Automatic ear recognition based on contour curve and local feature

A two-step matching method based on an improved Hausdorff distance for human ear recognition is proposed. An edge tracking method is performed first to extract the contour curve fo outer ear in a side face. and then a contour curve alignment method based on the improved Hausdorff distance is adopted for image matching so that a few well-matched ear images are picked out, the local feature extraction method is used for the final refined match. The recognition method can be performed automatically by use of contour curve and local key-points as feature vector, and the improved Hausdorff distance as a matching measurement. The method can efficiently overcome the influence of illumination change and partial occlusion on the ear recognition. Our simulation experiments on two ear image databases demonstrate that the proposed method is simple, efficient, robust and practical.