This work studies the basic constraint relating changes in image irradiance and relative motion of an image-sensing system with respect to the environment. The solution for the motion, as well as the shapes of the objects in the environment, given time-varying images, is based on a calculus of variation formulation. It is expected that the discrete versions of the resulting Euler equations can be solved using iterative methods applied to a grid. Algorithms to recover motion and shape are tested, and a design for real-time hardware implementation is developed. The planned approach uses information from whole image regions instead of special feature points. This approach also avoids the computation of optical flow, an intermediate product used in one previous partial solution of this problem. An important goal is the development of algorithms which can be executed in a cooperative fashion on a grid, utilizing recently developed architectures for parrallel computing.
