结合L^1拟合项的Chan-Vese模型  

An Improved Chan-Vese Model Integrated with L^1 Fitting Term

作　　者：唐利明[1] 方壮[1,2] 向长城[1] 黄大荣[3] 陈世强[1] 

机构地区：[1]湖北民族学院理学院,恩施445000 [2]武汉大学数学与统计学院,武汉430072 [3]重庆交通大学信息科学与工程学院,重庆400074

出　　处：《计算机辅助设计与图形学学报》2015年第9期1707-1715,共9页Journal of Computer-Aided Design & Computer Graphics

基　　金：湖北省自然科学基金(2015CFB262);国家科技支撑计划课题(2015BAK27B03);湖北民族学院博士启动基金(MY2015B001)

摘　　要：为了提高Chan-Vese(CV)模型对椒盐噪声的鲁棒性,提出一个结合L1拟合项的CV模型.首先采用L1拟合和L2拟合的线性组合构造一个新的拟合项,然后通过调整这2个拟合的权重以提升该模型对不同噪声图像分割的灵活性,最后利用交替迭代算法对模型进行求解.采用被不同噪声污染的人造图像和自然图像进行实验的结果表明,该模型对噪声图像可以取得较好的分割结果,并且对于椒盐噪声污染图像的分割,比CV模型、LBF模型和VFCMS模型更具优势.An improved CV model integrated withL1 fitting term is proposed in this paper to enhance the robustness of the model for salt and pepper noise. First, a new fitting term is defined as a combination of L1 fitting and L2 fitting. Then, by appropriately choosing the weights of fitting, our proposed model allows flexible segmentation under various noise conditions. Finally, an alternating iterative algorithm is employed to solve the model numerically. Experiments on some synthetic and real images contaminated by different kinds of noise demonstrate that the proposed model is effective and robust for noise image segmentation. Moreover, compared with CV model, LBF model and VFCMS model, our model can achieve superior seg-mentation results for image corrupted by salt and pepper noise.

关 键 词：图像分割 CHAN-VESE模型 椒盐噪声 高斯噪声 L^1范数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]