The analysis of time-varying image sequences

The analysis of time-varying image sequences is a classical problem of machine vision, which is likely to be rather useful in several fields, such as robotics and passive navigation. In this paper a general approach to the problem is presented, which is based on the computation of optical flow obtained by a similar procedure to the one proposed by Girosi et al.. By exploiting mathematical properties of the 2D motion field several 3D motion parameters, such as time-to-collision and angular velocity, can be recovered with a precision which is dependent on the image texture.