Intermediate-Level Motion Representations Account for the Hollow Face Illusion S Tschechne, H Neumann |
|---|
Institute for Neural Information Processing, University of Ulm, Germany
|
| Three-dimensional surface structure can be deferred from motion fields and their gradients [Treue and Andersen, 1996, Visual Neuroscience]. In the hollow face illusion (HFI) an unresolved convex/concave ambiguity leads to the percept of a concave face mask being convex when viewed frontally. It has been argued [Heard and Chugg, 2003, Perception] that this demonstrates the use of top-down knowledge to override local feature interpretations. We suggest that local mechanisms of motion computation may already account for the illusory effect. We extend a biologically inspired model that incorporates early and intermediate stages of cortical motion processing to indicate rotations of rigid object around its axes [Raudies et al., 2013, NECO]. Network components sensitive to motion direction/speed, speed gradients and their nonlinear combination to motion curvature build a robust representation of the spatio-temporal input. The model is probed by input sequences with rotating facial masks. Simulated motion responses are integrated at the stage of nonlinear motion curvature cells [Orban, 2008, Physiology Review]. Motion curvature cells selectively respond to the apparent image motion gradient pattern, reflecting the HFI. The major effect of the HFI can thus be accounted for by intermediate motion representations signaling motion gradients and curvature patterns. Acknowledgement: SFB/TRR62, DFG. |
| Up Home |
|---|