13.3.5 Graph Matching, Continuous Relaxation

Chapter Contents (Back)
Object Recognition. Constraint Satisfaction. Matching, Graphs. Matching, Relaxation. Relaxation, Continuous.

Barnard, S.T., and Thompson, W.B.,
Disparity Analysis of Images,
PAMI(2), No. 4, July 1980, pp. 333-340. BibRef 8007
Earlier: TR-79-1, CSD, Univ. of MinnesotaJanuary 1979. Relaxation, Results. Matching, Points. Matching for motion. This program finds corresponding pairs of points in a motion analysis system using the similarity of motion with neighboring points. Feature points (such as corners) in both views are used rather than the single view used in Moravec, and a relaxation procedure finds the final global match between the two sets of feature points. The initial assignments of possible matches for the set of feature points is simply all the features with a similar (nearby) position in the second image. Thus, small motions are assumed. An iterative (relaxation based) procedure uses the disparities of the nearby points to eliminate the unlikely assignments from the set of possible assignments. These include points with disparities different from the others in the neighborhood. The formulation of the algorithm is very simple and thus it works for any kind of disparity (such as from observer motion, multiple object motions, or stereo) and it does not require any detailed camera models. This provides a basic matching method to find disparity for a moderate number of points (the feature points) that are generally consistent with the other nearby points (i.e. smooth surfaces), but allowing for edges or changes in the disparity field. See also Lower-level Estimates and Interpretation of Visual Motion. BibRef

Barnard, S.T.,
The Image Correspondence Problem,
Ph.D.Thesis (CS), U Minn, 1979. The thesis version of his work. BibRef 7900

Kitchen, L.,
Relaxation Applied to Matching Quantitative Relational Structures,
SMC(10), February 1980, pp. 96-101. Fuzzy Logic. Introduction of a new operator defined in terms of fuzzy logic with some examples on synthetic structures. Experiments with the operator on more general problems indicate that there may be problems which are not indicated by the synthetic problems. BibRef 8002

Yamamoto, H.,
Some Experiments on LANDSAT Pixel Classification Using Relaxation Operators,
CGIP(13), No. 1, May 1980, pp. 31-45.
WWW Version. BibRef 8005

Kirby, R.L.,
A Product Rule Relaxation Method,
CGIP(13), No. 2, June 1980, pp. 158-189.
WWW Version. BibRef 8006

Hwang, J.J., and Hall, E.L.,
Matching of Featured Objects Using Relational Tables from Stereo Images,
CGIP(20), No. 1, September 1982, pp. 22-42.
WWW Version. Matching, Regions. Features include regions, lines and vertices. The example is a complex block-like UT. The structure is simply adjacencies. The arrays are used to simplify the search for the matching subset. They use precise knowledge of the camera locations to get search lines in the second image. BibRef 8209

Hwang, J.J., and Hall, E.L.,
Scene Representation Using the Adjacency Matrix and Sampled Shapes of Regions,
PRIP78(250-261). BibRef 7800

Faugeras, O.D., and Price, K.E.,
Semantic Description of Aerial Images Using Stochastic Labeling,
PAMI(3), No. 6, November 1981, pp. 633-642. BibRef 8111 USC Computer Vision BibRef
And: ICPR80(352-357). BibRef
And: DARPA80(89-94). Matching, Regions. Relaxation, Results. The use of an optimization based relaxation method with structural descriptions. This work uses a relaxation approach very similar to that of ( See also Improving Consistency and Reducing Ambiguity in Stochastic Labeling: An Optimization Approach. ) for finding corresponding regions in two images of the same scene and finding regions in the image corresponding to elements in a model of the scene. The relaxation matching procedure has two major steps: finding initial potential matches and computing the updated match rating based on the matches for the neighboring regions. These steps are combined by: (1) Compute the match rating for each region in the model with all regions in the image. Order these and keep only the best (15) matches. (2)Compute the compatibility for each of these possible matches with the current most likely match for all the neighboring (related in the network) regions. (3) Update the match ratings so that compatible matches improve and incompatible ones decrease. (4) If some match is very likely, make the assignment permanent, and continue with the initialization step. Otherwise continue with the compatibility computation step. This procedure works by finding the most obvious match (e.g. largest regions, and all other features match) and building around this one by making assignments to regions related to the obvious match. This matching system makes few assumptions about the types of scenes, though assumptions can be used to improve the efficiency of the match, and is applicable to a variety of tasks. See also Symbolic Image Registration and Change Detection. BibRef

Price, K.E.,
Hierarchial Matching Using Relaxation,
CVGIP(34), No. 1, April 1986, pp. 66-75.
WWW Version. BibRef 8604 USC Computer VisionDiscussion of the use of group level descriptions to aid relaxation. BibRef

Price, K.E.,
Relaxation Matching Techniques - A Comparison,
PAMI(7), No. 5, September 1985, pp. 617-623. BibRef 8509 USC Computer Vision BibRef
And: ICPR84(987-989). Relaxation, Evaluation. Comparison of several relaxation methods, for accuracy and time. BibRef

Price, K.E.,
Symbolic Matching of Images and Scene Models,
DARPA82(299-308). BibRef 8200 USC Computer Vision BibRef
And: CVWS82(105-112). Several discussions on relaxation techniques in one paper. The See also Relaxation Matching Techniques - A Comparison. and See also Hierarchial Matching Using Relaxation. supersede this one. BibRef

Price, K.E.,
Relaxation Matching Applied to Aerial Images,
DARPA81(22-25). BibRef 8100 USC Computer VisionDiscussion of more recent results. Not much else. BibRef

Price, K.E.,
Symbolic Matching and Analysis with Substantial Changes in Orientation,
DARPA78(93-99). BibRef 7800 USC Computer Vision BibRef
And: PRAI-78(19-21). BibRef

Hummel, R.A.[Robert A.],
A Design Method for Relaxation Labeling Applications,
AAAI-83(168-171). BibRef 8300
Earlier: NYUCS Dept., TR 68, March 1983. A discussion of how to set up a relaxation labeling system. BibRef

Pelkowitz, L.,
A Continuous Relaxation Labeling Algorithm for Markov Random Fields,
SMC(20), 1990, pp. 709-715. BibRef 9000

Li, S.Z.,
Matching: Invariant to Translations, Rotations, and Scale Changes,
PR(25), No. 6, June 1992, pp. 583-594.
WWW Version. BibRef 9206

Ogawa, H.,
A Fuzzy Relaxation Technique For Partial Shape-Matching,
PRL(15), No. 4, April 1994, pp. 349-355. BibRef 9404

Qin, C., Luh, J.Y.S.,
Ambiguity Reduction by Relaxation Labeling,
PR(27), No. 1, January 1994, pp. 165-180.
WWW Version. BibRef 9401

Ranganath, H.S., and Chipman, L.J.,
Fuzzy Relaxation Approach for Inexact Scene Matching,
IVC(10), No. 9, November 1992, pp. 631-640.
WWW Version. Matching, Regions. BibRef 9211

Cooper, P.R.[Paul R.], Swain, M.J.[Michael J.],
Arc Consistency: Parallelism and Domain Dependence,
AI(58), No. 1-3, 1992, pp. 207-23.5
WWW Version. BibRef 9200

Cooper, P.R.[Paul R.], Swain, M.J.[Michael J.],
Domain Dependence in Parallel Constraint Satisfaction,
IJCAI89(54-59). BibRef 8900

Swain, M.J.[Michael J.], Cooper, P.R.[Paul R.],
Parallel Hardware for Constraint Satisfaction,
AAAI-88(682-686). BibRef 8800

Gold, S.[Steven], Rangarajan, A.[Anand],
A Graduated Assignment Algorithm for Graph Matching,
PAMI(18), No. 4, April 1996, pp. 377-388.
IEEE Abstract. IEEE Top Reference.
WWW Version. BibRef 9604 YaleDCS/RR-1062, January 1995. BibRef
And:
Graph Matching by Graduated Assignment,
CVPR96(239-244).
IEEE Abstract. IEEE Top Reference.
WWW Version. Matching O(lm). Similar to relaxation (annealing) approach. (But not quite). Uses hand labeled features in the image for matching (multiple features on an object). They note that relaxation labeling does poorly on pure subgraph isomorphism (no attributed nodes), and does poorly when noise is high for attributed graph matching. (Though the comparison is with the most basic relaxation methodology.) 9605 BibRef

Gold, S.[Steven],
Matching and Learning Structural and Spatial Representation with Neural Networks,
Ph.D.Thesis, Yale, 1995. BibRef 9500

Gold, S.[Steven], Rangarajan, A.[Anand], and Mjolsness, E.,
Learning with Preknowledge: Clustering with Point and Graph Matching Distance Measures,
NeurComp(8), 1966, pp. 787-804. BibRef 6600

Sitaraman, R., Rosenfeld, A.,
Probabilistic Analysis of Two Stage Matching,
PR(22), No. 3, 1989, pp. 331-343.
WWW Version. BibRef 8900

Finch, A.M.[Andrew M.], Wilson, R.C., Hancock, E.R.[Edwin R.],
Matching Delaunay Graphs,
PR(30), No. 1, January 1997, pp. 123-140.
WWW Version. 9702 BibRef
Earlier: A1, A3 only: CIAP95(56-61).
Springer DOI Reference 9509 BibRef

Finch, A.M.[Andrew M.], Wilson, R.C.[Richard C.], Hancock, E.R.[Edwin R.],
Matching delaunay triangulations by probabilistic relaxation,
CAIP95(350-358).
Springer DOI Reference 9509 BibRef

Finch, A.M., Hancock, E.R.,
Matching Deformed Delaunay Triangulations,
SCV95(31-36).
IEEE Top Reference. Univ. of York. Relaxation applied to matching graphs composed of triangles. BibRef 9500

Bhattacharya, P.,
Some Remarks on Fuzzy Graphs,
PRL(6), 1987, pp. 297-302. BibRef 8700

Pelillo, M., Fanelli, A.M.,
Autoassociative Learning in Relaxation Labeling Networks,
PRL(18), No. 1, January 1997, pp. 3-12. 9704 BibRef
Earlier: ICPR96(IV: 105-110).
IEEE DOI Reference 9608(Univ. Ca Foscari Venezia, I) BibRef

Do, K.H., Kim, Y.S., Uam, T.U., Ha, Y.H.,
Iterative Relaxational Stereo Matching Based on Adaptive Support Between Disparities,
PR(31), No. 8, August 1998, pp. 1049-1059.
WWW Version. 9807 Stereo, Matching. BibRef

Skomorowski, M.[Marek],
Use of random graph parsing for scene labelling by probabilistic relaxation,
PRL(20), No. 8, August 1999, pp. 949-956. BibRef 9908

Torsello, A.[Andrea], Pelillo, M.[Marcello],
Continuous-time relaxation labeling processes,
PR(33), No. 11, November 2000, pp. 1897-1908.
WWW Version. 0011 BibRef

Medasani, S., Krishnapuram, R., Choi, Y.S.,
Graph Matching by Relaxation of Fuzzy Assignments,
Fuzzy(9), No. 1, 2001, pp. 173-182. BibRef 0100

Bengoetxea, E.[Endika], Larraņaga, P.[Pedro], Bloch, I.[Isabelle], Perchant, A.[Aymeric], Boeres, C.[Claudia],
Inexact graph matching by means of estimation of distribution algorithms,
PR(35), No. 12, December 2002, pp. 2867-2880.
WWW Version. 0209 BibRef
Earlier: A1, A2, A3, A4, Only:
Estimation of Distribution Algorithms: A New Evolutionary Computation Approach for Graph Matching Problems,
EMMCVPR02(454 ff.).
HTML Version. 0205 BibRef

Perchant, A., Bloch, I.,
Graph Fuzzy Homomorphism Interpreted as Fuzzy Association Graphs,
ICPR00(Vol II: 1042-1045).
IEEE DOI Reference
HTML Version. 0009 BibRef

Aldea, E.[Emanuel], Fouquier, G.[Geoffroy], Atif, J.[Jamal], Bloch, I.[Isabelle],
Kernel Fusion for Image Classification Using Fuzzy Structural Information,
ISVC07(II: 307-317).
Springer DOI Reference 0711 BibRef

Aldea, E.[Emanuel], Atif, J.[Jamal], Bloch, I.[Isabelle],
Image Classification Using Marginalized Kernels for Graphs,
GbRPR07(103-113).
Springer DOI Reference 0706 BibRef

Fouquier, G.[Geoffroy], Atif, J.[Jamal], Bloch, I.[Isabelle],
Local Reasoning in Fuzzy Attribute Graphs for Optimizing Sequential Segmentation,
GbRPR07(138-147).
Springer DOI Reference 0706 BibRef

Cesar, Jr., R.M.[Roberto M.], Bengoetxea, E.[Endika], Bloch, I.[Isabelle], Larraņaga, P.[Pedro],
Inexact graph matching for model-based recognition: Evaluation and comparison of optimization algorithms,
PR(38), No. 11, November 2005, pp. 2099-2113.
WWW Version. 0509 BibRef
Earlier: A1, A2, A3, Only:
Inexact graph matching using stochastic optimization techniques for facial feature recognition,
ICPR02(II: 465-468).
IEEE DOI Reference 0211 BibRef

Sminchisescu, C.[Cristian], Triggs, B.[Bill],
Building Roadmaps of Minima and Transitions in Visual Models,
IJCV(61), No. 1, January 2005, pp. 81-101.
WWW Version. 0410 BibRef
Earlier:
Building Roadmaps of Local Minima of Visual Models,
ECCV02(I: 566 ff.).
HTML Version. 0205Avoiding local minima in searching techniques. BibRef

Richards, J.A.[John A.], Jia, X.P.[Xiu-Ping],
A Dempster-Shafer Relaxation Approach to Context Classification,
GeoRS(45), No. 5, May 2007, pp. 1422-1431.
IEEE DOI Reference 0704 BibRef

Schellewald, C.[Christian], Roth, S.[Stefan], Schnorr, C.[Christoph],
Evaluation of a convex relaxation to a quadratic assignment matching approach for relational object views,
IVC(25), No. 8, 1 August 2007, pp. 1301-1314.
WWW Version. 0706Quadratic assignment; Weighted graph matching; Combinatorial optimization; Convex programming; Object recognition BibRef

Werner, T.[Tomas],
A Linear Programming Approach to Max-Sum Problem: A Review,
PAMI(29), No. 7, July 2007, pp. 1165-1179.
IEEE DOI Reference 0706 Constraint Satisfaction. Maximization of a sum of binary functions. Explore a formulation from early Russian paper. BibRef

Werner, T.[Tomas],
High-arity interactions, polyhedral relaxations, and cutting plane algorithm for soft constraint optimisation (MAP-MRF),
CVPR08(1-8).
IEEE DOI Reference 0806 BibRef

Werner, T.[Tomas],
Combinatorial constraints on multiple projections of a set of points,
ICCV03(1011-1016).
IEEE DOI Reference 0311 BibRef

Potetz, B.[Brian], Lee, T.S.[Tai Sing],
Efficient belief propagation for higher-order cliques using linear constraint nodes,
CVIU(112), No. 1, October 2008, pp. 39-54.
WWW Version. 0810 BibRef
Earlier: A1, Only:
Efficient Belief Propagation for Vision Using Linear Constraint Nodes,
CVPR07(1-8).
IEEE DOI Reference 0706Belief propagation; Higher-order cliques; Non-pairwise cliques; Factor graphs; Continuous Markov random fields BibRef


Kumar, M.P.[M. Pawan], Torr, P.H.S.,
Fast Memory-Efficient Generalized Belief Propagation,
ECCV06(IV: 451-463).
Springer DOI Reference 0608 BibRef

Coito, F.J.[Fernando J.], Lemos, J.M.[João M.],
Adaptive Optimization with Constraints: Convergence and Oscillatory Behaviour,
IbPRIA05(II:19).
Springer DOI Reference 0509 BibRef

Yuille, A.L.[Alan L.],
A Double-Loop Algorithm to Minimize the Bethe Free Energy,
EMMCVPR02(3 ff.).
HTML Version. 0205 BibRef
Earlier:
A Double-Loop Algorithm to Minimize the Bethe and Kikuchi Free Energies,
SCTV01(xx-yy). 0106 BibRef

Yedidia, J., Freeman, W.T., Weiss, Y.,
Bethe free energy, Kikuchi approximations, and belief propagation algorithms,
SCTV01(xx-yy). 0106Stable points of belief propagation algorithms for graphs with loops correspond to extrema of the Bethe free energy. BibRef

Haddon, J., Boyce, J.,
Spatio-Temporal Relaxation Labelling Applied to Segmented Infrared Image Sequences,
ICPR96(II: 171-175).
IEEE DOI Reference 9608(Defence Res. Agency, UK) BibRef

Horiuchi, T., Yamamoto, K., Yamada, H.,
Robust Relaxation Method for Structural Matching Under Uncertainty,
ICPR96(II: 176-180).
IEEE DOI Reference 9608(Univ. of Tsukuba, J) BibRef

Shao, Z., Kittler, J.V.,
Fuzzy Non-Iterative ARG Labeling with Multiple Interpretations,
ICPR96(II: 181-185).
IEEE DOI Reference 9608(Univ. of Surrey, UK) BibRef

Hatef, M., Kittler, J.V.,
Combining symbolic with numeric attributes in multi-class object recognition problems,
ICIP95(III: 364-367).
IEEE DOI Reference 9510 BibRef

Deruyver, A., Hode, Y.,
Semantic graph and arc consistency in 'true' three dimensional image labelling,
ICIP95(II: 619-622).
IEEE DOI Reference 9510 BibRef

Choate, J.A., Gennert, M.A.,
Multiscale relaxation labeling of fractal images,
CVPR93(674-675).
IEEE Abstract. IEEE Top Reference. 0403 BibRef

McLean, C.R., Dyer, C.R.,
An Analog Relaxation Processor,
ICPR80(58-60). BibRef 8000

Chapter on Matching and Recognition Using Volumes, High Level Vision Techniques, Invariants continues in
Continuous Relaxation Theory .


Last update:Jan 1, 2009 at 17:09:16