Comparison of graph cuts with belief propagation for stereo, using identical MRF parameters

By Marshall F. Tappen, William T. Freeman. Proceedings Ninth IEEE International Conference on Computer Vision, 2003

528 citations recorded by the source corpus.

Subfields: Biophysics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/5648b62c444556ea86d615037167e648bb771451

How researchers see this paper

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

In Biophysics

In Computer Vision and Pattern Recognition

Citations in the corpus

MRF Energy Minimization and Beyond via Dual Decomposition (2011, unknown): “…each node p 2 V such that the following objective function, known as the MRF energy, is minimized: X p2V p ðl p Þ þ X pq2E pq ðl p ; l q Þ: ð1Þ Currently, two classes of methods are the most prominent ones in MRF optimization [46], [47]: those based on graph-cuts and those based on message-passing.”
Quality Assessment of Stereoscopic Images (2008, result): “The comparative study presented in [44] shows that performances of the two methods are close to each other and superior to those of other algorithms proposed in literature.”
Multi-target tracking by learning local-to-global trajectory models (2015, methodology): “During the LBP inference process, we update one node at a time and sweep all the nodes by a “left-right fashion”[47].”
Continuous Markov Random Fields for Robust Stereo Estimation (2012, background): “Most modern approaches frame the problem as inference on a Markov random field (MRF) and utilize global optimization techniques such as graph cuts or message passing [3] to reason jointly about the depth of each pixel in the image.”
Learning Conditional Random Fields for Stereo (2007, background): “Tappen and Freeman [12] compare GC and BP for stereo; Szeliski et al. [13] compare a larger set of MRF minimization techniques and provide software that we use in our implementation.”
Consistent Depth Maps Recovery from a Video Sequence (2009, methodology): “To accomplish these goals, this paper contributes a global optimization scheme, which we call bundle optimization, to resolve most of the aforemen- tioned difficulties in disparity estimation.”
Depth Estimation Using Monocular and Stereo Cues (2007, background): “, 2004; Kumar and Hebert, 2003] and smoothing disparity to compute depthmaps in stereo vision [Tappen and Freeman, 2003].”
Convergent Tree-Reweighted Message Passing for Energy Minimization (2006, background): “In the context of ordinary BP it was observed experimentally that sequential updates are superior to parallel updates [28] although convergence is still not guaranteed.”
A Comparative Study of Energy Minimization Methods for Markov Random Fields with Smoothness-Based Priors (2008, background): “Among the few exceptions are [9], which compared the efﬁciency of several different max-ﬂow algorithms for graph cuts, and [45], which compared graph cuts with LBP for stereo.”
Efficient Belief Propagation for Early Vision (2004, methodology): “Overall our method performs comparably to the original graph cuts energy minimiza tion approach [4, 9] that similarly used simple data and discontinuity costs, as well as to resul ts in [11] that compared belief propagation with graph cuts.”
Optical techniques for 3D surface reconstruction in computer-assisted laparoscopic surgery (2013, background): “, 2007) or over the entire image using belief propagation (Tappen and Freeman, 2003), graph cuts (Kolmogorov et al.”
Stereo Matching with Color-Weighted Correlation, Hierachical Belief Propagation and Occlusion Handling (2006, methodology): “The proposed matching technique can be applied for many 3D reconstruction applications 21 .”
Efficient Belief Propagation with Learned Higher-Order Markov Random Fields (2006, background): “In general finding exact solutions to these estimation problems is hard for loopy graphs, but approximation approaches based on graph cuts [1, 3, 17, 18] and loopy belief propagation [6, 18, 19] have been found to often work well in practice.”
Linear Programming Relaxations and Belief Propagation - An Empirical Study (2006, methodology): “The task of finding the most probable assignment (or MAP) in a graphical model c mes up in a wide range of applications including image understanding (Tappen and Freeman, 2003), error correcting codes (Feldman et al., 2003) and protein folding (Yanover and Weiss, 2002).”
A Comparative Study of Energy Minimization Methods for Markov Random Fields (2006, methodology): “Among the few exceptions are [14], which compared the efficiency of several different max flow algorithms for graph cuts, and [13], which compared graph cuts with LBP.”
Exact and Approximate Sampling by Systematic Stochastic Search (2009, methodology): “Finally, we applied our method to a larger-scale application: approximate joint sampling from a Markov Random Field model for stereo vision, using parameters from (Tappen and Freeman, 2003).”
Describing Clothing by Semantic Attributes (2012, methodology): “In our implementation, we use belief propagation [32] to minimize the attribute label cost.”
Ray Markov Random Fields for image-based 3D modeling: Model and efficient inference (2010, methodology): “For MRF, two approximate inference methods, graph cuts [10] and loopy belief propagation (BP) [9] [14] [18], are usually deployed for their good inference quality and speed.”