M. Lindenbaum, AN INTEGRATED MODEL FOR EVALUATING THE AMOUNT OF DATA REQUIRED FOR RELIABLE RECOGNITION, IEEE transactions on pattern analysis and machine intelligence, 19(11), 1997, pp. 1251-1264
Many recognition procedures rely on the consistency of a subset of dat
a features with a hypothesis as the sufficient evidence to the presenc
e of the corresponding object. We analyze here the performance of such
procedures, using a probabilistic model, and provide expressions for
the sufficient size of such data subsets, that, if consistent, guarant
ee the validity of the hypotheses with arbitrary confidence. We focus
on 2D objects and the affine transformation class, and provide, for th
e first time, an integrated model which takes into account the shape o
f the objects involved, the accuracy of the data collected, the clutte
r present in the scene, the class of the transformations involved, the
accuracy of the localization, and the confidence we would like to hav
e in our hypotheses. Interestingly, it turns out that most of these fa
ctors can be quantified cumulatively by one parameter, denoted ''effec
tive similarity,'' which largely determines the sufficient subset size
. The analysis is based on representing the class of instances corresp
onding to a model object and a group of transformations, as members of
a metric space, and quantifying the variation of the instances by a m
etric cover.