Topics in Unconventional Imagery

Abstract

Abstract : Two problems in unconventional imagery, were worked on, (a) an exact solution to the in age turbulence problem (also called the 'blind deconvolution' problem); and (b) closed-form maximum entropy (M.E.) image restoration Progress on (a) was as follows. It was found that by dividing the image spectra of two short-exposure images of an incoherent object viewed through random turbulence, a system of linear equations can be generated. The unknowns of the equations are the sampled values of the two point spread functions characterizing the two images. These can be found, with arbitrary precision, by simple inversion of the equations. Then the object is restored by inverse filtering the two images with transfer functions generated from the known point spread functions. The approach works perfectly in the absence of additional randomness due to noise of detection, and tolerates small amounts of such noise. Progress on (b) was as follows. Doctoral student David Graser tested out the closed-form M.E. approach by computer simulation. Two widely-used classes of test objects--point sources and edge sources--were used as inputs, and these were imaged using Gaussian spread functions of given halfwidths. The M.E. outputs were found to be, overall, superior to corresponding outputs using clipped inverse-filtering and Wiener filtering.