Spatially informed spectral unmixing

Spectral unmixing methods have traditionally relied on the plethora of information in the spectral domain to resolve subpixel components. Several new methods have utilised the spatial information contained within the hyperspectral image, however these are limited to the spatial resolution of the sensor. In this work, spectral data is unmixed from a spectrometer co-registered with a high spatial resolution RGB camera. The image from the camera is segmented, informing the unmixing process about the number of materials and their abundance. This new algorithm has been titled Image Segmentation Assisted Constrained Spectral-unmixing (ISACS). This presents an intensive computational search. A method to reduce the computational complexity is covered in this work. The performance of these informed spectral unmixing methods are compared to literature methods.