Paper Info

Title | ||
---|---|---|

Modeling Natural Images Using Gated MRFs |

Abstract | ||
---|---|---|

This paper describes a Markov Random Field for real-valued image modeling that has two sets of latent variables. One set is used to gate the interactions between all pairs of pixels, while the second set determines the mean intensities of each pixel. This is a powerful model with a conditional distribution over the input that is Gaussian, with both mean and covariance determined by the configuration of latent variables, which is unlike previous models that were restricted to using Gaussians with either a fixed mean or a diagonal covariance matrix. Thanks to the increased flexibility, this gated MRF can generate more realistic samples after training on an unconstrained distribution of high-resolution natural images. Furthermore, the latent variables of the model can be inferred efficiently and can be used as very effective descriptors in recognition tasks. Both generation and discrimination drastically improve as layers of binary latent variables are added to the model, yielding a hierarchical model called a Deep Belief Network. |

Year | DOI | Venue |
---|---|---|

2013 | 10.1109/TPAMI.2013.29 | Pattern Analysis and Machine Intelligence, IEEE Transactions |

Keywords | Field | DocType |

Gaussian processes,Markov processes,belief networks,image processing,Gaussian process,Markov random field,covariance,deep belief network,gated MRF,hierarchical model,latent variable,mean,natural image modeling,pixel intensity,real-valued image modeling,Boltzmann machine,Gated MRF,deep learning,denoising,density estimation,energy-based model,facial expression recognition,factored 3-way model,generative model,natural images,object recognition,unsupervised learning | Computer vision,Pattern recognition,Markov random field,Computer science,Deep belief network,Latent class model,Latent variable,Probabilistic latent semantic analysis,Artificial intelligence,Gaussian process,Covariance matrix,Covariance | Journal |

Volume | Issue | ISSN |

35 | 9 | 0162-8828 |

Citations | PageRank | References |

25 | 1.65 | 45 |

Authors | ||

4 |

Authors (4 rows)

Cited by (25 rows)

References (45 rows)

Name | Order | Citations | PageRank |
---|---|---|---|

Marc'Aurelio Ranzato | 1 | 5242 | 470.94 |

Volodymyr Mnih | 2 | 25 | 1.65 |

Joshua Susskind | 3 | 194 | 9.68 |

geoffrey e hinton | 4 | 40435 | 4751.69 |