Noise analysis and restrain in the fusion process of remote sensing

Considering the noise problem of multi-spectral images and SAR, several image noise models are described in this paper. The idea of non-negative matrix factorization and rough set are used to carry out the fusion experiment. Non-negative factorization is a matrix factorization method in the restrictive condition which all elements are non-factorization. The theory of rough set is mainly about the combination between knowledge and classification, which has an effect on de-noising aspect to the image noises. On the other side, the HIS and wavelet algorithm are adopted to the multi-spectral image fusion. It can not only hold more spectral information but also improve the spatial resolution greatly. The results from the experiment on the images specified the different types of noise, and get the conclusion in the help of some evaluation methods.