Sparse Reconstruction of Hyperspectral Image using Bregman Iterations

Abstract

Hyperspectral image processing plays an important role in satellite communication. Hyperspectral Image (HSI) processing generally requires very high ‘computational resources’ in terms of computational time and storage due to extremely large volumes of data collected by imaging spectrometers on-board the satellite. The bandwidth available to transmit the image data from satellite to the ground station is limited. As a result, Hyperspectral image compression is an active research area in the research community in past few years. The research work in the paper proposes a new scheme, Sparsification of HSI and reconstruction (SHSIR) for the reconstruction of hyperspectral image data acquired in Compressive sensing (CS) fashion. Compressed measurements similar to compressive sensing acquisition are generated using measurement matrices containing gaussian i.i. d entries. Now the reconstruction is solving the constrained optimization problem with non smooth terms. Adaptive Bregman iterations method of multipliers is used to convert the difficult optimization problem into a simple cyclic sequence problem. Experimental results from research work indicates that the proposed method performs better than the other existing techniques.