Face recognition using Zernike and complex Zernike moment features

Abstract

Selection of a good feature extraction method is the most important factor in achieving the higher recognition rate in face
recognition. This paper presents the analysis of two moment based feature extraction methods namely Zernike moments (ZMs)
and Complex Zernike moments (CZMs) in application to face image recognition. We have intensively analyzed these methods in
terms of their reconstruction ability and invariance to rotation, scale and size. Almost all existing methods use only magnitude
component of the moments as invariant features in recognition task. Recently it is proposed that the phase component of moments
also captures useful information for image representation. In this paper, we have analyzed the performance of both magnitude
and phase coefficients of ZMs and call it CZMs. These methods are tested separately on suitable databases. The databases used
are UMIST pose database for rotation variation, JAFFE expression database for size and scale variations, and popular ORL and
FERET databases for comparison of recognition results. It can be concluded from the experimental results that the performance
of CZMs is not only better than ZMs but also it is the descriptor that gives best recognition rate amongst the descriptors
well known for face recognition.

Keywordsface recognition–moments–zernike moments–complex zernike moments–invariance