From a Sequence of Images to a Depth Map (NOT PUBLISHED)

In a robotic environment objects in the workspace are often manipulated only by the robot. Thus a stable vision field is presented to a robot mounted camera. Given exact knowledge of the physical motion of the camera, depth information can be readily extracted from the camera images. Closed-form solutions for velocities and depths are obtained without using any iterative method, and the computational cost is proportional to the resolution of the images.