Many recent energy-based methods for optical flow estimation rely on a good initialization that is typically provided by some kind of feature matching. So far, however, these initial matching approaches are rather general: They do not incorporate any additional information that could help to improve the accuracy or the robustness of the estimation. In particular, they do not exploit potential cues on the camera poses and the thereby induced rigid motion of the scene. In the present paper, we tackle this problem. To this end, we propose a novel structure-from-motion-aware PatchMatch approach that, in contrast to existing matching techniques, combines two hierarchical feature matching methods: a recent two-frame PatchMatch approach for optical flow estimation (general motion) and a specifically tailored three-frame PatchMatch approach for rigid scene reconstruction (SfM). While the motion PatchMatch serves as baseline with good accuracy, the SfM counterpart takes over at occlusions and other regions with insufficient information. Experiments with our novel SfM-aware PatchMatch approach demonstrate its usefulness. They not only show excellent results for all major benchmarks (KITTI 2012/2015, MPI Sintel), but also improvements up to 50% compared to a PatchMatch approach without structure information.
%0 Book Section
%1 conf/eccv/MaurerMGB18
%A Maurer, Daniel
%A Marniok, Nico
%A Goldluecke, Bastian
%A Bruhn, Andrés
%B Computer Vision – ECCV 2018. ECCV 2018. Lecture Notes in Computer Science
%D 2018
%E Ferrari, Vittorio
%E Hebert, Martial
%E Sminchisescu, Cristian
%E Weiss, Yair
%I Springer International Publishing
%K 2018 B04 B05 from:leonkokkoliadis sfbtrr161 visus visus:bruhnas visus:maurerdl
%P 575-592
%R 10.1007/978-3-030-01237-3_35
%T Structure-from-motion-aware PatchMatch for Adaptive Optical Flow Estimation
%U https://doi.org/10.1007/978-3-030-01237-3_35
%V 11212
%X Many recent energy-based methods for optical flow estimation rely on a good initialization that is typically provided by some kind of feature matching. So far, however, these initial matching approaches are rather general: They do not incorporate any additional information that could help to improve the accuracy or the robustness of the estimation. In particular, they do not exploit potential cues on the camera poses and the thereby induced rigid motion of the scene. In the present paper, we tackle this problem. To this end, we propose a novel structure-from-motion-aware PatchMatch approach that, in contrast to existing matching techniques, combines two hierarchical feature matching methods: a recent two-frame PatchMatch approach for optical flow estimation (general motion) and a specifically tailored three-frame PatchMatch approach for rigid scene reconstruction (SfM). While the motion PatchMatch serves as baseline with good accuracy, the SfM counterpart takes over at occlusions and other regions with insufficient information. Experiments with our novel SfM-aware PatchMatch approach demonstrate its usefulness. They not only show excellent results for all major benchmarks (KITTI 2012/2015, MPI Sintel), but also improvements up to 50% compared to a PatchMatch approach without structure information.
%@ 978-3-030-01237-3
@inbook{conf/eccv/MaurerMGB18,
abstract = {Many recent energy-based methods for optical flow estimation rely on a good initialization that is typically provided by some kind of feature matching. So far, however, these initial matching approaches are rather general: They do not incorporate any additional information that could help to improve the accuracy or the robustness of the estimation. In particular, they do not exploit potential cues on the camera poses and the thereby induced rigid motion of the scene. In the present paper, we tackle this problem. To this end, we propose a novel structure-from-motion-aware PatchMatch approach that, in contrast to existing matching techniques, combines two hierarchical feature matching methods: a recent two-frame PatchMatch approach for optical flow estimation (general motion) and a specifically tailored three-frame PatchMatch approach for rigid scene reconstruction (SfM). While the motion PatchMatch serves as baseline with good accuracy, the SfM counterpart takes over at occlusions and other regions with insufficient information. Experiments with our novel SfM-aware PatchMatch approach demonstrate its usefulness. They not only show excellent results for all major benchmarks (KITTI 2012/2015, MPI Sintel), but also improvements up to 50% compared to a PatchMatch approach without structure information.},
added-at = {2020-03-05T11:47:01.000+0100},
author = {Maurer, Daniel and Marniok, Nico and Goldluecke, Bastian and Bruhn, Andrés},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/23e4d40ecf46e9f262b9f18ec176ff2b0/sfbtrr161},
booktitle = {Computer Vision – ECCV 2018. ECCV 2018. Lecture Notes in Computer Science},
doi = {10.1007/978-3-030-01237-3_35},
editor = {Ferrari, Vittorio and Hebert, Martial and Sminchisescu, Cristian and Weiss, Yair},
ee = {https://doi.org/10.1007/978-3-030-01237-3_35},
interhash = {939cd39d1657ff8b9e8aae4d030ede61},
intrahash = {3e4d40ecf46e9f262b9f18ec176ff2b0},
isbn = {978-3-030-01237-3},
keywords = {2018 B04 B05 from:leonkokkoliadis sfbtrr161 visus visus:bruhnas visus:maurerdl},
pages = {575-592},
publisher = {Springer International Publishing},
timestamp = {2020-10-05T11:38:55.000+0200},
title = {Structure-from-motion-aware PatchMatch for Adaptive Optical Flow Estimation},
url = {https://doi.org/10.1007/978-3-030-01237-3_35 },
volume = 11212,
year = 2018
}