A 2-D motion estimation algorithm

The problem of motion estimation is reviewed, and a region matching motion estimation algorithm is presented. This algorithm estimates accurately the motion parameters in the case of linear 2-D motion and gives a very good linear approximation in the case of a nonlinear 2-D motion. It assumes that the images have been previously segmented. Its main advantage in comparison to other motion estimation algorithms is its robustness in the presence of noise. It does not require small motion or smooth intensity profiles but only segmented images. Its computation time can be significantly reduced by an appropriate choice of the initial values of the motion parameters. The performance of the algorithm is examined for different kinds of motion and various signal-to-noise ratios using computer generated and real images