Proceeding of the 2nd International Conference on Informatics, 2007
PRECISE FINGERPRINT ENROLMENT THROUGH PROJECTION INCORPORATED SUBSPACE BASED ON PRINCIPAL COMPONENT ANALYSIS (PCA) Md. Rajibul Islam, Md. Shohel Sayeed, Andrews Samraj Faculty of Information Science and Technology (FIST) Multimedia University, Jalan Ayer Keroh lama, 75450 Melaka, Malaysia E-mail: {md.rajibul.islam05, shohel.sayeed, andrews.samraj}@mmu.edu.my
ABSTRACT
Despite recent advances in the area of fingerprint identification, fingerprint enrolment continues to be a challenging pattern recognition problem. The first step to this problem is the enhancement of landmarks as well as precise minutiae points (ridge bifurcation and ridge ending), core, plain ridges from a print. Once enhanced, these fingerprint images are then ready to extract features and store into a database. Later these are compared to all sets on file in search of a match. The accurate fingerprint image is the basis for the entire identification and matching process. Various enhancement approaches have been proposed in the literature, each with its own merits and degree of success. The most common approach is to enhance and store the precise fingerprint image through normalization, orientation, frequencies calculation, contextual filtering and then binarisation and masking. Our emphasis in this paper is to enhance and store the fingerprint image accurately using Projection Incorporated Subspace based on Principal Component Analysis (PCA). In particular, we have implemented the methods based on eigenspace representations and neural network classifiers. Moreover, we present preliminary results of an attempt to mingle the outputs of these methods using a clustering algorithm unique to this type of problem.
Keywords: Fingerprint enrolment, Minutiae, Projection Incorporated Subspace, PCA, Region merging.
1.0 INTRODUCTION Fingerprint identification is one of the most admired biometric technologies and is used in biometric personal identification, criminal investigations, commercial applications, and so on. The performance of a fingerprint imagematching algorithm depends heavily on the quality of the input fingerprint images [1]. It is very significant to acquire good quality images but in practice a significant percentage of acquired images are of poor quality due to some environmental factors or user’s body condition [2]. The poor quality images cause two problems: (1) many spurious minutiae may be created and (2) many genuine minutiae may be ignored. Therefore, a novel approach is necessary to increase the performance of the minutiae extraction algorithm. In this paper, we have presented a subspace method based on Principle Component Analysis which incorporates the projection information of the fingerprint images. This method uses a subspace of lower dimension than that used by PCA. Also, its correct recognition rates are superior to PCA. The rest of this paper is structured as, section 2 briefly introduces the Projection Incorporated Subspace and PCA, section 3 the region merging technique (assume that section 2 and section3 are the proposed approach), section 4 presents some experimental results. Finally, section 5 concludes this article.
2.0 PROJECTION INCORPORATED SUBSPACE Projection Incorporated Subspace has been described by Wu Jianxin, Chen Zhaoqian and Zhon Zhihua in [4]. The reason we use the projection incorporated subspace method that it requires less eigenvectors. Using fewer eigenvectors means that fewer computational power and processing time is needed. In real-world applications the fingerprint database may have thousands of individuals or even more, therefore the saving of computational cost may be quite significant. Suppose P(x, y) be an intensity image of size N1×N2 , satisfies x[1, N1 ] , y[1,N2] , P(x, y)[0,1]. The vertical and horizontal integral projections are defined respectively as:
©FCSIT, UM 2007
T1-85
Proceeding of the 2nd International Conference on Informatics, 2007
Define the projection map Mp (x, y) for P(x, y) as
in which P is the image’s mean intensity, defined as
Then, a projection incorporated version of P(x, y) is defined as
in which is called combine parameter. Since P (x, y) may go out of [0, 1], exhibit of the fingerprint image may be imprecise although the recognition results will not be affected using projection map. For better display, the fingerprint image in Fig. 1 has been adjusted according to:
In the above process, PCA are performed on the projection incorporated version of the fingerprint image instead of on the original image. We call the subspace find by this technique projection incorporated subspace method. And all the subspaces are shown by some boundary boxes (see Fig. 1).
(m) (n) Fig. 1: Fingerprint image using projection incorporated subspace method based on PCA. Image (n) is the enhance part of image (m). 2.1 Principal Component Analysis (PCA) PCA is a useful statistical technique that has found application in fields such as fingerprint recognition and image compression, and is a common technique for finding patterns in data of high dimension [8]. The main effect of PCA is dimensionality reduction, that is, mapping the n-dimensional vector x into an m-dimensional space, where m<