Quality-based band selection for hyperspectral image visualization

Displaying a hyperspectral image on a trichromatic monitor poses the problem of reducing the large number of bands to just three colour channels. In this letter, a new visualization method is proposed, based on an image quality score computed on each band of the hyperspectral image. A detail enhancement post-processing step increases the prominence of small-scale information in the visualization. The experimental results obtained on three widely used hyperspectral datasets prove the effectiveness of the proposed approach in producing high-quality visualizations from the point of view of the chosen objective metrics.