Application of Several Non-negative Matrix Factorization-Based Methods in Remote Sensing Image Fusion

Multi-spectral image keeps good multi-spectral feature in spite of its low special resolution, which canpsilat describe the earthpsilas surface information in detail. In contrast, panchromatic image keeps high special resolution but with limitations in spectral feature. The advantages of both multi-spectral image and panchromatic image can be integrated in the same image with fusion technique, so the fusion image with high quality will make it easier to visually evaluate or process by computers. In this paper, several new non-negative matrix factorization (NMF)-based fusion algorithms are presented for fusing multi-spectral image and panchromatic image, including NMF and principal component transform (PCA)-based fusion method (N-PCA), NMF and lifting wavelet transform (LWT)-based fusion method (N-LWT). The capabilities of each method are summarized. The experiments prove that NMF applied in the field of remote sensing image fusion can improve the quality of fusion image. A comparison of all proposed new fusion methods with PCA fusion method shows that the formers are better than the later, and the N-LWT fusion method is the best one among all methods.