Stereo reconstruction serves many outdoor applications, and thus sometimes faces difficulties with foggy weather. However, fog provides depth cues for far away objects. By taking advantages of both stereo and fog cues, stereo reconstruction in fog can be improved. We propose a Markov Random Field model for this problem. The proposed model is tested on synthetic images and it shows that improved results can be achieved on both stereo reconstruction and visibility restoration.

This presentation is part of Minisymposium “MS45 - Mathematical techniques for bad visibility restoration”
organized by: Javier Vazquez-Corral (Information and Communication Technologies Department, Universitat Pompeu Fabra) .

Authors:

Jean-Philippe Tarel (Researcher. COSYS/LEPSiS , IFSTTAR)

Laurent Caraffa (IGN)

Keywords:

bayesian methods, computer vision, image enhancement, image reconstruction, inverse problems