Circular Motion Geometry Using Minimal Data
IEEE Transactions on Pattern Analysis and Machine Intelligence
Outward-Looking Circular Motion Analysis of Large Image Sequences
IEEE Transactions on Pattern Analysis and Machine Intelligence
Motion Recovery for Uncalibrated Turntable Sequences Using Silhouettes and a Single Point
ACIVS '08 Proceedings of the 10th International Conference on Advanced Concepts for Intelligent Vision Systems
Camera calibration with two arbitrary coaxial circles
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part I
| Hi-index | 0.00 |
This paper describes a new and simple method of recovering thegeometry of uncalibrated circular motion or single axis motionusing a minimal data set of 2 points in 4 images. This problem hasbeen solved using non-minimal data either by computing thefundamental matrix and trifocal tensor in 3 images, or by fittingconics to tracked points in 5 images. Our new method first computesa planar homography from a minimum of 2 points in 4 images. It isshown that two eigenvectors of this homography are the images ofthe circular points. Then, other fixed image entities and rotationangles can be straightforwardly computed. The crux of the methodlies in relating this planar homography from two different pointsto a homology naturally induced by corresponding points ondifferent conic loci from a circular motion. The experiments onreal image sequences demonstrate the simplicity, accuracy androbustness of the new method.