Motion estimation from optical flow using camera mimicking blowfly visual processing

Optical flow, the apparent motion of spatially varying intensity profiles produced by a moving observer, allows estimation of self-motion and is used by many organisms for locomotion through their natural environment. However, the details of how this is accomplished using visual signals derived from dynamic natural scenery is unknown. In this work, we computationally investigate this problem using natural stimuli recorded using a specially designed "FlEye" camera that mimics the blowfly (Calliphora vicina) visual system and is equipped with joint gyroscope and accelerometer to capture motion trajectories. We use the convective derivative of the filtered image intensities to obtain one- and two-dimensional estimators of wide-field motion. Comparing yaw rate estimators from these two approaches, we demonstrate that the two-dimensional estimation provides a better determination and make connection with motion estimation in the blowfly visual system.