Palmprint based recognition system using phase-difference information

https://doi.org/10.1016/j.future.2010.11.029Get rights and content

Abstract

This paper presents an efficient palmprint based human recognition system. Each palmprint is divided into several square overlapping blocks. Reconstruction error using principle component analysis (PCA) is used to classify these blocks into either a good block or a non-palmprint block. Features from each good block of a palmprint are obtained by binarising the phase-difference of vertical and horizontal phase. The Hamming distance is used to compute the matching score between the features of corresponding good blocks of enrolled and live palmprint. These matching scores are fused using weighted sum rule, where weights are based on the average discriminating level of a block relative to other blocks. The performance of the proposed system is analysed on different datasets of hand images and it has been observed that it achieves a Correct Recognition Rate of 100% with a low Equal Error Rate for all the datasets. The system is also evaluated for noisy and bad palmprint images. It is found to be robust as long as the noise density is less than 50% or the bad region is less than 64% of the images.

Research highlights

► An efficient palmprint based human recognition system is proposed. ► A proposed technique is based using local phase-difference features of palmprints. ► The system is tested for performance on three different datasets of hand images. ► The system is also evaluated for noisy and bad palmprint images. ► The proposed system achieves a very high Recognition Rate and low Equal Error Rate.

Introduction

Biometric based recognition systems have wide applications in the field of personal identification/verification. Fingerprint based systems are most widely used while iris based systems are considered to be most reliable [1]. The palmprint is the region between the wrist and fingers. It has features such as texture, wrinkles, principle lines, ridges and minutiae points that can be used for its representation [2]. Like other biometric traits, palmprints also satisfy the critical properties of biometric characteristics such as universality, uniqueness, permanence and collectability [3]. Moreover, the use of palmprints over other biometric technologies has several advantages. It provides relatively stable and unique features. Collection of data is easy and non-intrusive. Co-operation required from the subject to collect data is low. It uses a low cost device and provides high efficiency using low resolution images. Palmprint based systems have high user acceptability [3].

Recently palmprint based recognition systems have received wide attention from researchers and efforts have been made to build a palmprint based system based on structural features of palmprints such as principle lines, wrinkles, datum points, minutiae points, ridges and crease points. Sobel and morphological operations have been used in [4] to extract line-like features from palmprints. In [5], isolated points along the principle lines are used as features. A system based on ridges of the palmprint eliminating creases has been proposed in [6]. End points of principle lines have been used as features in [7]. These points are found to be location and directional invariant. In [8], a palmprint based system has been proposed which uses crease points that are related to diseases of a person. In [9], directional line energy features which are characterised with the help of crease points are considered for identification through palmprints. Features such as ridges and minutiae are used for matching latent palmprint images in [10]. But the technique to extract features and to match features is found to be computationally expensive compared to fingerprints as its region is larger than a fingerprint. Further, most of the these systems are based on structural features and are not invariant to occlusion of the palmprint image.

Systems based on statistical features of palmprints include Principle Component Analysis (PCA) [11], Linear Discriminant Analysis (LDA) [12], Independent Component Analysis (ICA) [13], Discrete Cosine Transformation (DCT) [14], Fourier Transforms [15], Stockwell Transform [16], Moments [17], [18], Scale Invariant Feature Transform (SIFT) [19], [20], [21], Speeded-up Robust Features (SURF) [22], Gabor filter [23], [24], [25], Fusion code [26], Competitive code (Comp-code) [27], Ordinal code [21] and Wavelets [28], [29], [30]. There are systems [31], [32], [33] which have applied kernel PCA, LDA, ICA, PCA along with wavelets, DCT, FFT and Gabor filters. In [34], the statistical signatures like the center of gravity, density, spatial dispersivity and energy of the wavelet transformed image are considered as features. In [30], standard deviation of the small square region is used as a feature while mean and standard deviation of the small circular region is treated as features in [35] after transforming using a Gabor filter. In [36], a histogram pattern of a local binary image has been used as a palmprint feature. Global energy, global texture, local texture and fuzzy interest lines of the palmprint are considered as features in [37]. An identification system has been proposed in [17] which uses Hu invariant moments as features on an Otsu binarised palmprint image. A recognition system fusing the phase and orientation information has been proposed in [25]. A modified finite Radon transform has been given in [38], [39] to extract features of palmprints and matching is performed with an enlarged enrolled data set and pixel to area matching. A recognition system fusing the phase and orientation information has been proposed in [25]. In [40], [41], [42], a phase based image matching is proposed to achieve good results for verification. But these systems require to store the entire dataset along with features. Apart from a dramatic increase in storage requirements, two dimensional operations are found to be computationally more intensive and slow down the verification/identification process. Further, there also exist multimodal systems fusing features of palmprint with those of other traits like fingerprints [43], [44], palm veins [45], hand geometry [46], [47], face [14], [48], and iris [49] to improve the accuracy of the system. However, all the well-known identification/verification systems have been tested on noise-free and non-occluded palmprint images. Further, all these systems can identify the subject with his/her complete palmprint. If the subject is not able to expose the palmprint to the scanner completely due to medical injuries or being physically challenged, the systems fail to perform at their best.

