An integration-based method for depth estimation
Daniel Raviv · 1991
A closed-form, integration-based, and massively-parallel algorithm for determining depth of points in 3-D using one moving camera is presented.It is based on analyzing a sequence of images that result from a known rectilinear motion of a camera (with no rotation) in a stationary environment.A traceable point in an image sequence is reconstructed using an integration operation (no differentiation operator is involved).The method arose from two simple observations:(1) Stationary points in the 3-D scene appear to move away from the focus of expansion (FOE).(2) The distance of a point in 3-D space from the camera motion-axis is the same at all instants of time.Any visible moving point in the image can be processed independently of, and concurrently with, any other point.Laboratory results for the case where the optical axis is parallel to the motion axis show an error of less than 0.6% in absolute distance.