Abstract

Is perceived motion based on a similarity metric (subtending the classical motion detector whose spatially sampled inputs are processed by identical sensors), or is it based, as recently suggested,¹ on a covariance metric implying a unique, undifferentiated motion detector allowing for spatiotemporal matching between qualitatively distinct elements to override spatiotemporal matching between identical elements? We will display the stimulus used to answer this question as it relates to the combination of chromatic contrast (CC) and luminance contrast (LC) information. Its spatiotemporal structure is such that CC-defined elements (red or green on a yellow background) will carry rightward motion, while the potential matching between CC- and LC- (positive or negative luminance increments) defined elements will carry leftward motion. According to the covariance model, leftward motion should be perceived any time that CC×LC>CC×CC, i.e., when LC is larger than the equivalent luminance contrast of the CC-defined element (~5%). For LC ranging from 0 to ±80%, leftward motion was never perceived, clearly reinforcing (together with additional experiments to be reported) the original view of a similarity metric for motion perception.

© 1992 Optical Society of America