Variational Simultaneous Stereo Matching and Defogging in Low Visibility
Yining Ding (Heriot-Watt University),* Andrew Wallace (Heriot Watt University), Sen Wang (Imperial College London)
The 33rd British Machine Vision Conference
Abstract
Given a stereo pair of daytime foggy images, we seek to estimate a dense disparity map and to restore a fog-free image simultaneously. Such tasks remain extremely challenging in low visibility, partially preventing modern autonomous vehicles from operating safely. In this paper, we propose a novel simultaneous stereo matching and defogging algorithm based on variational continuous optimisation. It effectively fuses depth cues from disparity and scattering to achieve accurate depth estimation as the first step. Then the depth information is used to help restore a defogged image by leveraging a photo-inconsistency check. Extensive experiments on both synthetic and real data show the proposed algorithm outperforms comparative methods in all metrics on depth estimation, and produces visually more appealing defogged images.Video
Citation
@inproceedings{Ding_2022_BMVC,
author = {Yining Ding and Andrew Wallace and Sen Wang},
title = {Variational Simultaneous Stereo Matching and Defogging in Low Visibility},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0394.pdf}
}
Copyright © 2022
The British Machine Vision Conference is organised by