True Multi-Image Alignment and Its Application to Mosaicing and Lens Distortion Correction
IEEE Transactions on Pattern Analysis and Machine Intelligence
Spherical Mosaics with Quaternions and Dense Correlation
International Journal of Computer Vision
Automated Mosaics via Topology Inference
IEEE Computer Graphics and Applications
Hi-index | 0.00 |
Given a video sequence acquired by an uncalibratedcamera rotating about a fixed center of projection, it is desiredto estimate the appearance of the scene in all possibledirections, recover the camera orientations, and determinethe constant focal length and radial lens distortion ofthe camera, without recourse to physical scene measurements.While parts of this problem have been studied before,there has not yet been a comprehensive algorithmicsolution that accommodates a variety of trajectories nor acharacterization of the prerequisites for successful reconstruction.This paper studies a computationally efficient algorithmthat takes maximum advantage of all informationavailable locally among pairs or small groups of framesas well as global closed cycle constraints. It works withany rotational trajectory, not just panning about a fixedaxis. Analyzing the algorithm's error surface predicts thesensitivity of estimation to lens distortion and global constraints.Both the analysis and experiments with naturalimages demonstrate that it is crucial to use either lens distortioncompensation or closed cycle constraints to get accurateparameter estimates and well-aligned mosaic images.