基于非局部总变差的图像分割活动轮廓模型*

在一般活动轮廓模型的连续全局极小化方法基础上，利用四种非局部总变差，给出了一种具有连续全局极小解的非局部活动轮廓模型。由于该模型的非局部特性，在分割过程中能有效地去除图像中的噪声，同时保留那些重复的精细结构。数值实验证明，该模型能将图像中的主体结构和精细结构很好地分割出来，而标准活动轮廓模型的分割结果中则丢掉了许多小的精细结构。

Image segmentation active contour model based on nonlocal total variation

Based on the continuous global minimizing method for the general active contour model, this paper proposed a nonlocal active contour model having continuous global minimizing solutions by using four nonlocal total variation. Due to the nonlocal property of the model proposed here, noise could be effectively removed in the course of segmentation, and the repeated fine structures could be preserved simultaneously. Numerical experiment shows this model can segment the main structures and the fine structures very well, however many small fine structures are lost in the results of the standard active contour model.