Computational and Applied Math Proseminar

Thursday, April 24, 2008, 12:15 p.m. PSA 206

Wolfgang Stefan

Dept. Math. & Stats.

Improved total variation-type regularization using higher-order edge detectors

Abstract We present a novel deconvolution approach that simultaneously deblurs and detects edges in piecewise smooth signals. Both edges and smooth regions, separated by jump discontinuities, are preserved. The method uses a two step procedure: The polynomial annihilation method combined with total variation (TV) deconvolution obtains an estimate of the location of jump discontinuities in blurred noisy data. This information is used to determine the order for a higher-order TV regularization which is then utilized in the signal restoration. As compared to those obtained with standard TV, signal restorations are more accurate representations of the true signals, as measured in a relative $l^2$ norm, and can also be used to obtain a more accurate estimation of the locations and sizes of the true jump discontinuities.

For further information please contact: mittelmann@asu.edu