A Novel Algorithm For Blind Deconvolution Applied To The Improvement Of Radiographic Images

A novel algorithm for blind deconvolution is proposed in this work, which does not require any previous information concerning the image to be unfolded but solely an assumed shape for the PSF. This algorithm, incorporating a Richardson-Lucy unfolding procedure, assesses the overall contrast for each image unfolded with an increasing w, seeking for the highest value. The basic idea behind this concept is that when the spatial resolution of the image is improved, the contrast is improved too, because the pixel overlapping diminishes. Trials with several different images acquired with neutron and gamma-ray transmission radiography have been carried out in order to evaluate the correctness of the proposed algorithm. It has been found that for a steadily increasing w, the overall contrast increases, reaches a maximum and then decreases. The w-value yielding the highest contrast can be achieved after 1 to 3 iterations and further iterations do not affect it. Images deconvoluted with this value, but with a higher number of iterations, exhibit a better quality than their companions deconvoluted with neighbor values, corroborating thus the best w-value. Synthetic images with known resolutions return the same w-values used to degrade them, showing thus the soundness of the proposed algorithm.