摘要
从具有全局最优解的几何活动轮廓方法出发,分别提出了两种基于齐次Besov空间与小波变换的图像分割算法,并给出了解的存在性证明.数值求解利用小波软阈值以及分裂Bregman方法,能够有效提高计算效率.由于小波变换具有多分辨特性,对于包含较多细节信息的图像,采用新算法能够得到更好的分割效果.数值实验表明采用新算法能够获得较好的分割效果,并具有较高的计算效率.
Two image segmentation algorithms were proposed based on homogeneous Besov space, inspired from geodesic active contors methods with global minimizers. Meanwhile the existence of the solution was proved. Those new algorithms can be resolved efficiently through wavelet soft threshold thanks to the conections between wavelet transform and Besov s- pace semi-norm. And split Bregman method was also adopted in one of algorithms. Dwe to the multiresolution character of wavelet, they are proper to segment images containing more fine information. Numerical simulations show that the proposed methods can improve computation efficiency and result in good segmentation.
出处
《计算数学》
CSCD
北大核心
2011年第1期103-112,共10页
Mathematica Numerica Sinica
基金
国家自然科学基金(10601068)
全国优秀博士论文资助基金(2005043).
