Poisson-Gaussian mixed noise removing for hyperspectral image via spatial-spectral structure similarity

Traditional hyperspectral denoising methods assumed that the noise to be removed follows the additive Gaussian model, which is not true for real situation. The noise in hyperspectral data is signal dependent, Poisson-Gaussian mixed noise model is more accurate to describe it. On the other hand, the noise in hyperspectral data distributes on spatial and spectral dimension, panchromatic imagery denoising method can not be used directly to hyperspectral imagery. There are many similar spatial-spectral structures in every scene, through utilizing these similarities into denoising process, the spatial and spectral redundancy and correction would be exploited, thus the denoising performance can be improved greatly. Based on these ideal, we propose hyperspectral Poisson-Gaussian mixed noise removing method based on spatial-spectral structure similarity. Numerical experiments on different testing data and theoretical illustration demonstrate that proposed denoising method obtain higher performance than the state-of-art methods.