Scale-Space Properties for Discrete Cross-Diffusion Filtering
摘要
In this paper we consider \(2\times 2\) nonlinear cross-diffusion models for image restoration. Digital images are sampled on a pixel grid, so it is important to ascertain whether the outcomes obtained in a continuous framework carry over to the practical relevant fully-discrete setting. Here we explore fully-discrete filters which reveal similar features as their continuous cross-diffusion counterparts as scale-space invariance properties as well as the existence of Lyapunov functions.