Why a Gradient Model Can See Second-Order Motion Though an Energy Model Can’t

Alan Johnston

Dept. of Psychology, University College London

Abstract:
Second-order motion sequences are interesting because the motion we see cannot be accounted for by motion energy models. Second-order motion can be made visible to motion energy analysis by introducing a nonlinear operation (such as rectification) prior to motion energy computation. In essence the rectification stage “throws away” the sign of the luminance contrast signal. We have shown that a gradient model can recover the direction of motion of second-order pattern. A spatio-temporal gradient model combines the signs of the spatial and temporal derivatives of image luminance in such a way that the sign of the product indicates direction of motion independent of the sign of the luminance contrast. It is this property that appears to important for recovery of second-order motion. The model is tested by comparing model predictions against psychophysical data on the percieved speed and direction of induced carrier motion, which is generated by drifting a contrast modulation over a static binary noise carrier. Induced carrier motion is reduced when a first-order luminance variation is added to the modulation. This observation is predicted by a the gradient model but is difficult to account for on the basis of the standard nonlinear rectification model. Ref: Johnston, A, Benton, C.P. and McOwan, P.W. (1999) Induced motion at texture-defined motion boundaries. Proc. R. Soc. Lond., B., 266, 2441-2450.

The lecture will take place:

in the Lecture Hall, Room 203, 44 Cummington St.
on Tuesday, August 22, 2000
at 11:00 am