CS700:Graduate Seminar in Computer Science & Informatics

Estimation of PSF parameters for hyperspectral imaging
Sebastian Berisha, Department of Mathematics and Computer Science

Hyperspectral imaging degradation is generally a result of axial optical aberrations in the image acquisition system. Axial optical aberrations arising from wavelength-dependent variations in the index of refraction of the incident light can lead to significant blurring of image intensities in certain parts of the spectral range. The estimation of the overall system dependent point spread function (PSF) is necessary for hyperspectral imaging data processing and analysis methods. We use a spectrally varying PSF model pertaining to ground-based hyperspectral astronomical observations. The overall PSF identification process is reduced to finding the PSF parameters from the noisy hyperspectral datacube of an isolated star. We pose the hyperspectral PSF estimation problem in a separable nonlinear least squares framework. The proposed numerical approach allows for a joint estimation of the hyperspectral PSF parameters and the star spectrum without prior estimation of the noise variance. We do not assume the PSF to be constant in any wavelength range and thus avoid the need for determining the number of wavelengths used for spectral binning. The experimental results illustrate the effectiveness of the resulting numerical scheme.