Face contour detection and tracking with complex backgrounds

A novel method combining a motion detection technique and geometric active contours for face contour detection and tracking with complex backgrounds is presented in this paper. First, Wronskian change detector is used to segment moving regions from complex background. Second, via projection histograms of moving silhouette and geometric constraints of facial parts, possible face region can be determined very quickly. Third, a modified geometric active contour model incorporating prior face shape into the Chan-Vese model is applied to further detect face contour accurately. And finally, via first-order linear motion estimation, the detected face contour is tracked correctly in the following frames. Experimental results show the efficiency of our method.