基于核函数与局部信息的凸优化分割模型

发布时间：2018-06-20 12:16

本文选题：核函数 + 图像分割　；参考：《计算机应用研究》2017年05期

【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。
[Abstract]:In view of the fact that the C-V model can not accurately segment non-homogeneous and high-noise images, and the computational efficiency is low, the following improvements are made: for each point in the region, The local energy term is defined by using the kernel function metric of the average gray value of the region in which the point is located and the gray value of the other points in the neighborhood. Then the global energy term is defined by integrating the local energy of all points in the image domain. Because of the introduction of local information and kernel function, the region mean updating has strong anti-noise ability and improves the segmentation robustness, and then the model is transformed into a global convex segmentation model. At the same time, the total variation norm weighted by the boundary edge detection function is introduced to obtain the boundary position of the target more accurately, so as to improve the segmentation accuracy of the model. Finally, the split Bregman iteration is used to solve the problem. The experimental results show that the proposed model can effectively segment heterogeneous and high-noise images. Compared with C-V RSF and DRLSE models, the speed and segmentation accuracy of the proposed model are greatly improved.
【作者单位】：太原理工大学计算机科学与技术学院;
【基金】：国家自然科学基金资助项目(61472270,61402318)
【分类号】：TP391.41

【相似文献】