A common problem in Bayesian object recognition using marked point pro
cess models is to produce a point estimate of the true underlying obje
ct configuration: the number of objects and the size, location and sha
pe of each object. We use decision theory and the concept of loss func
tions to design a more reasonable estimator for this purpose, rather t
han using the common zero-one loss corresponding to the maximum a post
eriori estimator. We propose to use the squared Delta-metric of Baddel
ey (1992) as our loss function and demonstrate that the corresponding
optimal Bayesian estimator can be well approximated by combining Marko
v chain Monte Carlo methods with simulated annealing into a two-step a
lgorithm. The proposed loss function is tested using a marked point pr
ocess model developed for locating cells in confocal microscopy images
.