SPOT image fusion and application based on Non-negative Matrix Factorization

Change detection has attracted much attention for the application of disaster monitoring. Multi-band SPOT remote sensing images are wildly used because of the abundant spectrum information. But the data redundancy needs to be eliminated using image fusion technique, which has developed rapidly. Non-negative matrix factorization (NMF) has been proven to be a very effective image fusion tool to extract useful information. In this paper, a time-sharing fusion method based on NMF is proposed for the purpose of change detection. First, the 3 band SPOT images of the same time were fused using NMF algorithm for different time period, respectively. Then the residual image, which was generated using the fused images of different time, was used to indicate the changed area. The results are able to present changed area clearly and show better performance compared to the traditional methods. It demonstrates that first using image fusing for images of different time period is able to exclude redundancy and keep useful information which helps to detect the changed area with higher accuracy.