A HIERARCHICAL OPTICAL-FLOW ESTIMATION ALGORITHM-BASED ON THE INTERLEVEL MOTION SMOOTHNESS CONSTRAINT

An optical flow estimation algorithm is proposed based on the multiresolution representation of the image, i.e. the image pyramid. Since the image pyramid consists of several low-pass filtered versions of the image sequence, the motion vector at each pyramid level is also the low-pass filtered versions of the motion vector at the finest level. Thus, the interlevel motion smoothness constraint is introduced in the motion estimation by defining the operators to establish an appropriate relationship between two adjacent pyramid levels. The employment of the image pyramid allows us to estimate the large flow vectors and to adopt easily the multigrid technique to improve the convergence speed of the proposed algorithm. The simulation results reveal that the proposed algorithm provides more accurate motion estimation, compared to the methods using the smoothness constraints such as Horn and Schunck's algorithm. In addition, the theoretical analysis of the convergence behavior of the proposed algorithm is also provided in this paper.