Phase-based 3D optical flow sensors for motion detection

Abstract

We demonstrate a computationally efficient technique for sensing optical flow using angle sensitive pixels. Angle-sensitive pixels function as spatial Gabor filters as a result of their periodic response to angle. Measuring temporal derivatives in the phase response of quadrature filters recovers information about object motion. These sensors work without a lens for global motion, or, with a lens to create a 3D map of object motion under ambient light.