一种改进的局部区域特征医学图像分割方法

水平集分割方法中的Chan-Vese模型能够处理具有模糊边界和复杂拓扑结构的图像,但没有充分利用图像局部灰度的变化信息,致使其不能准确分割强度不均匀物体。针对这一问题对模型做了改进,引入局部灰度均值替换全局均值,以边界指示函数作权进行加权长度积分,加入使用双阱势的距离正则项来避免水平集重新初始化。试验结果表明:改进后的模型能够有效提高分割精度与效率,可以有效应用在医学图像的分割领域。

An Improved Medical Image Segmentation Model with Local Region Features

Although the conventional Chan-Vese model could process images with fuzzy boundary and complex topology,it could not segment objects with nonuniform grayscale accurately for the underutilization of the local diversity in an image.In this paper,a local weighted mean grayscale was brought in to replace the globle one. A boundary indicator was used to make the weighted length integral,and a distance regulation term with double-well potential was added to avoid the reinitialization,which makes some improvements to the CV model.Experiments show that the proposed model can effectively improve the segmentation precision and efficiency,and it is useful for medical image segmentation.