Optical flow estimation based on spatiotemporal plane fitting of event point clouds

Event data is the asynchronous acquisition of the increase or decrease in brightness of each pixel. Events are mainly generated near the edges along the motion of the subject and camera, and the event point clouds form a plane in the spatiotemporal 3D domain. Therefore, the surface formed by the event point clouds in the spatiotemporal domain can be regarded as the trajectory of motion (optical flow). In this paper, we propose a method for optical flow estimation using parameters of planes estimated from event point clouds in a segmented spatiotemporal subregion.