Hyper Spectral Image Denoising via Sparse Representations over Learned Dictionaries

Abstract

Hyperspectral pictures ar corrupted by noise throughout their acquisition and Hyperspectral image (HSI) denoising may be a crucial preprocessing procedure to enhance the performance of the following HSI interpretation and applications. In an HSI, there's an oversized quantity of native and international redundancy in its abstraction domain that may be wont to preserve the main points and texture. additionally, the correlation of the spectral domain is another valuable property that may be used to get smart results. during this work, we have a tendency to propose to expeditiously denoise hyperspectral pictures underneath the idea that the image patches ar thin in an exceedingly correct illustration domain outlined through a lexicon. we have a tendency to propose to rather learn the lexicon from Hyperspectral pictures, a task unremarkably referred to as lexicon learning. variety of HSI knowledge sets are employed in our analysis experiments and that we show that the lexicon learning approach is additional economical to denoise hyperspectral pictures than state-of the- art HSI denoising strategies with mounted dictionaries, at the price of a bigger computation time.