Contact
joseph.garrett@ntnu.no

NTNU's [HYPSO-1](https://www.ntnu.edu/web/smallsat/ntnu-smallsat-lab) satellite records images of the ocean in 120 spectral bands which can resolve many unexpected features in the earth's oceans. However, its spatial resolution is not quite sufficient to see into the narrow channels of many fjords. Therefore, we would like to develop a super-resolution technique that can clearly separate the land from the water in narrow fjords (~200 m) while also preserving the integrity of the water color data. Attaining this resolution is important for HYPSO-1's main objective, which is to monitor harmful algae blooms. 

Many types of hyperspectral super-resolution exist. This topic will focus on super-resolution techniques that only rely on the hyperspectral image, and not on data fusion with multispectral images. Non-negative matrix factorization is the suggested the basis for the algorithm, but other techniques can also be explored. 

This topic is related to the ELO-Hyp project, in which NTNU is collaborating with several Romanian institutions. As part of the project, a simple version of non-negative matrix factorization has been developed, which can serve as a starting point for the development of the super-resolution algorithm. 

Links:
https://blog.paperspace.com/image-super-resolution/