Image correspondence from motion subspace constraint and epipolar constraint

Authors:
Shigeki Sugimoto;Hidekazu Takahashi;Masatoshi Okutomoi
Affiliations:
Department of Mechanical and Control Engineering, Graduate School of Sceience and Engineering, Tokyo Institute of Technology, Tokyo, Japan;Diesel Inj. Eng. Dept. 1, DENSO Corporation, Kariya-shi, Aichi, Japan;Department of Mechanical and Control Engineering, Graduate School of Sceience and Engineering, Tokyo Institute of Technology, Tokyo, Japan
Venue:
ACCV'07 Proceedings of the 8th Asian conference on Computer vision - Volume Part II
Year:
2007

Citing 9
Cited 0

The Design and Use of Steerable Filters

IEEE Transactions on Pattern Analysis and Machine Intelligence
Shape and motion from image streams under orthography: a factorization method

International Journal of Computer Vision
A Paraperspective Factorization Method for Shape and Motion Recovery

IEEE Transactions on Pattern Analysis and Machine Intelligence
Determining the Epipolar Geometry and its Uncertainty: A Review

International Journal of Computer Vision
A Multibody Factorization Method for Independently Moving Objects

International Journal of Computer Vision
Stereo-Motion with Stereo and Motion in Complement

IEEE Transactions on Pattern Analysis and Machine Intelligence
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography

Communications of the ACM
Distinctive Image Features from Scale-Invariant Keypoints

International Journal of Computer Vision
Generalized Principal Component Analysis (GPCA)

IEEE Transactions on Pattern Analysis and Machine Intelligence

Quantified Score

Hi-index	0.01

Visualization

Abstract

In this paper, we propose a novel method for inferring image correspondences on the pair of synchronized image sequences. In the proposed method, after tracking the feature points in each image sequence over several frames, we solve the image corresponding problem from two types of geometrical constraints: (1) the motion subspace obtained from the tracked feature points of a target sequence, and (2) the epipolar constraints between the two cameras. Dissimilarly to the conventional correspondence estimation based on image matching using pixel values, the proposed approach enables us to obtain the correspondences even though the feature points, that can be seen from one camera view, but can not be seen (occluded or outside of the view) from the other camera. The validity of our method is demonstrated through the experiments using synthetic and real images.