Paul Rodriguez, LANL/T-7

Total Variation (TV) regularization has become a very popular method for a wide variety of image restoration problems, including denoising and deconvolution. Recently, a number of authors have noted the advantages, including superior performance with certain non-Gaussian noise, of replacing the standard l2 data fidelity term with an l1 norm.

A novel method, the Iterative Reweighted Norm algorithm, is proposed for solving the generalized TV functional, which includes the l2-TV and and l1-TV problems. The proposed method is rather simple but very fexible and computationally efficient.

Results for denoising, deconvolution, and demosaicing will be presented, with time-performance comparisons with other well-known, established algorithms.