基于核函数与局部信息的凸优化分割模型

摘要

针对C-V模型不能准确分割非同质和高噪声的图像,且计算效率比较低的特点,作出如下改进:对于区域中的每一点,利用该点所在区域的平均灰度值和其邻域内其他点的灰度值的核函数度量定义局部能量项,然后对图像域上所有点的局部能量进行积分定义全局能量项,由于局部信息和核函数的引入使得区域均值的更新具有较强的抗噪能力,提高分割鲁棒性;然后将该模型转换为全局凸分割模型,同时引入边界边缘检测函数加权的总变差范数(total variation,TV)更加准确地获取目标的边界位置,以提高模型的分割精度;最后,使用split Bregman迭代进行数值求解.实验结果表明,该模型能够有效地分割非同质和高噪声图像,与C-V、RSF和DRLSE模型相比,在运行速度和分割精度上有了很大的提升.%This paper was intended to put forward some improvements against the defects of the C-V model that couldn't accurately segment the images with intensity inhomogeneity and the images with high noise and its computational efficiency was not very high.So this paper made improvements against these defects.First,for each point in a region,it defined a local energy according to the kernel function metric between the intensities of all points within its neighborhood and the intensity average of the region.Then for the whole image domain,it defined a global energy as a data term to integrate the local energy with respect to the neighborhood center.The kernel metric and the local intensity information that were incorporated into the energy made the updating of region mean values more robust against the noise and improved the robustness of segmentation.The convex optimization was applied to this new model,which used a weighted TV-norm given by the edge indicator function to detect the boundaries more accurately.Finally,it used the split Bregman iterative method for numerical solution.Experimental results show that the proposed model can obtain better results with respect to images with noise and inhomogeneous,compared with C-V model,RSF model and DRLSE model.This method greatly increases the computation efficiency and accuracy.