PhIT-Net: Photo-consistent Image Transform for Robust Illumination Invariant Matching
Abstract
We propose a new and completely data-driven approach for generating a photo-consistent image transform. We show that simple classical algorithms which operate in the transform domain become extremely resilient to illumination changes. This considerably improves matching accuracy, outperforming the use of state-of-the-art invariant representations as well as new matching methods based on deep features. The transform is obtained by training a neural network with a specialized triplet loss, designed to emphasize actual scene changes while attenuating illumination changes. The transform yields an illumination invariant representation, structured as an image map, which is highly flexible and can be easily used for various tasks.
Keywords
Cite
@article{arxiv.1911.12641,
title = {PhIT-Net: Photo-consistent Image Transform for Robust Illumination Invariant Matching},
author = {Damian Kaliroff and Guy Gilboa},
journal= {arXiv preprint arXiv:1911.12641},
year = {2021}
}
Comments
Paper accepted for publication at BMVC 2021. This version has the same content as in the published version, including the supplementary material