INB-Lunch-Seminar

Image Deconvolution with Sparse Priors

Jens Hocke

Optical systems often produce blurred images due to imperfections.

This degradations can be removed by deconvolution. Current deconvolution

methods suffer from artifacts and are therefore limited to a moderate degree of blurring.

The deconvolution problem is formulated here as an underdetermined system

of equations, similar to the lately very popular compressed sensing

framework. A sparseness constraint is used to select a plausible solution

out of an infinite set of possible solutions.

The method is tested in a setting with a simulated pinhole camera.