Abstract: According to the problem that the C-V model can’t segment much grayscale image and noise image effectively, it introduces a Gaussian function for nuclear function of local binary fitting energy term and an edge stop function. Using local window function within the weighted mean instead of C-V model of the global average, and joining the distance function compensation term to avoid the level set function to reinitialize, at the same time putting the image edge information into the C-V model, so it can overcome the problem of traditional C-V model that can’t use the information of gradient image. Experiments show that the improved model and LBF model are significantly better than the C-V model in the vascular graph. The improved model is also less than the LBF model and C-V model on splitting time; no matter on splitting speed or splitting effect the improved model has better advantages over C-V model and LBF model.

Key words: Gaussian kernel function, edge stop function, distance function compensation term, C-V model, Local Binary Fitting（LBF） model

摘要： 针对C-V模型不能有效地分割多灰度级图像以及抗噪性不强的问题，分别引入了一个以高斯函数为核函数的局部二值拟合能量项和一个边缘停止函数。利用局部窗函数内的加权均值取代C-V模型的全局均值，并加入了距离函数补偿项，避免了水平集函数的重新初始化，同时将图像的边缘信息融入到C-V模型中，克服了传统的C-V模型无法利用图像梯度信息的不足。实验证明，改进的模型和LBF模型在血管图的分割的效果明显好于C-V模型，在分割时间上，改进的模型也少于LBF模型和C-V模型；对于灰度分布不均匀以及含有噪声的图像，无论是在分割的速度还是分割的效果上，改进的模型均明显优于C-V模型和LBF模型。

关键词: 高斯核函数, 边缘停止函数, 距离函数补偿项, C-V模型, 局部二值拟合（LBF）模型