Subregion Technique with Opimal Local Parameter Sets for Image Deconvolution
Karn Patanukhom, Akinoki Nishihara · 2005
A subregion technique for image deconvolution is proposed. This technique improves restoration results of conventional deconvolution methods by separately restoring each local region of an image with different restoration processes. Optimal local restoration parameter set is designed for each restoration process to optimize a restoration error in its corresponding local region. An iterative algorithm to estimate the optimal parameter sets for subregion scheme is introduced. Instead of using an original pair, which is a pair of original and degraded images, to find a characteristic between the restoration error and the parameter set, the proposed algorithm uses a pseudo pair, which is a pair of an estimated original image and its degraded version having the same degradation condition as the original pair. Simulation results are presented to demonstrate the promising performance of the subregion technique.