Image inpainting using geostatistical Kriging

Datum konání: 02.03.2018
Přednášející: Jiří Dvořák
Odpovědná osoba: Kotera

Image inpainting is a classical task of restoring missing or damaged parts of an image. Many automatic methods have been proposed using a variety of techniques. We discuss the possibility to use a statistical interpolation technique called "kriging". In this context we assume that sub-blocks of the image constitute a realization of a random field with constant mean, estimate its covariance structure and use it for prediction of the missing values (interpolation). A mask determining the missing/damaged pixels must be available. This approach produces extremely acceptable results in a range of different situations. The talk is based on the paper Jassim (2013), Image Inpainting by Kriging Interpolation Technique.