COMPUTATION OF IMAGE VELOCITY FIELD USING LINEAR AND NONLINEAR OBJECTIVE FUNCTIONS.

Image velocity (optical flow) fields are computed and interpreted from an image sequence by incorporating knowledge of object kinematics. Linear and nonlinear objective functions are defined by considering the kinematics, and the function parameters are estimated simultaneously with the computation of the velocity field by relaxation. The objective functions provide an interpretation of the dynamic scenes, and at the same time serve as the smoothness constraints. The difficulty of the aperture problem is alleviated by introducing iterative adjustment of measurement directions. The approach lends itself to the analysis of 3-D motion of a solid object, using piecewise objective functions. An example using images of an automobile in motion generated velocity fields that are consistent with visual perception is presented.