Texture Inpainting Using Efficient Gaussian Conditional SimulationCP7

For texture images, the inpainting problem can be formulated as a random field conditional simulation within the masked area given the unmasked pixels. We propose such an approach for stationary Gaussian textures and show that the traditional algorithm for Gaussian conditional simulation can be implemented efficiently using the Fourier representation of the covariance operator. The resulting algorithm is able to inpaint large holes of any shape in a texture. Joint work with Arthur Leclaire ; http://epubs.siam.org/doi/10.1137/16M1109047

This presentation is part of Contributed Presentation “CP7 - Contributed session 7

Bruno Galerne (Université Paris Descartes)
Arthur Leclaire (CMLA, ENS Cachan)
gaussian random fields, image reconstruction, inpainting, inverse problems, texture