Frame-to-frame image motion estimation using image features

Frame-to-frame image motion estimation is a problem of practical interest. Its uses include biomedical imaging, communications systems, and military targetting. In this paper a new methodology of motion estimation is presented which uses random features of an image as the basis for estimating frame-to-frame image motion. The purpose of using features is to avoid the difficulty with the non-invertibility of the image field data. It is this non-invertibility of image fields that makes the problem difficult and interesting to solve.<
>