基于SML和PCNN的NSCT域多聚焦图像融合

Multi-focus Image Fusion Based on SML and PCNN in NSCT Domain

针对融合规则带来的虚假边缘、伪影等问题,提出了改进拉普拉斯能量和(Sum-modified Laplacian,SML)和脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)相结合的非下采样Contourlet变换(Non-Sampled Contourlet Transform,NSCT)域融合方法。首先,采用NSCT将每幅源图像分解成包含基本信息的低频子带图像和多幅包含细节信息的带通子带图像。然后,计算各尺度分解图像的SML值,根据值的大小对低频子带图像各像素点进行像素选择。对于带通子带部分,将计算的SML作为PCNN的输入激励,PCNN输出的点火映射图用来选择各子带图像的像素值。最后,将处理后的各子带系数进行NSCT重构得到融合图像。实验结果表明,此算法能很好地改善融合图像的聚焦清晰度,并且与现有的SIDWT,DTCWT,NSCT以及基于PCNN的融合方法相比,所提算法在互信息量、结构相似度以及边缘信息保留量等客观指标方面得到了提高。

Aiming at the false edges and artifacts resulted by the existing fusion rules,a new fusion method based on sum-modified Laplacian（SML）and pulse coupled neural network（PCNN）in non-sampled contourlet transform（NSCT）domain was proposed.Firstly,the source images are decomposed into low frequency sub bands which include basic information and multiple high frequency sub bands with detail information via NSCT.Then,the SML of multiscale coefficients are calculated and used to fuse the low frequency sub-band coefficients.For the high frequency subband coefficients,the calculated SML is taken as the input motivation of the PCNN,and the relative output fire map is adopted to select the pixels in the sub-band images.Finally,the dealt coefficients are reconstructed by the NSCT to get the fused image.The experimental results show that the proposed algorithm excellently improves the focus clarity.Compared with the existing algorithms such as SIDWT,DTCWT,NSCT and PCNN-based method,the objective indexes of mutual information,structural similarity and transferred edge information have been increased.