Simulation of hyperspectral image with existing Sentinel and AVIRIS data using distance functions

Hyperspectral data may provide abundant and fine surface spectral details. Their applications can, however, be constrained by their limited bandwidth and small coverage area. As the data having low bandwidth, each band has its significance for studies such as mineral composition, crop monitoring, and identification of materials. But hyperspectral data is expensive, and availability is less compared to multispectral data. Simulation of hyperspectral data with existing Sentinel and AVIRIS data will be an advantage for the analysis. This study concentrates on obtaining similar spectra using distance functions. The Chebyshev distance and spectral angle mapper (SAM) distances are combined to get the advantage of both the distance of vector coordinates and the pattern of the spectra. Each pixel of the test image is verified for the similarity of the whole reference image using distance functions to get similar spectra. These similar spectra are all combined to construct the simulated hyperspectral image. Finally, multispectral image is simulated to hyperspectral imagery. The relevance and novelty of this study is that it uses distance functions to simulate hyperspectral data, and it was discovered that the suggested methodology has good spectral correlation accuracy when compared to the current AVIRIS-NG dataset. The simulated hyperspectral image is validated with the AVIRIS image using normalized cross-correlation to obtain each pixel’s correlation. The normalized cross-correlation of test site-a is obtained as 95.35% and test site-b is 82.28% under 0.9 to 1, and the colour of the false colour composite is identical to the original AVIRIS image. © 2021, Saudi Society for Geosciences.