Numerical analysis of shell-based geometric image inpainting algorithms and their semi-implicit extension

TL;DR

This paper develops a theoretical model for shell-based geometric image inpainting algorithms, analyzing artifacts and proposing a semi-implicit extension that eliminates kinking artifacts, supported by continuum limit analysis and connection to random walks.

Contribution

It introduces a continuum limit model for shell-based inpainting, explaining artifact origins and demonstrating that a semi-implicit method removes kinking artifacts.

Findings

Semi-implicit extension eliminates kinking artifacts.

Continuum limit model explains artifact formation.

Analysis applicable to low-regularity images.

Abstract

In this paper we study a class of fast geometric image inpainting methods based on the idea of filling the inpainting domain in successive shells from its boundary inwards. Image pixels are filled by assigning them a color equal to a weighted average of their already filled neighbors. However, there is flexibility in terms of the order in which pixels are filled, the weights used for averaging, and the neighborhood that is averaged over. Varying these degrees of freedom leads to different algorithms, and indeed the literature contains several methods falling into this general class. All of them are very fast, but at the same time all of them leave undesirable artifacts such as "kinking" (bending) or blurring of extrapolated isophotes. Our objective in this paper is to build a theoretical model, based on a continuum limit and a connection to stopped random walks, in order to understand why these artifacts occur and what, if anything, can be done about them. At the same time, we consider a semi-implicit extension in which pixels in a given shell are solved for simultaneously by solving a linear system. We prove (within the continuum limit) that this extension is able to completely eliminate kinking artifacts, which we also prove must always be present in the direct method. Although our analysis makes the strong assumption of a square inpainting domain, it makes weak smoothness assumptions and is thus applicable to the low regularity inherent in images.