Disocclusion: a variational approach using level lines

Abstract

Object recognition, robot vision, image and film restoration may require the ability to perform disocclusion. We call disocclusion the recovery of occluded areas in a digital image by interpolation from their vicinity. It is shown in this paper how disocclusion can be performed by means of the level-lines structure, which offers a reliable, complete and contrast-invariant representation of images. Level-lines based disocclusion yields a solution that may have strong discontinuities. The proposed method is compatible with Kanizsa's amodal completion theory.

Threshold dynamics for high order geometric motions

“…For instance, the active contours (or "snakes") model [18] of Kass, Witkin, and Terzopoulos originally calls for minimizing an energy that includes the square of the curvature integrated along the curve. Another important example is the segmentation with depth involving disocclusion [20,21] model of Nitzberg, Mumford, and Shiota [24,10], where curvature dependent functionals are to be minimized in order to extract information about the three-dimensional arrangement of objects making up a scene from a single two-dimensional image of the scene. Yet another related problem of the field that leads to higher order geometric flows is the image inpainting problem of Bertalmio et al, where the goal is to repair damaged regions in an image by connecting level lines of intensity using smooth curves [3].…”

Curvature Regularization for Curves and Surfaces in a Global Optimization Framework

“…[2,3,8]. Experimental results are mainly given for segmentation, but other applications include surface completion [9] and inpainting [3,10].…”

Curvature Regularity for Multi-label Problems - Standard and Customized Linear Programming

“…In principle there are infinitely many levels to consider, but one commonly [15] deals only with those levels that actually occur on the image boundary, i.e. a subset of L = {0, ..., 255}.…”

“…This has turned out to be much more challenging. Here, in a region-based context only the inpainting approach of Masnou and Morel [15] finds global optima.…”