minimizing the difference between the measured image and estimated image in a local region, the im-proved method can obtain more accurate segmentation result. Finally, we extend this model to a three-phase level set formulation for brain MR image segmentation The experimental results on synthetic and real MR images with intensity inhomogeneity and multi-target show the effectiveness of our method.

关 键 词：图像分割 水平集 灰度不均匀 多目标 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]