An unsupervised hyperspectral image super-resolution technique trains a neural network on fully synthetic abundance maps that mimic real statistics via the dead leaves model, then reconstructs the high-resolution image from upsampled abundances and endmembers.
Dataset & Setup The Urban dataset, a widely-used dataset where there is an exploitable ground-truth of the abundance maps and end- members, is considered here
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.IV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Unsupervised Super-Resolution of Hyperspectral Remote Sensing Images Using Fully Synthetic Training
An unsupervised hyperspectral image super-resolution technique trains a neural network on fully synthetic abundance maps that mimic real statistics via the dead leaves model, then reconstructs the high-resolution image from upsampled abundances and endmembers.