Reconstruction of Images From a Compact Spectral Camera

Abstract

Hyperspectral imaging provides a lot of information about the imaged scene, and has several applications. The purpose of this work is to reconstruct hyperspectral images from pictures captured with a compact hyperspectral camera. The camera has a linearly variable filter placed over the image sensor, so that different pixels sample the incoming light at different wavelengths. Spectral information is collected from the entire scene by scanning the camera, but extensive processing of the raw data is required to reconstruct a regular hyperspectral image.An important basis for reconstructing images is an accurate model for the camera's response to the incoming radiance. This report describes the method used for calibrating the camera, and the results from the calibration experiments.Methods for reconstructing hyperspectral images are discussed. The ideal method for image reconstruction was found not to be feasible, because a very large amount of computer memory would be needed. Two alternative methods are presented, and these methods have been implemented in software for image reconstruction. Finally, successfully reconstructed images from an experiment are presented. For further work on the camera it is recommended to investigate the possibility of applying parallel computing to the image reconstruction in order to cut down on processing time.