Markov Random Field Modeling in Image Analysis

By S. Li. Computer Science Workbench, 2001

2,000 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/b9701ad65e256bd8841c4f80ced09b4ca1d5e331

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

Discriminative Random Fields (2006, methodology): “Other discontinuity models have also been proposed for adaptive smoothing (Li, 2001), but all of them require the labels to be either continuous or ordered.”
Image change detection algorithms: a systematic survey (2005, background): “We stop at this point, though the discussion could range into many broad, widely studied computer vision topics such as segmentation [110], [111], tracking [35], and layered motion estimation [112], [113].”
Texture classification: are filter banks necessary? (2003, methodology): “tential functions and then use the Hammersley-Clifford theorem to calculate the joint probability p(I) [12].”
Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising (2016, neutral): “random field (MRF) models [10], [11], [12].”
Eyeblink-based Anti-Spoofing in Face Recognition from a Generic Webcamera (2007, methodology): “Using the Hammersley and Clifford theorem [ 20 ], the joint distribution over the label sequence Y given the observation S can be written as the following form:”
Using Combination of Statistical Models and Multilevel Structural Information for Detecting Urban Areas From a Single Gray-Level Image (2007, methodology): “μλ, full covariance matrices ∑ λ, and the mixing parameters 3λ were estimated through standard ML approximated by the expectation maximization [10].”
Minimizing Energies with Hierarchical Costs (2012, background): “Energies with these terms are quite standard in vision, particularly in models based on Markov random fields (MRFs) and when performing maximum a posteriori (MAP-MRF) inference (Li 1994).”
A Statistical Approach to Material Classification Using Image Patch Exemplars (2009, background): “Both these results show that image patches cope as well with scale changes as the MR8 ﬁlter bank, and that features do not have to be extracted across a large range of scales for successful classiﬁcation.”
Creating Large-Scale City Models from 3D-Point Clouds: A Robust Approach with Hybrid Representation (2012, neutral): “Simulated annealing techniques (Li, 2001), graph-cut based algorithms e.g. (Boykov et al., 2001) or belief propagation methods e.g. (Weiss and Freeman, 2001) could provide a good approximation of the solution but at the expense of high computing time.”
Modeling and Detection of Geospatial Objects Using Texture Motifs (2006, background): “The MRF framework only provides local properties, namely a pixel’s conditional probability [41].”
Systematic image-driven analysis of the spatial Drosophila embryonic expression landscape (2010, methodology): “We developed a Markov Random Field (MRF)-based program (Geman and Geman, 1984; Li, 2001), TImrf, that minimizes the global energy score by iteratively labeling each triangle based on its staining intensities in relation to its neighboring triangles.”
Learning structured densities via infinite dimensional exponential families (2015, background): “Undirected graphical models, or Markov random fields [13], have been extensively studied and applied in fields ranging from computational biology [15, 28], to natural language processing [16, 20] and computer vision [9, 17].”
Multiscale conditional random fields for image labeling (2004, methodology): “MRFs are typically formulated in a probabilistic generative framework modeling the joint probability of the image and its corresponding labels [6, 12 ].”
Segmentation of X‐ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures (2009, methodology): “These methods are usually insensitive to local image noise and produce reasonable results, but may require a priori knowledge for construction of an appropriate image model, a supervised learning process or manual initialization [e.g., Berthod et al., 1996; Szirányi et al., 2000; Li, 2001].”
Deformable medical image registration: setting the state of the art with discrete methods. (2011, background): “Continuous (34) and discrete optimization (35, 36) methods are both often considered for the third aspect of parameter estimation where a trade-of between precision versus computational load is to be found.”
A Review of Recent Advances in Surface Defect Detection using Texture analysis Techniques (2008, methodology): “4 Model based approaches Model based methods include, among many others, fractal mod els [83], autoregressive models [85, 27], random field models [68], the epitome model [53], and the texem model [149].”