基于NSST域的改进加权非负矩阵分解的图像融合

针对加权非负矩阵分解中算法复杂度较高的问题,提出一种基于加权非负矩阵分解和双通道脉冲耦合神经网络的图像融合的改进算法。首先,对已经配准的两个源图像进行非下采样Shearlet变换;然后,对于图像低频子带,采用改进的WNMF的算法,动态更新权值矩阵,更好地提取图像特征信息。对于高频子带,采用改进双通道脉冲耦合神经网络的算法,链接强度值采用块的梯度值,更好地保留图像的微小细节信息;最后,经过非下采样Shearlet的逆变换得到融合图像。实验表明,将加权非负矩阵分解与双通道脉冲耦合神经网络相结合,不仅能很好的提取图像的特征信息,保留更多细节信息;同时双通道的脉冲耦合神经网络的方法能提高算法运行效率。

Image Fusion Based on Improved Weighted Nonnegative Matrix Decomposition Based on NSST Domain

Aiming at the problem of high complexity in weighted nonnegative matrix decomposition, an improved algorithm of image fusion based on weighted nonnegative matrix decomposition and dual channel pulse coupled neural network is proposed. Firstly, the Shearlet transform is applied to the two source images that have been registered. Then, the improved WNMF algorithm is used to dynamically update the weight matrix for the image low frequency subband and the image feature information is extracted better. In this paper, the algorithm of improving the dual channel pulse coupled neural network is used to improve the detail information of the image by using the gradient value of the block proposed in this paper. Finally, the fusion image is obtained by inverse transformation of the non-subsampled Shearlet. Experiments show that the combination of weighted nonnegative matrix decomposition and pulsed coupled neural network not only can extract the characteristic information of the image, but also keep more detailed information. At the same time, the dual channel pulse coupled neural network method can improve the efficiency of the algorithm.