Digital inpainting based on the Mumford–Shah–Euler image model

Abstract

Image inpainting is an image restoration problem, in which image models play a critical role, as demonstrated by Chan, Kang and Shen's recent inpainting schemes based on the bounded variation [10] and the elastica [9] image models. In the present paper, we propose two novel inpainting models based on the Mumford-Shah image model [37], and its high order correction-the Mumford-Shah-Euler image model. We also present their efficient numerical realization based on the ¡-convergence approximations of Ambrosio and Tortorelli [1, 2], and De Giorgi [18].

GPU CUDA accelerated Image Inpainting using Fourth Order PDE equation

“…This method makes improvement in inpainting longer distances than before. Esedoglu and Shen adapted image segmentation to the inpainting problem [6]. This model can be solved rather quickly but this model cannot connect image across large distance.…”

A Single-Image Super-Resolution Method for Texture Interpolation

“…Since these algorithms function in essence as low-pass filters, they tend to introduce blurring and blocking artifacts, which are visible especially in areas containing high frequency content such as edges as well as in textures. Several classes of more sophisticated algorithms were proposed in recent years [1]- [3]. Most of them adapt to the image and attempt to reconstruct the high frequency content in a manner that preserves the edges and introduces as few artifacts as possible.…”

Threshold dynamics for high order geometric motions

“…Explicit, semi-implicit, and implicit surface algorithms have been introduced to track the evolution of the curves and surfaces (see for example, [1,8,9,4,12,7,6]). Due to the high order derivatives involved in the motions, stability conditions impose a severe efficiency drawback for explicit discretization methods.…”