Proceedings. 1991 IEEE Computer Society Conference on Computer Vision and Pattern Recognition
DOI: 10.1109/cvpr.1991.139669
|View full text |Cite
|
Sign up to set email alerts
|

Establishing motion correspondence

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
56
0

Publication Types

Select...
5
2
1

Relationship

0
8

Authors

Journals

citations
Cited by 67 publications
(56 citation statements)
references
References 10 publications
0
56
0
Order By: Relevance
“…It uses the principle of prediction to select candidates for an association and the minimal distance criteria to choose the best one. The best associations were obtained by a greedy approach similar to the method proposed by Rangarajan and Shah (1991). An estimate for the particle position in the second image was obtained by adding the displacement vector of the preceding calculation step to each particle position in the first image.…”
Section: Initial Algorithmmentioning
confidence: 99%
“…It uses the principle of prediction to select candidates for an association and the minimal distance criteria to choose the best one. The best associations were obtained by a greedy approach similar to the method proposed by Rangarajan and Shah (1991). An estimate for the particle position in the second image was obtained by adding the displacement vector of the preceding calculation step to each particle position in the first image.…”
Section: Initial Algorithmmentioning
confidence: 99%
“…Measurements of rotation can be achieved through motion correspondence of the set of feature points on the object. The correspondence is achieved using proximal uniforrnity constraint [9]. The accuracy of the overall procedure is improved using a least squares error approach on an over-determined system of equations.…”
Section: Related Experiments: Particle Propertiesmentioning
confidence: 99%
“…[9] address this problem, but they assume availability of 3D data for points. Rangarajan and Shah [32] use their proximal uniformity constraint to predict the position of missing points. However, their method requires that all points be available in the first two frames.…”
Section: Missing Tokensmentioning
confidence: 99%
“…The question naturally arises as to how accurate does this estimate have to be. In [32], the authors use information from frames k and k -1, as well as the smoothness constrains to generate the missing point in frame k + 1. Our simulation results clearly show that even a model as simple and naive as uniformly distributed random points serves the purpose of establishing correspondence.…”
Section: 2mentioning
confidence: 99%
See 1 more Smart Citation