L. Herault et R. Horaud, SMOOTH CURVE EXTRACTION BY MEAN-FIELD ANNEALING, Annals of mathematics and artificial intelligence, 13(3-4), 1995, pp. 281-300
In this paper, we attack the figure-ground discrimination problem from
a combinatorial optimization perspective. In general, the solutions p
roposed in the past solved this problem only partially: either the mat
hematical model encoding the figure-ground problem was too simple or t
he optimization methods that were used were not efficient enough or th
ey could not guarantee to find the global minimum of the cost function
describing the figure-ground model. The method that we devised and wh
ich is described in this paper is tailored around the following contri
butions. First, we suggest a mathematical model encoding the figure-gr
ound discrimination problem that makes explicit a definition of shape
(or figure) based on cocircularity, smoothness, proximity, and contras
t. This model consists of building a cost function on the basis of ima
ge element interactions. Moreover, this cost function fits the constra
ints of an interacting spin system, which in turn is a well suited phy
sical model to solve hard combinatorial optimization problems. Second,
we suggest a combinatorial optimization method for solving the figure
-ground problem, namely mean field annealing which combines the mean f
ield approximation and annealing. Mean field annealing may well be vie
wed as a deterministic approximation of stochastic methods such as sim
ulated annealing. We describe in detail the theoretical bases of this
method, derive a computational model, and provide a practical algorith
m. Finally, some experimental results are shown for both synthetic and
real images.