Outline • Texture modeling - continued – FRAME model for textures Texture Modeling • The structures of images – The structures in images are due to the inter-pixel relationships – The key issue is how to characterize the relationships 5/29/2016 Visual Perception Modeling 2 FRAME Model • FRAME model – Filtering, random field, and maximum entropy – A well-defined mathematical model for textures by combining filtering and random field models 5/29/2016 Visual Perception Modeling 3 Maximum Entropy • Maximum entropy – Is an important principle in statistics for constructing a probability distribution on a set of random variables – Suppose the available information is the expectations of some known functions n(x), that is E p [ n ( x)] n ( x) p( x)dx n for n 1,, N – Let W be the set of all probability distributions p(x) which satisfy the constraints W { p( x) | E p [ n ( x)] n , n 1,, N } 5/29/2016 Visual Perception Modeling 4 Maximum Entropy – cont. • Maximum Entropy – continued – According to the maximum entropy principle, a good choice of the probability distribution is the one that has the maximum entropy p ( x) arg max p( x) log p( x)dx * subject to E p [ n ( x)] n ( x) p( x)dx n for n 1,, N p( x)dx 1 5/29/2016 Visual Perception Modeling 5 Maximum Entropy – cont. • Maximum Entropy – continued – By Lagrange multipliers, the solution for p(x) is 1 N p( x; ) exp n n ( x) Z () n 1 – where N (1 , 2 ,, N ) and Z( ) exp - n n ( x)dx n 1 5/29/2016 Visual Perception Modeling 6 Maximum Entropy – cont. • Maximum Entropy – continued – (1 , 2 ,, N ) are determined by the constraints – But a closed form solution is not available general • Numerical solutions dn E p ( x; ) [ n ( x)] n dt n 1, 2, , N 5/29/2016 Visual Perception Modeling 7 Maximum Entropy – cont. • Maximum Entropy – continued – The solutions are guaranteed to exist and be unique by the following properties 5/29/2016 Visual Perception Modeling 8 FRAME Model • Texture modeling – The features can be anything you want n(x) – Histograms of filter responses are a good feature for textures 5/29/2016 Visual Perception Modeling 9 FRAME Model – cont. • The FRAME algorithm – Initialization Input a texture image Iobs Select a group of K filters SK={F(1), F(2), ...., F(K)} Compute {Hobs(a), a = 1, ....., K} Initialize (ia ) 0, i 1,2,, L a 1,2,, K Initialize Isyn as a uniform white noise image 5/29/2016 Visual Perception Modeling 10 FRAME Model – cont. • The FRAME algorithm – continued – The algorithm Repeat calculate Hsyn(a), a=1,..., K from Isyn and use it as E p ( I ; K ,S K ) ( H (a ) ) Update (ia ) by d(ia ) / dt E p ( I ; ,S ) [ H (a ) ] H obs(a ) Apply Gibbs sampler to flip Isyn for w sweeps L obs(a ) syn (a ) 1 until | H H | for a 1, 2, , K i i 2 K K i 1 5/29/2016 Visual Perception Modeling 11 FRAME Model – cont. • The Gibbs sampler 5/29/2016 Visual Perception Modeling 12 FRAME Model – cont. • Filter selection – In practice, we want a small number of “good” filters – One way to do that is to choose filters that carry the most information • In other words, minimum entropy 5/29/2016 Visual Perception Modeling 13 FRAME Model – cont. • Filter selection algorithm – Initialization 5/29/2016 Visual Perception Modeling 14 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 15 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 16 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 17 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 18 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 19 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 20 FRAME Model – cont. 5/29/2016 Visual Perception Modeling 21