Magnitude and Phase of Discriminative Orthogonal Radial Moments for Face Recognition

Magnitude and Phase of Discriminative Orthogonal Radial Moments for Face Recognition

Neerja Mittal (CSIR- Central Scientific Instruments Organisation, India), Ekta Walia (South Asian University, India) and Chandan Singh (Punjabi University, India)
DOI: 10.4018/978-1-4666-6030-4.ch007
OnDemand PDF Download:
$37.50

Abstract

It is well known that the careful selection of a set of features, with higher discrimination competence, may increase recognition performance. In general, the magnitude coefficients of some selected orders of ZMs and PZMs have been used as invariant image features. The authors have used a statistical method to estimate the discrimination strength of all the coefficients of ZMs and PZMs. For classification, only the coefficients with estimated higher discrimination strength are selected and are used in the feature vector. The performance of these selected Discriminative ZMs (DZMs) and Discriminative PZMs (DPZMs) features are compared to that of their corresponding conventional approaches on YALE, ORL, and FERET databases against illumination, expression, scale, and pose variations. In this chapter, an extension to these DZMs and DPZMs is presented by exploring the use of phase information along with the magnitude coefficients of these approaches. As the phase coefficients are computed in parallel to the magnitude, no additional time is spent on their computation. Further, DZMs and DPZMs are also combined with PCA and FLD. It is observed from the exhaustive experimentation that with the inclusion of phase features the recognition rate is improved by 2-8%, at reduced dimensions and with less computational complexity, than that of using the successive ZMs and PZMs features.
Chapter Preview
Top

Introduction

The main objective of face recognition system is to identify a person in the database of stored face images. These days it is one of the most popular research topics because of having many commercial and law enforcement applications, e.g. criminal identification at public places, human computer interaction, video surveillance, bankcard verification, etc. Although there exist some good reliable methods of human identification such as iris recognition, fingerprint recognition, etc., these methods depend on the cooperation of individuals, whereas a face recognition system can work without the support or knowledge of the targets. Most of the face recognition approaches lack in satisfactory results under the conditions of lighting directions, pose variation, aging, occlusion and real world problems, etc. (Zhao et al., 2003).

In literature, the existing face recognition approaches are broadly classified into two categories, namely, local methods and the global methods (Hjelmas & Low, 2001). In local methods, structural features of facial shape like eyes, nose and mouth are analyzed. These methods are not affected by irrelevant information in the images, but are sensitive to unpredictability of face appearance and noise. Global methods are statistics based methods which use the complete information of faces and are less sensitive to noise. Principal component analysis (PCA) (Burton et al., 2005; Neerja & Walia, 2008; Moon & Phillips, 2001; Turk, 2001), Fisher linear discriminant (FLD) (Belhumeur et al., 2005; Etemad & Chellappa, 1997; Martinez & Kak, 2001), two-dimensional PCA (2DPCA) (Xu et al. 2008), two-directional two-dimensional PCA (2D2PCA) (Daoqiang & Zhi-Hua, 2005) and moment invariants (Singh, 2006; Teh & Chin, 1988; Wee & Paramesran, 2007) fall under the category of global methods. Orthogonal radial moments and their functions have been used as useful image features in a number of applications (Teague, 1980). Among these orthogonal radial moment based methods, Zernike moments (ZMs) and pseudo-Zernike moments (PZMs) are particularly useful because they are less susceptible to information redundancy and image noise. The magnitude of these orthogonal radial moments is invariant to rotation and under certain geometric transformations they can be made invariant to scale and translation. Some work has already been done to present the usefulness of moment based invariant features for the recognition of face images (Nor’aini et al., 2006; Singh et al., 2011a; Singh et al., 2011b; Singh et al., 2011c). A comparative analysis of different moment invariants is presented in (Faez & Farajzadeh, 2006; Nabatchian et al., 2008). Nor’aini et al. (2006) have observed that the ZMs are able to recognize face images better than PCA because of its characteristics of being invariant to rotation and insensitivity to image noise. PZMs using neural network classifier (Haddandnia et al., 2003) and the ZMs/ PZMs using optimal similarity measure (Singh et al., 2011b) gave better results as compared to the Euclidean distance classifier.

Complete Chapter List

Search this Book:
Reset