Hyper-Spectral Remote Sensing Image De-Noising with Three Dimensional Wavelet Transform Utilizing Smooth Nonlinear Soft Thresholding Function

A hyper-spectral image can be subject to additive noise during the acquisition process. The main objective in noise removal is to enhance the visual quality of the corrupted image using de-noising techniques. Most of the techniques try to discard the noise in the pre-processing stage prior to further analysis. The main focus in this paper is removing the noise from hyper-spectral remote sensing images. A new method is proposed for image de-noising by applying a smooth nonlinear soft threshold on high frequency sub-bands of the images after applying 3D un-decimated wavelet transform (3D-UWT). This proposed threshold function is referred to as the improved soft (smooth nonlinear)thresholding function. The proposed method is compared with de-noising based on 3D-UWT using standard hard and soft thresholding techniques. Results show the superiority of the proposed method over the standard and alternative methods in the literature by means of visual quality and peak signal to noise ratio (PSNR).