Purchase this article with an account.
Karl Muller, Jonathan Matthis, Kathryn Bonnen, Lawrence Cormack, Mary Hayhoe; Visual motion statistics during real-world locomotion. Journal of Vision 2018;18(10):1059. doi: https://doi.org/10.1167/18.10.1059.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The middle temporal visual area (MT/V5) has been implicated in the representation of visual motion in primates. MT neurons encode orientation and speed of visual motion, but the distribution of speed and orientation preferences of MT neurons is not uniform. In this study, we explored whether this non-uniform distribution reflects the statistics of the natural input derived from self-motion. We measured motion patterns generated in the context of locomotion in natural outdoor environments, using a Pupil Labs mobile eye tracker to record both binocular eye position and high-resolution scene video. We approximate a cyclopean retinal input by aligning each frame from the scene camera to the point of fixation, converting the scene data to a retinal reference frame. We used the Farneback optic flow estimation algorithm to measure speed and orientation of visual motion at each pixel location across the visual field at each video frame. The distribution of motion directions is strongly biased towards downward motion in the lower visual field, and the distribution of velocities peaks at about 10 deg/s. We then calculated the motion distributions at different locations across the visual field within regions matching the spatial characteristics of MT receptive fields. This analysis showed that distribution of motion in individual receptive fields is biased as a function of both eccentricity and direction (relative to the fovea). Specifically, the distribution of motion for more eccentric receptive fields has greater motion velocities, and the vector average of motion direction is pointed away from the fovea. This pattern is consistent with a pattern of continuous outward flow from the point of fixation. This analysis allows a comparison with distributions of speed and orientation preferences measured electrophysiologically to examine whether characteristics of MT cells are shaped by the natural image motion statistics.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only