[Previous] | [Session 8] | [Next]
F. Kamalabadi (Center for Space Physics and Department of Electrical and Computer Engineering), T. Cook, V. Taylor, S. Chakrabarti (Department of Astronomy and Center for Space Physics)
We present a novel concept for spectral imaging suitable for the study of the interstellar medium at ultraviolet wavelengths. The technique is capable of performing simultaneous multi-object spectroscopy on point sources and imaging spectroscopy of faint diffuse sources. Unlike conventional specral imaging techniques, our method encodes three dimensions of data (RA, DEC, wavelength) into a two-dimensional signal, using an optical system where photons from the entire scence are collected at all times. Inversion of this data back into the three-dimensional space is accomplished by formulating a tomographic reconstruction problem. Intrinsic presence of noise, however, poses difficulties on conventional inversion methods such as the filtered back-projection (FBP) or the minimum-norm least square methods such that they often produce unacceptable results. This limitation is overcome by incorporating sophisticated regularization schemes. We investigate issues concerning the required inversion, and develop a suitable approach to image reconstruction using stochastic regularization by minimizing appropriate L1 and L2-norm functionals. We demonstrate the performance of our technique quantitatively in comparison with conventional spectral imaging techniques.
If the author provided an email address or URL for general inquiries, it is a
s follows:
http://net.bu.edu/spinr.html