In this paper we propose a technique to recursively estimate structure and motion from a sequence of images acquired by an hand held camera. The algorithm is based on estimation of instantaneous optical flow for each frame of the sequence and on the use of the differential epipolar constraint to compute motion and structure. An alternative procedure to manage scale ambiguity across frames is also proposed. Neither user interaction nor any kind of scene knowledge is required and the algorithm, given its low complexity, is adaptable to real time applications
Home Contents Author index Keyword index