改进的NSCT与NMF结合的图像融合方法

Improved image fusion method by combining NSCT with NMF

基于人类视觉系统和源图像特性,对基于非下采样Contourlet变换与非负矩阵分解(NMF)图像融合算法进行了改进。在非负矩阵分解过程中,适当地选取特征空间的维数能够获得原始数据的局部特征,低频部分使用非负矩阵分解的方法进行融合,高频部分使用活性测度和一致性验证的方法进行融合。实验结果表明,该算法具有较强的鲁棒性,融合图像边缘的清晰度和连续性也较理想。

This algorithm of image fusion with nonsubsampled contourlet transform and non-negative matrix factorization is improved based on human visual system（HVS） and source image characteristics.It is shown that the local feature of original data can be obtained by choosing a suitable dimension of the feature subspace in non-negative matrix factorization.Therefore NMF is used in low-frequency,and activity measurement and the consistency of the method is used in high-frequency.Experiment results show that the proposed fusion technique is robust and the fusion images have ideal clear and continue edges.