Impulse Denoising From Hyper-Spectral Images: A Blind Compressed Sensing Approach

TL;DR

This paper introduces a novel blind compressed sensing method for removing impulse noise from hyperspectral images by learning sparsifying dictionaries adaptively, resulting in significant PSNR improvements.

Contribution

It presents a new BCS-based algorithm that learns spatial and spectral dictionaries during denoising, unlike traditional methods with fixed dictionaries.

Findings

Over 5 dB PSNR improvement compared to existing techniques

Effective removal of sparse impulse noise from hyperspectral images

Utilizes spatial redundancy and spectral correlation for enhanced denoising

Abstract

In this work we propose a technique to remove sparse impulse noise from hyperspectral images. Our algorithm accounts for the spatial redundancy and spectral correlation of such images. The proposed method is based on the recently introduced Blind Compressed Sensing (BCS) framework, i.e. it empirically learns the spatial and spectral sparsifying dictionaries while denoising the images. The BCS framework differs from existing CS techniques - which assume the sparsifying dictionaries to be data independent, and from prior dictionary learning studies which learn the dictionary in an offline training phase. Our proposed formulation have shown over 5 dB improvement in PSNR over other techniques.