New ear contour extraction method based on edge feature

Ear detection is an important step in ear recognition and the exact localization for ear directly affects the final recognition rate. According to the feature of ear and its position on the side face, a method based on edge-tracking was adopted for human ear localization and extraction from side face images. At first, binary image was generated by using an adaptive thresholding segmentation method, and difference image was obtained by using a difference model. Then outer ear was divided into three regions depending on different orientation tendency. The binary image and the difference image were combined with the original image to find the contour of the three regions respectively and the information of ear's position in side face. Experimental results show that our method is simple, its localization accuracy is high, and the extracted ear contour edge is clear, connected and non-disturbed.