Face Tracking Using Motion-Guided Dynamic Template Matching

摘要: Combining two sophisticated techniques of motion detection and template matching, this paper proposes a simple but effective algorithm for tracking human faces. First, we use statistical model skin color shape information to detect face in the first frame, initialize it as an appearance-based intensity subsequent tracking. Second, incorporating background subtraction, projection histograms moving silhouette geometric constraints body parts, can quickly determine good approximation search region corresponding head location. Finally, correlation-based matching procedure is applied further localize accurately, current be dynamically updated size content adapt temporal changes tracked face’s scale orientation. Moreover, confidence measure representing template’s reliability presented guide possible re-initialization continuous Experimental results demonstrate validity our proposed method.