A Novel Algorithm of Multi-Sensor Image Fusion Using Non-negative Matrix Factorization

The theory of NMF(Non-negative Matrix Factorization) technique is investigated and a novel method based on NMF is presented for image fusion in this paper. It is shown that the local feature could be obtained by choosing suitable dimension of the feature subspace in NMF. In this paper, the principle of extracting the global feature of an image using NMF is analyzed and a visual example is given. We point out that when using NMF, if the dimension of the feature subspace is set to 1, the resultant feature base is just the fusion result of the original input images and the feature base contains the global feature of the original images. Experiments show that the method proposed in the paper performs better in preserving the feature information for the test images than wavelet transform does.