Automatic face feature tracking

A reliable face tracking can be obtained by integrating of the region detection, feature locating and tracking. To detect the face feature region, we used the mean shift method that is a clustering technique for multivariate data, and to track face landmark features, we also used the STK tracking method. STK tracking is very efficient for feature tracking, but the Newton-Raphson iteration scheme has the initial coordinate problem for tracking features. To solve the problem, we proposed a new BMA-NR method of the STK algorithm for face landmark features. Preliminary results indicate that this method solves the local minimum problem which occurs by NR iteration