Abstract: In allusion to the low contrast and the unconspicuous infrared target can easily arise in infrared and visible image fusion, a new algorithm is proposed. The algorithm innovatively applies improved PCNN and region characters to the selection of low frequency subband coefficients and bandpass directional subband coefficients in NSCT domain. The subband coefficients of images to be fused are obtained by using NSCT. The decomposed subband coefficients are processed by using PCNN and the subband coefficients’ fire mapping images are obtained. Low frequency subband coefficients’ fire mapping image uses the method based on region standard deviation to choose the fusion coefficients. Bandpass directional subband coefficients’ fire mapping image uses the method based on region energy to choose the fusion coefficients. The fusion image can be obtained by inverse transform of NSCT. Compared with other algorithms, simulation experiment shows that the proposed algorithm can get a prominent infrared target and better quality fusion image and the objective evaluation indexes of fusion image improve obviously.

Key words: Pulse Coupled Neural Network（PCNN）, region characters, image fusion, Nonsubsampled Contourlet Transform（NSCT）, infrared and visible

摘要： 针对红外图像与可见光图像融合中容易产生红外目标不明显、对比度不高的问题，提出了一种新的融合算法。该算法创新地将PCNN与区域特征应用到NSCT域内低频和带通子带系数的选择上。通过NSCT分解得到待融合图像的子带系数。运用PCNN对分解后的子带系数进行处理，得到子带系数的点火映射图。低频子带点火映射图采取基于区域标准差的方法选取融合系数。带通子带点火映射图采取基于区域能量的方法选取融合系数。融合图像通过NSCT逆变换可以得到。仿真实验表明与其他算法相比，该算法能够得到红外目标突出、质量更好的融合图像，图像客观评价指标提升明显。

关键词: 脉冲耦合神经网络（PCNN）, 区域特征, 图像融合, 非下采样Contourlet变换（NSCT）, 红外与可见光