A GENERIC GROUPING ALGORITHM AND ITS QUANTITATIVE-ANALYSIS

This paper presents a generic method for perceptual grouping and an an alysis of its expected grouping quality. The grouping method is fairly general: It may be used for the grouping of various types of data fea tures, and to incorporate different grouping cues operating over featu re sets of different sizes. The proposed method is divided into two pa rts: constructing a graph representation of the available perceptual g rouping evidence, and then finding the ''best'' partition of the graph into groups. The first stage includes a cue enhancement procedure, wh ich integrates the information available from multifeature cues into v ery reliable bifeature cues. Both stages are implemented using known s tatistical tools such as Wald's SPRT algorithm and the Maximum Likelih ood criterion. The accompanying theoretical analysis of this grouping criterion quantifies intuitive expectations and predicts that the expe cted grouping quality increases with cue reliability. It also shows th at investing more computational effort in the grouping algorithm leads to better grouping results. This analysis, which quantifies the group ing power of the Maximum Likelihood criterion, is independent of the g rouping domain. To our best knowledge, such an analysis of a grouping process is given here for the first time. Three grouping algorithms, i n three different domains, are synthesized as instances of the generic method. They demonstrate the applicability and generality of this gro uping method.