Improving the speed of convergence of a maximum-likelihood motion estimation algorithm of a human face

For human face motion estimation the shape of a human face is considered to be rigid and described by a triangular mesh. The motion is described by six parameters: one three-dimensional translation vector and three rotation angles. The motion parameters are estimated by maximizing the conditional probability of the frame to frame intensity differences at observation points. The speed of convergence is improved by detecting outliers in the observation points and excluding them from motion estimation. For outlier detection an algorithm based on random sample consensus (RANSAC) has been developed. Experimental results reveal that the average processing time for motion estimation per frame is reduced by 67.49%.