Hyperspectral image spectral denoising using pure and wavelet with sparse restoration

Hyper Spectral Image (HSI) denoising are usually classified as signal dependent and signal independent methods of denoising. In an HSI data, both signal dependent and signal independent noises co-exist and it is said as a mixed noise condition. The proposed method in this paper denoises an HSI data, in which a signal dependent Poisson noise is degraded by a signal independent AWGN. This method uses lasso regression along with Poisson based Stein Unbaised Risk Estimation to remove the Poisson degraded by AWGN. The regularization parameter is updated for every band of data and it uses wavelet for image denoising. This method is evaluated in terms SNR and IEF and compared with the other existing methods of denoising.