Abstract: Curvelet has the ability of acquiring high dimension singularity of image. Utilizing this characteristic, an algorithm of image fusion based on Pulse Coupled Neural Networks（PCNN） and optimization of evaluated rule in curvelet domain is proposed. By using curvelet transform, the input images are decomposed into sub-images with various scales. Then, based on the global coupled character of PCNN, a fusion rule is given and the high frequency fused coefficients can be generated by the rule. After that, the low frequency fused coefficients can be generated by optimization of some evaluated rules and the fused image is obtained by inverse curvelet transform. Its validity and advantage are shown by experimental results and the analysis of comparing with other algorithms.

Key words: image fusion, curvelet transform, pulse coupled neural networks, optimization of evaluated rule

摘要： 利用曲波变换能够准确捕获图像高维奇异信息的特点，提出了一种在曲波域中基于脉冲耦合神经网络和最优化评价准则的图像融合方法。该方法用曲波变换对输入图像进行多尺度分解，再利用脉冲耦合神经网络的全局耦合特性对高频子带曲波系数进行选取，定义图像融合的目标函数，根据最优化目标函数确定低频曲波系数的融合权值，进行曲波逆变换得到融合图像。实验结果以及与其他算法的比较分析表明了算法的有效性和优越性。

关键词: 图像融合, 曲波变换, 脉冲耦合神经网络, 最优化评价准则