An estimation model of figure segregation based on human visual perception

When observing an overlapped figure, a human tends to interpret the figure by decomposing it into simpler subfigures. In psychology, this attitude is called figure segregation. In general, an overlapped figure can be segregated in many ways, depending on the observer. In the engineering realization of the figure segregation, the frequency of selection for each segregation candidate must be estimated. Up to now, such factors as symmetry and continuity have been considered qualitatively as the factors affecting the decision about the segregation of the figure. This paper presents first a method to describe quantitatively those factors. Then the significance of the decision factor is determined by comparing the characteristic values of the decision factor in the segregation candidate and the selection frequency of that candidate in the psychological experiment. Furthermore, it is shown that the figure segregation characteristics can be estimated accurately, based on the prediction expression derived from the linear multiple regression analysis.