摘要
提出了一种基于归一化互信息相似性判据,并采用模糊自适应粒子群优化算法(particle swam optimization,PSO)作为搜索策略的图像自动配准方法。由于互信息方法不能解决图像缩放的问题,该方法在计算图像互信息之前,先对图像进行尺寸相同化操作;同时针对互信息方法中目标函数易陷入局部极值及搜索速度慢的问题,该方法采用归一化互信息作为相似性准则,并提出以模糊自适应PSO算法作为优化策略来提高配准速度和精度的方法。实验表明,采用归一化互信息作为配准测度,可提高配准的鲁棒性,而且,引入了模糊推理机之后,配准效率得到大幅提高,用该方法对具有仿射变换的图像进行配准能得到快速、精确的配准结果,证明了该算法的可行性和有效性。
A new automatic image registration method, using fuzzy adaptive Particle Swam Optimization （PSO） algorithm as the search strategy, based on normalized Mutual Information （MI） as the similarity criterion is presented. In MI method, to solve the problem of scale in image registration, the two images are produced to have same size before calculating the MI. Meanwhile, the target function gets into local extremes easily and the search speed is slow, in order to solve these problems, normalized MI as the similarity criterion and adopts fuzzy adaptive PSO as the optimization strategy to improve the speed and the precision are used. Experimental result shows that normalized MI can get high robustness. Moreover, the introducing of fuzzy reasoning machine improves the registration efficiency greatly. This method has fast and accurate registr.ation result for images that with affine transformation. The experiment proves the feasibility and validity of this algorithm.
出处
《计算机科学》
CSCD
北大核心
2008年第6期175-177,共3页
Computer Science
基金
国家自然科学基金(60702063)
国防预研项目
广西区青年科学基金(桂科青0640067)资助
关键词
图像配准
归一化互信息
粒子群优化算法
模糊自适应PSO
Image registration, Normalized mutual information, Particle swarm optimization, Fuzzy adaptive PSO
