Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (7): 177-180.DOI: 10.3778/j.issn.1002-8331.1509-0099

Previous Articles     Next Articles

Adaptive image fusion algorithm based on pulse coupled neural networks

WANG Hongmei, FU Hao   

  1. School of Astronautics, Northwestern Polytechnical University, Xi’an 710072, China
  • Online:2017-04-01 Published:2017-04-01

脉冲耦合神经网络自适应图像融合算法研究

王红梅,付  浩   

  1. 西北工业大学 航天学院,西安 710072

Abstract: As a new network model, Pulse Coupled Neural Networks(PCNN) has been widely used in many fields. Aiming at the insufficient of traditional image fusion by pulse coupled neural networks algorithm, a novel adaptive PCNN image fusion algorithm is proposed. Two features are used as external stimulus of PCNN to extract the complementary information of source images. The linking strength parameters of PCNN are determined adaptively according to contrast of source images. Through analyzing the traditional PCNN algorithms in obtaining the optimal fusion result, a novel method is proposed which can acquire the optimal fusion result by comparing the structural similarity of the fused image at each iteration. Experimental results for visible and infrared images show that the proposed algorithm is effective for image fusion.

Key words: image fusion, pulse coupled neural networks, structural similarity, objective evaluation

摘要: 作为一种新型的神经网络模型,脉冲耦合神经网络(PCNN)已经在众多领域得到了应用。针对现有脉冲耦合神经网络图像融合算法存在的不足,提出了一种新的自适应PCNN图像融合算法。提取原始待融合图像的互补特征作为PCNN的外部输入,并通过提取待融合图像的对比度特征自适应确定PCNN的链接强度参数;分析了传统PCNN获取最优图像融合结果的方法,探索性地将结构相似度引入到PCNN融合结果的评价中,为PCNN最优融合结果的获取提供了很好的借鉴作用。通过红外和可见光等图像的仿真实验结果表明,提出的融合算法是有效的。

关键词: 图像融合, 脉冲耦合神经网络, 结构相似度, 客观评价