We introduce a generalized image osmosis model, enhancing the quality of the reconstructed image in the shadow removal problem by means of directional diffusion weights. Analogies with the second order linear and nonlinear diffusion inpainting equations (e.g. Harmonic, AMLE, Total Variation) acting on the shadow boundaries are further shown. Numerical details of efficient implementations for the model are provided, as well as applications to real images and to cultural heritage conservation challenges.

This presentation is part of Minisymposium “MS15 - Nonlinear Diffusion: Models, Extensions and Algorithms (2 parts)”
organized by: Ke Chen (University of Liverpool) , Joachim Weickert (Saarland University) , Xue-Cheng Tai (Hong Kong Baptist University) .

Authors:

Simone Parisotto (University of Cambridge)

Luca Calatroni (CMAP, École Polytechnique CNRS)

Marco Caliari (University of Verona)

Carola-Bibiane Schönlieb (University of Cambridge)

Keywords:

image enhancement, image reconstruction, image registration, inverse problems