Applications of CS based spectrum recovery in hyperspectral images

In this paper, we consider the question regarding high spectral redundancy in hyperspectral image (HSI) which causes problems of storage as well as transmission. Efficient spectrum sampling and recovery approaches are needed. This paper addresses the application of compressed sensing (CS) based spectrum recovery in spectra and HSIs. We firstly analyze the sparsity of spectra. Then we design the appropriate processes of spectrum recovery from efficiency and speed aspects. The proposed strategy was tested using the spectral data from USGS dataset and an AVIRIS HSI data. The experiments demonstrate that it results in a high reconstruction rate and low mean square errors compared with traditional interpolation method. Besides, the reconstructed HSIs have been applied into target detection to show the feasibility and applicability, illustrating the spectrum recovery´s superiority.