English

Image Processing Variations with Analytic Kernels

Analysis of PDEs 2012-04-06 v1

Abstract

Let fL1(Rd)f\in L^1(\R^d) be real. The Rudin-Osher-Fatemi model is to minimize uBV˙+λfuL22\|u\|_{\dot{BV}}+\lambda\|f-u\|_{L^2}^2, in which one thinks of ff as a given image, λ>0\lambda > 0 as a "tuning parameter", uu as an optimal "cartoon" approximation to ff, and fuf-u as "noise" or "texture". Here we study variations of the R-O-F model having the form infu{uBV˙+λK(fu)Lpq}\inf_u\{\|u\|_{\dot{BV}}+\lambda \|K*(f-u)\|_{L^p}^q\} where KK is a real analytic kernel such as a Gaussian. For these functionals we characterize the minimizers uu and establish several of their properties, including especially their smoothness properties. In particular we prove that on any open set on which uW1,1u \in W^{1,1} and u0\nabla u \neq 0 almost every level set {u=c}\{u =c\} is a real analytic surface. We also prove that if ff and KK are radial functions then every minimizer uu is a radial step function.

Keywords

Cite

@article{arxiv.1204.1097,
  title  = {Image Processing Variations with Analytic Kernels},
  author = {John B. Garnett and Triet M. Le and Luminita A. Vese},
  journal= {arXiv preprint arXiv:1204.1097},
  year   = {2012}
}

Comments

18 pages

R2 v1 2026-06-21T20:44:56.421Z