Skip to Main Content
The paper proposes a novel method for chin contour estimation. First, we estimate several possible chin locations. A modified Canny edge detector is then used to extract edges from the face image. Using these estimated chin edges and chin locations, we obtain a number of curves as candidates for the chin contour. A confidence score that measures the likeliness of a curve being the actual chin contour is used to select the most likely candidate. If the confidence score of the most likely candidate is high enough, then the estimated chin contour may be used as a feature for face recognition. Experimental results show that the proposed algorithm can extract the chin contours of human faces with good accuracy and the reliable estimated chin contours can improve the face recognition rate.
Date of Conference: 6-10 April 2003