''Minimum Volume Simplex Analysis: A Fast Algorithm for Hyperspectral Unmixing''

Hyperspectral unmixing is the process in which a hyperspectral image is expressed by its constituent components. The specialized area of the linear unmixing model consists of expressing an image as a product of two matrices, namely the endmember matrix which contain the signatures of the components of the hyperspectral image and the abundance matrix which contain the fraction of an endmember contribution for each band of the hyperspectral image. In this talk I will present a two year progress of my research in hyperspectral unmixing and specifically I will present the linear unmixing algorithm based on the minimum volume enclosing theory called Minimum Volume Simplex Analysis (MVSA). I will present how its computational complexity can be dramatically reduced and how it can be applied in High Performance Computing. I will also present a very recent research on a Robust version of MVSA which take into consideration also the noise present in the Hyperspectral Image and how this consideration can dramatically improve the performance of MVSA for particularly noisy scenarios.