Abstract
3D surface structure is not fully specified by the Optic Flow (OF) produced by the relative motion between an observer and a planar surface. Current theories postulate that the visual system interprets the OF by combining retinal measurements of image velocities with extra-retinal information about the observer's egomotion. By introducing some assumptions in the interpretation process, veridical estimates of Euclidean 3D structure can, in principle, be derived from the OF. However, recent empirical findings hinder the biological plausibility of such approach by showing that perceived 3D structure is not veridical and that egomotion signals are seldom used in the perceptual interpretation of the OF. Here, we present a new model for perceived 3D structure from the OF that (1) disregards the linear motion of the observer’s Point of View (PoV), (2) relies on the assumption that the OF is produced by a generic linear motion of the PoV, (3) derives the most likely 3D structure by integrating over the missing motion parameters. A tenet of our approach, which we term the Generic Linear Ego Motion (GLEM) assumption, is that the translational components of the OF are diagnostic of egomotion. This model makes counterintuitive predictions that cannot be reconciled with current theories postulating a veridical analysis of the OF. We tested the model's predictions by asking observers to passively or actively view the local OF produced by a rotating planar surface. Results from these experiments, which will be illustrated at the present conference (Mancuso et al., VSS 2012; Fantoni, et al., VSS 2012), confirm the predictions of the model. Our findings provide further empirical evidence that the perceptual interpretation of the OF discards extraretinal signals of linear egomotion.
Meeting abstract presented at VSS 2012