inpainting_mcfinpainting_mcfinpainting_mcfInpaintingMcfInpaintingMcf (Operator)

Name

inpainting_mcfinpainting_mcfinpainting_mcfInpaintingMcfInpaintingMcf — Perform an inpainting by smoothing of level lines.

Signature

Description

The operator inpainting_mcfinpainting_mcfinpainting_mcfInpaintingMcfInpaintingMcf extends the image edges that adjoin the region RegionRegionRegionRegionregion of the input image ImageImageImageImageimage into the interior of RegionRegionRegionRegionregion and connects their ends by smoothing the level lines of the gray value function of ImageImageImageImageimage.

This happens through the application of the mean curvature flow or intrinsic heat equation

  u_t = div(grad u/|grad u|) |grad u| = curv(u) |grad u|

on the gray value function u defined in the region RegionRegionRegionRegionregion by the input image ImageImageImageImageimage at a time t_0 = 0. The discretized equation is solved in IterationsIterationsIterationsIterationsiterations time steps of length ThetaThetaThetaThetatheta, so that the output image InpaintedImageInpaintedImageInpaintedImageInpaintedImageinpaintedImage contains the gray value function at the time IterationsIterationsIterationsIterationsiterations * ThetaThetaThetaThetatheta .

A stationary state of the mean curvature flow equation, which is also the basis of the operator mean_curvature_flowmean_curvature_flowmean_curvature_flowMeanCurvatureFlowMeanCurvatureFlow, has the special property that the level lines of u all have the curvature 0. This means that after sufficiently many iterations there are only straight edges left inside the computation area of the output image InpaintedImageInpaintedImageInpaintedImageInpaintedImageinpaintedImage. By this, the structure of objects inside of RegionRegionRegionRegionregion can be simplified, while the remaining edges are continuously connected to those of the surrounding image matrix. This allows for a removal of image errors and unwanted objects in the input image, a so called image inpainting, which is only weakly visible to a human beholder since there remain no obvious artefacts or smudges.

To detect the image direction more robustly, in particular on noisy input data, an additional isotropic smoothing step can precede the computation of the gray value gradients. The parameter SigmaSigmaSigmaSigmasigma determines the magnitude of the smoothing by means of the standard deviation of a corresponding Gaussian convolution kernel, as used in the operator isotropic_diffusionisotropic_diffusionisotropic_diffusionIsotropicDiffusionIsotropicDiffusion for isotropic image smoothing.

Parallelization

• Multithreading type: reentrant (runs in parallel with non-exclusive operators).
• Multithreading scope: global (may be called from any thread).
• Automatically parallelized on tuple level.

Parameters

Alternatives

harmonic_interpolationharmonic_interpolationharmonic_interpolationHarmonicInterpolationHarmonicInterpolation, inpainting_ctinpainting_ctinpainting_ctInpaintingCtInpaintingCt, inpainting_anisoinpainting_anisoinpainting_anisoInpaintingAnisoInpaintingAniso, inpainting_cedinpainting_cedinpainting_cedInpaintingCedInpaintingCed, inpainting_textureinpainting_textureinpainting_textureInpaintingTextureInpaintingTexture

References

M. G. Crandall, P. Lions; “Convergent Difference Schemes for Nonlinear Parabolic Equations and Mean Curvature Motion”; Numer. Math. 75 pp. 17-41; 1996.
G. Aubert, P. Kornprobst; “Mathematical Problems in Image Processing”; Applied Mathematical Sciences 147; Springer, New York; 2002.

Module

Foundation