Home – Home – Jornals – Avtometriya 2020 number 1

An approach to filtration of hyperspectral images distorted by the Gaussian additive noise is proposed. The approach is based on using the property of interchannel redundancy of such images. The developed algorithm of noise filtration allows maintaining the planimetric and brightness portraits of objects in individual components of the hyperspectral image, in contrast to algorithms of linear component-by-component and vector filtration, as well as the algorithm of averaging over a set of components. The numerical results obtained in the study testify to the advantage provided by interchannel gradient reconstruction in terms of the accuracy of recovery of hyperspectral image components distorted by additive noise. The efficiency of the proposed approach is demonstrated by an example of processing of real hyperspectral images.