Article Preview
Top1. Introduction
Nowadays, an individual is recognized based on his/her biometric characteristics in many places like offices, institute, airports, etc. These biometric characteristics are divided into two types: physical and behavioral (Jain and Kumar, 2012; Jain and Nandakumar, 2012). The examples of physical characteristics of an individual are fingerprint, face, iris, palm print and ear. The examples of behavioral characteristics of an individual are speech, signature, and gait. The systems based on biometrics recognize an individual automatically using various computer vision and pattern recognition algorithms (Jain and Kumar, 2012; Jain and Nandakumar, 2012). The biometric systems are used mainly for two operations: verification and authentication. In these operations, the query biometric image is compared and matched with its closest image in the database. While the query image is compared with all the database images in case of authentication, it is compared with its enrolled image in case of verification.
There are many recognition schemes described by researchers in last two decades. Jain and his research team suggested recognition schemes for physical characteristics such as ear, gait, etc. (Jain and Kumar, 2012; Abaza et al., 2013). Usage of the ear image for recognition of an individual has many advantages: (a) Ear shape does not change during human life. (b) The ear is easy to be acquired from distance (Abaza et al., 2013). On the other hand, usage of the ear images as biometric is challenging due to the levels of illumination during image acquisition, and also due to the occlusions in the acquired images which includes hair at head portion, hair on the ear, neck portion, headscarf worn by Muslim women, turban worn by Panjabi man etc.
Researchers have provided various ear databases, and presented different approaches for ear image detection, and recognition. The authors (Abaza et al., 2013; Pflug and Busch, 2012) gave a description on different ear databases such as West Virginia University (WVU), University of Science and Technology Beijing (USTB), University of California Riverside (UCR), University of Notre Dame (UND), XM2VTS, UMIST, NIST Mugshot Identification, and FERET. The authors listed out different recognition schemes based on intensity, force field, 2D curve geometry, Fourier descriptor, wavelet transformation, Gabor filters, Scale-Invariant Feature Transform (SIFT) and 3D features. The details of various ear databases are given below: