*Faugeras, O.D.*, and
*Berthod, M.*,

**Improving Consistency and Reducing Ambiguity in Stochastic Labeling:
An Optimization Approach**,

*PAMI(3)*, No. 4, July 1981, pp. 412-424.
This paper describes the essence of the gradient optimization
approach to relaxation labeling. A global criterion is defined that combines
the concepts of ambiguity and consistency. The relaxation procedure is used to
minimize the criterion, by moving in the direction of the strongest gradient.
This basic technique was used by other researchers in matching problems as in
Bhanu (

See also Shape Matching of Two-Dimensional Objects. ) and
Faugeras-Price (

See also Semantic Description of Aerial Images Using Stochastic Labeling. ).
BibRef
**8107**

*Faugeras, O.D.*, and
*Berthod, M.*,

**Scene Labeling: An Optimization Approach**,

*PR(12)*, No. 5, 1980, pp. 339-347.

Elsevier DOI
BibRef
**8000**

Earlier:
*PRIP79*(318-326).
*Relaxation, Theory*.
BibRef

*Faugeras, O.D.[Olivier D.]*,

**Decomposition and Decentralization Techniques in Relaxation Labeling**,

*CGIP(16)*, No. 4, August 1981, pp. 341-355.

Elsevier DOI
BibRef
**8108**

*Faugeras, O.D.[Olivier D.]*,

**An Optimization Approach for Using Contextual Information in
Computer Vision**,

*AAAI-80*(56-60).
BibRef
**8000**

*Faugeras, O.D.*, and
*Berthod, M.*,

**Using Context in the Global Recognition of a Set of Objects:
An Optimization Approach**,

presented at the 8th World Computer Congress,
*IFIP*1980.
More of the basic optimization approach.
BibRef
**8000**

*Yu, S.*,
*Berthod, M.*,

**A Game Strategy Approach for Image Labeling**,

*CVIU(61)*, No. 1, January 1995, pp. 32-37.

DOI Link
BibRef
**9501**

*Berthod, M.*,

**Semi-Consistency: A Solution to the No-Label Problem**,

*CVPR83*(555-556).. Similar:
BibRef
**8300**

**Global Optimization of a Consistent Labeling**,

*IJCAI83*(1065-1067).
BibRef

*Faugeras, O.D.*,

**Relaxation Labeling and Evidence Gathering**,

*PRIP82*(672-677).
BibRef
**8200**

And:
*ICPR82*(405-412).
The problem is how to represent non-support (negative support) and
ignorance. Normalization of the Q vectors is not needed.
Interesting-new numbers and computation-try it out to see what it
really means! The second version says about the same as the first,
but there is more of it.
BibRef

*Berthod, M.*,
*Long, P.*,

**Graph Matching by Parallel Optimization Methods:
An Application to Stereo Vision**,

*ICPR84*(841-843).
BibRef
**8400**

Chapter on Matching and Recognition Using Volumes, High Level Vision Techniques, Invariants continues in

Discrete Relaxation Methods .

Last update:Aug 28, 2024 at 16:02:19