On the Statistical Analysis of Dirty Pictures

By Julian Besag. 1986

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Generating Sentences from a Continuous Space (2015, methodology): “Inference method To generate imputations from the two models, we use beam search with beam size 15 for the rnnlm and approximate iterated conditional modes (Besag, 1986) with 3 steps of a beam size 5 search for the vae.”
Barrier Frank-Wolfe for Marginal Inference (2015, methodology): “The models are close to submodular and the local relaxation is a good approximation to the marginal polytope.”
Dynamic Coattention Networks For Question Answering (2016, methodology): “start and end positions in a similar fashion to the Iterative Conditional Modes algorithm (Besag, 1986).”
Inverse Reinforcement Learning in Swarm Systems (2016, background): “Essentially, this is equivalent to a synchronous version of the iterated conditional modes algorithm [4] which, in the sequential case, is guaranteed to translate the system to a state of locally minimum energy.”
Discrete optimization of ray potentials for semantic 3D reconstruction (2015, methodology): “The remaining variables can be labelled using Iterated Conditional Modes (ICM) [1] iteratively per variable.”
Bayesian areal wombling via adjacency modeling (2007, methodology): “Geman and Geman 1984; Besag 1986 )u ses up tofour-way neighbor relations to address this problem.”
Bilinear Sparse Coding for Invariant Vision (2005, methodology): “ICM is a deterministic method shown to generally converge quickly, albeit to a local minimum.”
An Algorithm for Total Variation Minimization and Applications (2004, unknown): “In practice, rather than sampling, it is found that choosing the values that maximise the local conditional yields a useful result, and at a much lower computational cost. This is ICM [ 2 ].”
MAP estimation via agreement on trees: message-passing and linear programming (2005, background): “…taking expectations of these indicators with respect to some distribution p ( · ) yields marginal probabilities for each node s ∈ V and for each edge Note that equations (5) and (6) deﬁne a d -dimensional vector µ = { µ α , α ∈ I} of marginals, indexed by elements of I deﬁned in equation (4).”
A Comparative Study of Energy Minimization Methods for Markov Random Fields with Smoothness-Based Priors (2008, background): “One of the reasons this framework is so popular is that it can be justiﬁed in terms of maximum a-posteriori estimation of a Markov random ﬁeld (MRF) [5], [19].”
Medical image segmentation on GPUs - A comprehensive review (2015, methodology): “One method is iterative conditional modes (ICM), which was introduced by Besag (1986). ICM starts with an initial segmentation S, and optimizes the local energy of each pixel deterministically.”
ReSeg: A Recurrent Neural Network for Object Segmentation (2015, methodology): “…& Huttenlocher, 2004; Boykov & Funka-Lea, 2006) that represent the pixels of the image as nodes of the graph, greedy approaches (such as, e.g., Besag, 1986; Cai et al., 2007) and continuation methods (such as, e.g., Blake & Zisserman, 1987; Geman & Geman, 1984) are usually more popular.”
Tree consistency and bounds on the performance of the max-product algorithm and its generalizations (2004, methodology): “…max-product algorithm and its generalizations MARTIN WAINWRIGHT∗, TOMMI JAAKKOLA† and ALAN WILLSKY† ∗Electrical Engineering & Computer Science, UC Berkeley, Berkeley, CA wainwrig@eecs.berkeley.edu †Electrical Engineering & Computer Science, MIT, Cambridge, MA tommi@ai.mit.edu;…”
Magnetic resonance image tissue classification using a partial volume model. (2001, methodology): “We use the iterated conditional modes (ICM) algorithm developed by Besag (1986) to search for an optimal labeling.”
Slice-to-volume medical image registration: a survey (2017, methodology): “Simpler move-making algorithms not using max-flow, but still applying the strategy of starting with an initial labeling and iteratively moving to a better one until a convergence criterion is met, 23 are the classical Iterated conditional modes (ICM) (Besag, 1986) and its generalization Lazy Flipper”
Using Combination of Statistical Models and Multilevel Structural Information for Detecting Urban Areas From a Single Gray-Level Image (2007, background): “For computational tractability, the observed data is assumed to be conditionally independent [10], [11].”