Title :
Image-based pose estimation using a compact 3D model
Author :
Heisterklaus, Iris ; Ningqing Qian ; Miller, Artur
Author_Institution :
Inst. fur Nachrichtentechnik, RWTH Aachen Univ., Aachen, Germany
Abstract :
Image-based localization finds the pose an image was taken in. Our method uses the MPEG Compact Descriptors for Visual Search framework [1] to build a compact 3D model from the 3D reconstruction of a city. Then using an image retrieval the closest match for a query image is found and the pose is estimated. Our method is able to determine an estimate of the pose within a few meters of the actual one.
Keywords :
image reconstruction; image retrieval; pose estimation; solid modelling; video coding; 3D city reconstruction; MPEG compact descriptors; compact 3D model; image retrieval; image-based localization; image-based pose estimation; visual search framework; Cameras; Computational modeling; Estimation; Image retrieval; Solid modeling; Three-dimensional displays; Visualization;
Conference_Titel :
Consumer Electronics ??? Berlin (ICCE-Berlin), 2014 IEEE Fourth International Conference on
Conference_Location :
Berlin
DOI :
10.1109/ICCE-Berlin.2014.7034307
