English

Multiscale Higher Order TV Operators for L1 Regularization

Numerical Analysis 2018-03-30 v2

Abstract

In the realm of signal and image denoising and reconstruction, 1\ell_1 regularization techniques have generated a great deal of attention with a multitude of variants. A key component for their success is that under certain assumptions, the solution of minimum 1\ell_1 norm is a good approximation to the solution of minimum 0\ell_0 norm. In this work, we demonstrate that this approximation can result in artifacts that are inconsistent with desired sparsity promoting 0\ell_0 properties, resulting in subpar results in {some} instances. With this as our motivation, we develop a multiscale higher order total variation (MHOTV) approach, which we show is related to the use of multiscale Daubechies wavelets. We also develop the tools necessary for MHOTV computations to be performed efficiently, via operator decomposition and alternatively converting the problem into Fourier space. The relationship of higher order regularization methods with wavelets, which we believe has generally gone unrecognized, is shown to hold in several numerical results, although notable improvements are seen with our approach over both wavelets and classical HOTV.

Cite

@article{arxiv.1703.02404,
  title  = {Multiscale Higher Order TV Operators for L1 Regularization},
  author = {Toby Sanders and Rodrigo B. Platte},
  journal= {arXiv preprint arXiv:1703.02404},
  year   = {2018}
}
R2 v1 2026-06-22T18:38:30.818Z