Abstract

In a recent model for the measurement of 2-D image motion, each 2-D sensor responds to edge motion and generates multiple outputs representing a subset of all possible edge velocities.¹ A second stage determines which output is the correct edge velocity by using a cosine weighted voting scheme. Voting is carried out over the whole field, and the winning vote determines the final output direction. In the majority of cases involving uniform motion over the whole image, this scheme detects the correct velocity of the pattern as a whole and solves the aperture problem. For certain patterns, however, the model output departs significantly from that predicted by the constraint line solution.^2,3 In particular, for moving plaids made up of two unequal contrast sine wave gratings of different orientations, the model predicts a bias in the perceived direction of motion. Data are presented that show that this Is consistent with human psychophysical results. We show that in the case of unequal contrast, the nodes in a plaid become skewed. Their asymmetric shape is responsible for the directional bias in the model output and may be involved in the perceptual bias as well. Our findings suggest a reassessment of the constraint line solution and the role of the nodes in the determination of the direction of motion of plaids.

© 1988 Optical Society of America