This paper proposes an efficient palmprint based recognition system using phase difference information. The whole palmprint image is divided into some overlapping square blocks. The phase-difference of the vertical and horizontal phases of each of the square blocks is binarised using zero crossing to generate binary features. The nearest-neighbour approach is used for identification with the Hamming distance to measure the amount of similarity. The proposed system is tested on various datasets like IITK, CASIA and PolyU datasets.

The rest of the paper is organised as follows. Section 2 describes the phase which is used to extract palmprint features. Also it discusses a method of reconstructing the image using PCA. Section 3 proposes a system to extract features from the extracted palmprint obtained from the hand image. Two matching strategies have been proposed in Section 4. Performance of the proposed system is evaluated and results are analysed in Section 5. Conclusions are presented in the last section.

Section snippets

Background

This paper partitions the extracted palmprint into several overlapping blocks and makes use of the technique of image reconstruction using PCA to classify these blocks as good or bad. Features from each good block are extracted with the help of phase information of the block. This section describes the method of phase information which can be used for feature extraction and the PCA based image reconstruction technique.

System description

This section presents an efficient biometric system for recognition using phase-difference of a local region as features of a palmprint. The system acquires the hand image using a flat bed scanner at 200 dpi spatial resolution and 256 gray levels. The device is free of pegs. The region of interest or palmprint is extracted from the hand image. Stable valley points which are between the ring finger and the little finger, and between the index finger and the middle finger are used to extract the

Matching

The binary features extracted from the blocks of each palmprint are used for matching. There is a possibility of having some non-palmprint blocks in a live palmprint. While matching, these blocks are not considered. The PCA reconstruction method has been used for determining whether a block contains a palmprint or a non-palmprint image. A block containing a palmprint image is termed as a good block while that with non-palmprint image is considered as a bad block. Schematic diagram for the

Datasets

The proposed system has been tested on three different datasets of hand images. These datasets which are obtained from the Indian Institute of Technology Kanpur (IITK), Hong Kong Polytechnic University (PolyU) and Chinese Academy of Sciences Institute of Automation (CASIA) form a diverse representation of palmprints in terms of regions (ethnicity), device used to capture the images, resolution, lightning conditions and constraints (with or without pegs) used for placing the hand. This makes the

Conclusion

This paper has proposed an efficient palmprint based recognition system. The palmprint extracted from the acquired hand image is enhanced to correct the variation of brightness and to improve the texture of the palmprint using a coarse estimate of reflection and histogram equalization. The enhanced palmprint is divided into overlapping square blocks. The phase variations of all blocks are considered as features of the palmprint for its representation. The square blocks of the live palmprint are

Acknowledgements

Authors are thankful to the anonymous reviewers for their valuable comments and suggestions to improve the quality of the paper.

G.S. Badrinath received B.E. Degree in Computer Science and Engg. from Visveswaraiah Technological University, India, in 2003 and M.E. Degree in Information Technology and Engg. from Bangalore University, India, in 2005. He is currently a Ph.D. student in the Dept. of Computer Science and Engg. at Indian Institute of Technology Kanpur, India. His research interest includes biometrics, pattern recognition, computer vision and networks.

References (60)

  • D.D. Zhang

    Palmprint Authentication

    (2004)
  • J. Funada, N. Ohta, M. Mizoguchi, T. Temma, K. Nakanishi, A. Murai, T. Sugiuchi, T. Wakabayashi, Y. Yamada, Feature...
  • J. Chen, C. Zhang, G. Rong, Palmprint recognition using creases, in: International Conference on Information...
  • X. Wu, K. Wang, D. Zhang, Fuzzy directional element energy feature based palmprint identification, in: International...
  • A.K. Jain et al.

