EM, MCMC, and Chain Flipping for Structure from Motion with Unknown Correspondence.
Frank DellaertSteven M. SeitzCharles E. ThorpeSebastian ThrunPublished in: Mach. Learn. (2003)
Keyphrases
- structure from motion
- epipolar constraint
- feature correspondences
- bundle adjustment
- generative model
- camera parameters
- image correspondences
- computer vision
- image sequences
- point cloud
- expectation maximization
- feature tracking
- camera motion
- rank constraint
- markov chain
- missing data
- reprojection error
- scene structure
- monte carlo
- markov chain monte carlo
- epipolar lines
- camera positions
- em algorithm
- reconstruction from multiple images
- camera pose
- maximum likelihood
- posterior distribution
- parameter estimation
- image based localization
- machine learning
- importance sampling
- rigid motion
- approximate inference
- focal length
- probabilistic model
- three dimensional