Complex-domain Joint Broadband Hyperspectral Image Denoising

Abstract

In this paper, we propose a novel complex domain denoising algorithm for hyperspectral data. The algorithm is based on the Complex Domain Block-Matching 3D (CDBM3D) filter and on similarity of hyperspectral data, which are usually slow varying for close values of wavelengths. Singular Value Decomposition (SVD) of the hyperspectral data is used in order to define an optimal small dimension data subspace. The CDBM3D is applied for 2D images of this subspace. The efficiency of the algorithm is demonstrated in simulation tests and for experimental data obtained by spectrally resolved digital holography of a transparent color object. It is proved that the proposed filtering algorithm retrieves amplitude and phase distributions even from very noisy data.