    Latent palmprint matching

    IEEE Transactions on Pattern Analysis and Machine Intelligence

    (2009)
  • X. Wu et al.

    based palmprint recognition

    Pattern Recognition Letters

    (2003)
  • X. Jing et al.

    A face and palmprint recognition approach based on discriminant dct feature extraction

    Systems, Man, and Cybernetics-B

    (2004)
  • L. Wenxin et al.

    Palmprint identification by fourier transform

    International Journal of Pattern Recognition and Artificial Intelligence

    (2002)
  • G.S. Badrinath, P. Gupta, Stockwell transform based palm-print recognition, Applied Soft Computing (2010), in press...
  • J.S. Noh, K.H. Rhee, Palmprint identification algorithm using hu invariant moments and otsu binarization, in: Fourth...
  • G.S. Badrinath et al.

    Palmprint based verification system robust to occlusion using low-order zernike moments of sub-images

    Telecommunication Systems

    (2010)
  • J. Chen, Y. Moon, Using sift features in palmprint authentication, in: International Conference on Pattern Recognition,...
  • G.S. Badrinath, P. Gupta, Palmprint verification using sift features, in: International Workshop on Image Processing...
  • Z. Sun et al.

    Ordinal palmprint representation for personal identification

  • G.S. Badrinath, P. Gupta, Robust biometric system using palmprint for personal verification, in: International...
  • D. Zhang et al.

    Online palmprint identification

    IEEE Transactions on Pattern Analysis and Machine Intelligence

    (2003)
  • R. Chu, Z. Lei, Y. Han, R. He, S. Li, Learning gabor magnitude features for palmprint recognition, in: Asian Conference...
  • X. Wu et al.

    Fusion of phase and orientation information for palmprint authentication

    Pattern Analysis and Applications

    (2006)
  • A.W.K. Kong, D. Zhang, Feature-level fusion for effective palmprint authentication, in: International Conference on...
  • A. Kong, D. Zhang, Competitive coding scheme for palmprint verification, in: International Conference on Pattern...
  • Cited by (61)

    • Decade progress of palmprint recognition: A brief survey

      2019, Neurocomputing
      Citation Excerpt :

      Angle principle makes use of the principle that ROI boundary point-valley point connection line and valley point connection line have a constant angle. And according to many experiments, the angle selected to be 45° or 60° is suitable for precise feature extraction [47]. Paradigms of three distinctive methods are illustrated in Figs. 7–9.

    • Monitoring of pet animal in smart cities using animal biometrics

      2018, Future Generation Computer Systems
      Citation Excerpt :

      However, due to weak signal, management of GPS-enabled devices is a challenging problem [10,16,35]. Therefore, the traditional animal recognition and monitoring systems do not provide the required level of control and security of pets [11,13,14,36–38]. The animal biometrics has provided a new paradigm for the recognition and monitoring of pet animals in the smart cities [37].

    • Vein Pattern Classification using Convolutional Neuronal Network and moment invariants

      2023, Proceedings of SPIE - The International Society for Optical Engineering
    View all citing articles on Scopus

    G.S. Badrinath received B.E. Degree in Computer Science and Engg. from Visveswaraiah Technological University, India, in 2003 and M.E. Degree in Information Technology and Engg. from Bangalore University, India, in 2005. He is currently a Ph.D. student in the Dept. of Computer Science and Engg. at Indian Institute of Technology Kanpur, India. His research interest includes biometrics, pattern recognition, computer vision and networks.

    Phalguni Gupta received the Doctoral degree from Indian Institute of Technology Kharagpur, India in 1986. Currently he is a Professor in the Department of Computer Science & Engineering, Indian Institute of Technology Kanpur (IITK), Kanpur, India. He works in the field of biometrics, data structures, sequential algorithms, parallel algorithms, on-line algorithms. He is an author of 2 books and 10 book chapters. He has published more than 200 papers in International Journals and International Conferences. He is responsible for several research projects in the area of Biometric Systems, Image Processing, Graph Theory and Network Flow. Prior to joining IITK in 1987, he worked in Space Applications Centre Ahmedabad, Indian Space Research Organization, India.

    View full text