Depth map from the combination of matched points with active contours

This paper describes the analysis of an active contour fitted to a target in a sequence of images recorded by a freely moving uncalibrated camera. The motivating application is the visual guidance of a robot towards a target. Contour deformations are analysed to extract the scaled depth of the target, and to explore the feasibility of 3D egomotion recovery. The scaled depth is used to compute the time to contact, which provides a measure of distance to the target, and also to improve the common depth maps obtained from point matches, which are a valuable input for the robot to avoid obstacles