Local edge-preserving image inpainting with mixed weighted anisotropic regularization methods

RUBEN SPIES; FRANCISCO IBARROLA

Tandil

Congreso; V Congreso de Matemática Aplicada, Computacional e Industrial; 2015

The image inpainting problem consists of restoring an image from which data from one or more regionsis missing. Several inpainting models to perform this task have been developed, depending their performances on thecharacteristics of the image. For instance, if the image to be restored is smooth, the inpainting can be performed withgood results by means of classical Tikhonov-Phillips methods. For non-smooth images, however, such an approachmay not work quite well, as it precludes edge preservation inside the occluded regions. Whenever this is a priority, theCurvature-Driven Diffusion (CDD) method results effective, although it is unsuitable for inpainting smooth regions,as it tends to produce piecewise-constant restorations. In this work we make use of a CDD inpainting method to geta-priori information for introducing anisotropy into a mixed weighted regularization method, which is finally used forinpainting. Results and comparison tools will be shown.