摘要
目前的多聚焦图像融合方法对于融合模型的建立主要依赖于经验,其参数配置存在主观性.提出了一种基于IMOPSO算法的多目标多聚焦图像融合方法,简化了多聚焦图像融合模型,克服了参数配置对经验的依赖性.首先给出了多聚焦图像融合有效的评价指标,然后构造了统一的小波域多聚焦图像融合模型,最后以模型参数作为决策变量,采用IMOPSO算法进行多目标优化搜索.IMOPSO算法不但引入变异算子以避免早熟,而且引入拥挤算子,使Pareto优解尽可能均匀分布于Pareto前端,并采用一种新的自适应惯性权重提高寻优能力.实验结果表明,IMOPSO算法具有更快的收敛速度和更好的寻优能力,同时基于该算法的融合方法也实现了Pare-to最优多聚焦图像融合.
In most of multifocus image fusion methods, the parameter configuration of fusion model is usually based on experience. In this paper, a new multiobjective optimization method of multifocus image fusion based on IMOPSO ( Improved Multiobjective Particle Swarm Optimization) is presented, which can simplify the model of multifocus image fusion and overcome the limitations of traditional methods. First the proper evaluation indices of multifocus image fusion are given,then the uniform model of multifocus image fusion in DWT (Discrete Wavelet Transform) domain is constructed,in which the model parameters are selected as, the decision variables, and finally IMOPSO is designed to optimize the decision variables. IMOPSO not only uses a mutation operator to avoid earlier convergence, but also uses a crowding operator to improve the distribution of nondominated solutions along the Pareto front, and uses a new adaptive inertia weight to raise the optimization capacities. Experiment results demonstrate that IMOPSO has a higher convergence speed and better search capacities, and that the method of multifocus image fusion based on IMOP- SO achieves the Pareto optimal image fusion.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2006年第9期1578-1583,共6页
Acta Electronica Sinica
基金
国防科技大学博士创新基金(No.050303)
关键词
多聚焦图像融合
多目标优化
多目标粒子群算法
离散小波变换
multifocus image fusion
multiobjective optimization
improved multiobjective particle swarm optimization (IMOPSO)
discrete wavelet transform ( DWT )