Open AccessData Adaptable Sparse Reconstruction for Hyperspectral Image Recovery from Compressed Measurements
Author(s) -
H. S. Prasantha,
M. Tech
Publication year - 2020
Publication title -
international journal of innovative technology and exploring engineering
Language(s) - English
Resource type - Journals
ISSN - 2278-3075
DOI - 10.35940/ijitee.c9007.109320
Subject(s) - hyperspectral imaging , compressed sensing , prior probability , computer science , algorithm , regularization (linguistics) , iterative reconstruction , sparse matrix , matrix (chemical analysis) , rate of convergence , computer vision , artificial intelligence , gaussian , key (lock) , materials science , bayesian probability , physics , computer security , quantum mechanics , composite material
Hyperspectral image compression using compressive sensing technique is very much important in the area of satellite image compression because it can greatly en hance the compression rate. The research work proposes a novel data adaptable sparse reconstruction algorithm for the reconstruction of hyperspectral images from compressive sensing measurements. In the proposed algorithm, compressive sensing technique is used for the compression of HSIs, where Gaussian i.i.d. matrix is used to generate compressive sensing measurements. The algorithm solves the optimization problem containing total variation regularization and data adaptable parameter terms. The regularization terms are added to provide hyperspectral data characteristics as priors into the objective function. The addition of priors helps in convergence of the algorithm into the desired solution. The algorithm alternatively reconstructs the end member matrix and abundance matrix instead of reconstructing the entire HSI at once, thereby reducing the computational complexity at the reconstruction process. To nullify the effect of modelling errors, debiasing is performed.