稀疏正则化的多目标图像分割变分模型

Sparsity Regularized Variational Model for Multiphase Image Segmentation

基于隶属度函数的稀疏正则化,本文提出一个新的多目标图像分割变分模型和相应求解算法.该模型和算法有以下主要优点:首先,稀疏正则可以更好地保持分割区域的边界,克服了全变差正则导致分割边界模糊的缺点.其次,利用多尺度几何分析工具可以更好地保持图像的几何形状.最后,提出算法简单、易实现、运行速度快.一系列实验结果验证了提出方法的可行性与有效性.

Based on sparse regularization to the membership functions,this paper proposes a novel multiphase variational model and corresponding algorithm for image segmentation.The proposed model and algorithm has three main advantages.Firstly,the sparse regularizer performs better than total variation regularizer.It protects edges from oversmoothing which is a common drawback of the total variation regularizer.Secondly,the multi-scale geometric analysis tool well preserves geometric shape of the segmentation regions.Finally,the proposed algorithm is simple and has rapid running speed.A series of experimental results demonstrate the feasibility and effectiveness of the proposed method.