10-12 topics：
1.Tips of the iceberg: introduction to the statistics of natural images
(1/f power laws, marginal statistics, high-kurtosis, scale invariance, scaling laws, manifolds, image universes )
2.From image statistics to image models
(Maximum entropy, Gibbs fields, Markov random fields)
3.Learning and estimation with Markov random fields
(MLE=ME, Ising/Potts, various likelihoods)
4.Minimax entropy learning and model pursuit
(texture, shape, face examples)
5.Conceptualization and ensemble equivalence
(types, typical sets, etc)
6.Generative modeling
(PCA, TCA, textons, and sparse coding, connecting to Dr. Jackie Shen’s lectures on wavelets)
7.Manifold learning: implicit vs explicit manifolds
(data ensemble, Kolmogorov episilon-entropy, capacity dimension)
8.Primal sketch modeling
(integrating the descriptive and generative models)
9.From primal sketch to 2.1D sketch and 2.5D sketches
(layered representation, shape from stereo, and shading)
10.Context free and context sensitive grammars and language
11. The integration of grammars and Markov random fields
(graphlets and composite templates)
12.A big unification in visual modeling and remaining questions |