We extend Weickert's drift-diffusion osmosis model (2011) to the anisotropic and non-linear case by weighting appropriately the diffusion component in terms of suitable contraction and rotation matrices encoding local directional information in the image. The model is amenable for several imaging applications such as shadow removal, face fusion and image cloning and can effectively be applied for the analysis of large image data in cultural heritage.

This is poster number 6 in Poster Session

Authors:

Luca Calatroni (CMAP, École Polytechnique CNRS)

Simone Parisotto (University of Cambridge)

Keywords:

computer vision, image enhancement, image reconstruction, image representation, partial differential equation models