Hyperacuity is the phenomenon whereby differences in stimuli smaller than than the size of individual photo-receptors in the can be reliably detected. This is similar to pixel sub-sampling used in computer vision applications. This work applies vision modeling to hyperacuity stimuli that go beyond simple stimuli to include more complexity.
Simple hyperacuity phenomena can be modeled with a small number of oriented filters. Such models break down when the stimuli become more complex, such as opposite polarity with respect to the background, or masked by sinusoidal gratings, or when the gap between stimulus elements is increased. How can these stimuli and combinations of these stimuli be modeled? We use a version of the Boundary Contour System as the Vision Model, and connect to a modified Hebbian Learning module.
Figure 1: A complex hyperacuity stimulus.