Biometric systems have achieved a great deal of success for identity recognition of individuals in

most of the civilian, law-enforcement, and forensic applications in recent years. The ever increasing

popularity of biometric systems offer reliable identity recognition than traditional possession

and knowledge based approaches, as biometric characteristics cannot be shared, forgotten, or lost.

Biometric recognition operation refer to verification or identification of individuals based on physiological

or behavioral characteristics such as fingerprint, palmprint, iris, face, ear, gait, signature,

voice etc. Over last two decades, the ear has been predominantly attracted many researchers as an

emerging biometric trait due to its encouraging features such as uniqueness, consistent shape, high

acceptability, easy collectability, and passive biometrics.

Despite of several inherent advantages of ear biometrics, issues in uncontrolled scenarios such as

illumination variation, pose changes, poor contrast, partial occlusion, and presence of noise restrict

to increase recognition performance. This opportunity gives sufficient chance for the recognition

improvement in ear biometric system. This factor motivates us to investigate the potential of ear

biometric characteristic with 2-D imagery. Objective of this thesis is to improve recognition performance

of the ear based unimodal and multimodal biometric systems.

Since performance of the ear biometric system depends on accurate ear localization and proper ear

image enhancement operations, we propose automatic ear localization and ear image enhancement

methods. In this thesis, our first contribution is an automatic ear image enhancement approach which

is used to enhance the degraded input ear images prior to use for feature extraction and recognition

operations. Otherwise, it is difficult to extract more detail local features from the ear images that

impart a negative effect on the recognition performance. Hence, it is desirable to enhance the quality

of low contrast ear images before ear recognition task. In this work, we propose an computationally

efficient and parameter free Jaya meta-heuristic optimization algorithm for ear image enhancement.

In addition to enhance convergence rate, we incorporate mutation operator in the proposed enhancement

approach named as enhanced Jaya algorithm.