一种新的多聚焦图像融合方法  

Novel Image Fusion Method For Multi-focus Images

作　　者：荣传振[1] 贾永兴[1] 杨宇[1] 朱莹[1] 王渊[1] 

机构地区：[1]解放军理工大学通信工程学院,江苏南京210007

出　　处：《军事通信技术》2017年第1期72-77,共6页Journal of Military Communications Technology

摘　　要：论文提出一种非下采样轮廓波变换(NSCT)域内基于自适应单位链接脉冲耦合神经网络(UL-PCNN)的多聚焦图像融合方法。首先,利用NSCT对源图像进行多尺度和多方向分解;对于低频子带,利用一种基于边缘的图像融合方法;对于高频方向子带,采用局部邻域改进拉普拉斯能量作为UL-PCNN的外部激励,同时利用各子带图像改进的拉普拉斯能量和自适应调节UL-PCNN的链接强度,并选取具有较大点火幅度的系数作为融合图像的高频子带系数;最后,经逆NSCT变换重构融合图像。实验结果表明该方法无论在主观视觉还是客观评价标准上都要优于传统的基于多尺度分解的图像融合方法。An image fusion method for multi-focus images based on Non-subsampled Countourlet Transform（NSCT）and Unit-Linking pulse Coupled Neural Network（UL-PCNN）was proposed.Firstly,the source images were decomposed via NSCT.For the low-frequency sub-images,the fusion was based on the edges of the images,while for the high-frequency directional subimages,the local neighborhood modified-Laplacian were used as the external stimuli of ULPCNN.The sum-modified-Laplacian of each sub-band image was used to adjust the liking strength adaptively,and then the coefficients with larger firing magnitude were selected according to the firing times.Finally,the fused image can be obtained via the inverse NSCT.Experimental results show that the proposed method is better than the typical image fusion algorithms both in subjective and objective evaluation.

关 键 词：图像融合 脉冲耦合神经网络 改进的拉普拉斯能量和 链接强度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]