Estimate the noise of high-spectral data with multiplicative noise model

The multiplicative noise model for HS data cube was introduced in this paper. Based on this model, an algorithm was also developed to estimate the noise of the HS data in spatial and spectral domain. The comparison among the classical additive noise model, Poisson noise model and multiplicative was also discussed. The noise estimation experiments show that the multiplicative noise model is reasonable and suitable for HS data. The good performance of the NSR algorithms validated the effectiveness of the multiplicative noise model. The experiments also show that the multiplicative noise model have unique characteristics in information extraction of land cover. Based on the Multiplicative Noise model, we found that some natural objects like waters can be easily distinguished and extracted from the HS data using the NSR algorithm that we proposed in this study.