Deep Optical Flow Estimation Via Multi-Scale Correspondence Structure Learning.
Shanshan ZhaoXi LiOmar El Farouk BourahlaPublished in: CoRR (2017)
Keyphrases
- optical flow estimation
- structure learning
- multiscale
- graphical models
- bayesian networks
- optical flow
- coarse to fine
- conditional independence
- feature matching
- parameter estimation
- transfer learning
- motion field
- flow field
- motion estimation
- multi frame
- point correspondences
- probabilistic model
- stereo matching
- markov networks
- image matching
- network structure
- image processing
- optic flow
- object recognition
- learning algorithm
- belief propagation
- random variables
- dynamic programming
- high quality
- image sequences
- maximum likelihood
- velocity field
- computer vision