Ear recognition based on 2D images

Research of ear recognition and its application is a new subject in the field of biometrics authentication. Ear normalization and alignment is a fundamental module in the ear recognition system. Traditional manually normalization method is not suitable for an automatic recognition system. In this paper, an automatic ear normalization method is proposed based on improved Active Shape Model (ASM). This algorithm is applied on the USTB ear database for ear normalization. Then Full-space Linear Discriminant Analysis (FSLDA) is applied for ear recognition on the normalized ear images with different rotation variations. Experiments are performed on USTB ear image database. Recognition rates show that based on the right ear images, the acceptable head rotation range for ear recognition is between the right rotation of 20 degree to the left rotation of 10 degree.