Movement Detection and Figure-Ground Discrimination

Abstract

Movement detectors of the so-called correlation type were proposed long ago to explain motion perception in insects (Hassenstein and Reichardt 1956; Reichardt 1957, 1961; Reichardt and Varju 1959; Varjú 1959). In the meantime, good evidence has been accumulated that this movement detection scheme can also be applied to motion detection in humans (e.g. van Doom et al. 1982a,b; van Santen et al. 1984, 1985; Wilson 1985; Baker and Braddick 1985). More recently our interest in movement computation has focused on dynamic aspects and on the dependence of the detector output on the structure of the stimulus pattern. In addition, the properties of two-dimensional arrays of pairs of movement detectors (Reichardt and Guo 1986; Egelhaaf and Reichardt 1987; Reichardt 1987) have been investigated in detail. Individual movement detectors, however, do not provide meaningful information on a moving pattern. In addition, some spatial, physiological integration is needed, for instance in connection with a solution of the figure and ground discrimination problem (Reichardt and Poggio 1979; Reichardt 1979, 1980; Poggio et al. 1981; Reichardt et al. 1983; Egelhaaf 1985).