Showing papers by "Gilles Simon published in 2018"
08 Sep 2018
TL;DR: It is shown that, in images of man-made environments, the horizon line can usually be hypothesized based on a-contrario detections of second-order grouping events, which allows constraining the extraction of the horizontal vanishing points on that line, thus reducing false detections.
Abstract: We show that, in images of man-made environments, the horizon line can usually be hypothesized based on a-contrario detections of second-order grouping events This allows constraining the extraction of the horizontal vanishing points on that line, thus reducing false detections Experiments made on three datasets show that our method, not only achieves state-of-the-art performance wrt horizon line detection on two datasets, but also yields much less spurious vanishing points than the previous top-ranked methods
30 citations
07 Oct 2018
TL;DR: A novel method for two-view epipolar and planar geometry estimation that first aims at detecting and matching physical vertical planes frequently present in these environments, before estimating corresponding homographies andferred local correspondences are finally used to improve fundamental matrix estimation.
Abstract: Given two views of the same scene, usual correspondence geometry estimation techniques exploit the well-established effectiveness of keypoint descriptors. However, such features have a hard time in poorly textured man-made environments, possibly containing repetitive patterns and/or specularities, such as industrial places. In that paper, we propose a novel method for two-view epipolar and planar geometry estimation that first aims at detecting and matching physical vertical planes frequently present in these environments, before estimating corresponding homographies. Inferred local correspondences are finally used to improve fundamental matrix estimation. The gain in precision is demonstrated on industrial and urban environments.
2 citations
25 Jun 2018
TL;DR: In this paper, a nouvelle methode for l'estimation des geometries epipolaire et planaire entre deux vues, which vise tout d'abord a detecter et mettre en correspondance les plans verticaux souvent presents dans ces environnements, avant d'estimer les homographies correspondantes.
Abstract: Etant donnees deux vues d'une meme scene, les techniques classiques de mise en correspondance geometrique s'appuient generalement sur l'efficacite maintes fois demontree des descripteurs de points d'interet. Cependant, de tels indices sont mis en difficulte dans des environnements peu textures, contenant possiblement des structures repetees et/ou des specularites, comme les environnements industriels. Dans ce papier, nous proposons une nouvelle methode pour l'estimation des geometries epipolaire et planaire entre deux vues, qui vise tout d'abord a detecter et mettre en correspondance les plans verticaux souvent presents dans ces environnements, avant d'estimer les homographies correspondantes. Les correspondances locales ainsi generees sont ensuite utilisees pour ameliorer l'estimation de la matrice fondamentale. Le gain en precision a ete demontre sur des images d'environnements industriel et urbain.
1 citations