Superresolution with compound Markov random fields via the variational EM algorithm.
Atsunori KanemuraShin-ichi MaedaShin IshiiPublished in: Neural Networks (2009)
Keyphrases
- em algorithm
- markov random field
- super resolution
- image segmentation
- expectation maximization
- low resolution
- maximum a posteriori
- image restoration
- high resolution
- image super resolution
- parameter estimation
- mixture model
- belief propagation
- maximum likelihood
- graph cuts
- image reconstruction
- map estimation
- low resolution images
- random fields
- generative model
- mrf model
- higher order
- maximum likelihood estimation
- hyperparameters
- gaussian mixture model
- high quality
- energy function
- free energy
- potential functions
- conditional random fields
- motion estimation
- image processing
- pairwise
- probability density function
- energy minimization
- prior model
- hidden variables
- mrf models
- image patches
- unsupervised learning
- parameter learning
- message passing
- depth map
- optical flow
- multiscale
- computer vision
- bayesian framework
- level set
- feature vectors