摘要
设计了一种用于图像配准的优化算法。根据图像灰度和空间结构信息,构造一个基于最大化互信息的配准测度函数,结合一种适用于图像自动配准的文化粒子群优化(CIPSO)算法。新的配准算法在搜索空间上进行优化设置,将整个搜索空间分成群体空间、信念空间两组。群体空间的粒子不仅根据自己空间的个体最优和全局最优来优化自己,还与信念空间中的最优个体进行交互学习,加速粒子群的收敛速度,克服了图像配准中计算量过大、搜索速度慢等问题。实验表明,与现有的PSO配准算法相比,提出的算法具有较好的鲁棒性和配准精确率。
This paper designed a kind of optimization algorithm for image registration. Depending on the image gray and spatial structure information,it constructed a registration measure function based on maximum mutual information. By combining with cultural improved particle swarm optimization （CIPSO）, it outlined a novel image registration algorithm. In the algorithm, it di- vided the search space into two spaces, group space and belief space. Particles of group space not only track individual extrema and global extrema to update themselves, but also exchange with the best individuals in belief space to speed up the conver- gence rate, which overcome the problems in image registration such as the large computation complexity, slow search speed and so on. Experimental results that compared with existing PSO registration algorithm, the proposed algorithm has good registration accuracy and good robustness.
出处
《计算机应用研究》
CSCD
北大核心
2015年第6期1874-1877,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(10972102)
国家博士点基金资助项目(200802870007)
江苏省科技支撑计划资助项目(BE2009163)
淮安市科技项目(HAG2012048)
江苏省高校优势学科建设工程资助项目(PAPD)
关键词
互信息
文化粒子群优化算法
测度函数
图像配准
mutual information
culture particle swarm optimization algorithm
measure function
image registration
