Hyperspectral Image Super-Resolution
The aim of this project is to combine information coming from a hyperspectral and a multispectral (or a RGB image). It is common that hyperspectral sensors lack in spatial resolution, whereas on the other hand multispectral sensors do not capture detailed information of the scene’s spectrum. Combining the previous two products can enable not only to super-resolve the hyperspectral image, but also to obtain a physically plausible explanation about the scene’s endmembers and abundances at high resolution (see image). This is done by imposing the constraints guided by linear spectral unmixing. As an extension of this project we estimate the relative spectral and spatial responses of the two sensors directly from the data. The relative sensor characteristics are required to perform the super-resolution.
C. Lanaras, E. Baltsavias, K. Schindler, Hyperspectral Super-Resolution by Coupled Spectral Unmixing, International Conference on Computer Vision (PDF, 4.9 MB) (ICCV), Santiago, Chile, 2015
C. Lanaras, E. Baltsavias, K. Schindler, Advances in Hyperspectral and Multispectral Image Fusion and Spectral Unmixing (PDF, 3.2 MB), ISPRS Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XL-3/W3, pp. 451–458.
C. Lanaras, E. Baltsavias, K. Schindler: Estimation of Relative Sensor Characteristics for Hyperspectral Super-Resolution (PDF, 952 KB), IEEE Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), Los Angeles, USA, 2016
, Emmanuel Baltsavias