01-08-2012, 12:48 PM
A New Method for ROI Extraction for Palmprint Recognition
A New Method for ROI Extraction for Palmprint Recognition.pdf (Size: 228.54 KB / Downloads: 79)
Abstract:
Palmprint verification system using Biometrics is one of the emerging technologies, which recognizes a person based on the principle lines, wrinkles and ridges on the surface of the palm. These line structures are stable and remain unchanged throughout the life of an individual. More importantly, no two palmprints from different individuals are the same, and normally people do not feel uneasy to have their palmprint images taken for testing. Therefore palmprint recognition offers a promising future for medium-security access control systems. Most of present system uses a specific ROI(Region of Interest) for extracting the surface of palm containing the above mentioned features. In this paper we proposed a novel method to acquire the ROI (Region of Interest).
Keywords: Palmprint recognition, ROI (Region of Interest), convex hull,
I. INTRODUCTION
The palm consists of many features, such as principal lines, ridges wrinkles, delta points. These features can be used sufficiently in verification processes. Compared with the other biometric technology, palm print authentication has several advantages. Firstly, the print patterns are unique, and the palm print characteristics are more abundant than fingerprint and iris. Secondly, the iris patterns required high resolution images. While the palmprint recognition use the principal and wrinkles which are also discriminating in low-resolution images. So the capture devise is less expensive than iris recognition. Moreover, palm print recognition system presents much higher user acceptability than iris and fingerprint.
Biometric recognition based on palm-print features contains different processing stages such as data acquisition, pre-processing, feature extraction and matching. This paper focuses on the pre-processing section which is quite important in providing high accuracy in pattern recognition. Among those papers which have discussed about palm-print recognition, seldom any of them discussed about pre-processing in detail. So, we pay more attention to this part and its most important stage which is extracting the ROI. This region that is the basis of pattern recognition in next processes (feature extraction and matching) is shown in Fig. 1(b). Generally there are two kind of images used in palm-print recognition: Online and Offline. Online images are those taken with digital cameras or scanners. Offline ones are those produced by ink on paper [5]. The database we use for testing our ethod is PolyU [6] that uses online images. The images in this database are low-resolution ones and are suitable for real-time application testing. A sample of the images in the discussed DB is shown in Fig. 1(a). In the following sections, we mention some prior work about the case in section 2. Section 3 provides the details of our proposed method. Section 4 discusses the experimental results including noise test. Section 5 concludes the major findings of this paper.
II PRIOR WORK
Most of the methods used to extract a region on palm for personal authentication use geometrical techniques [3, 7]. Nevertheless a paper which has extracted ROI using spectral approach is Han et al [8]. They have used wavelet-based segmentation to find the locations of finger tips and four finger roots. For attaining this, they have found boundary of palm image and then have transformed palm boundary’s coordinates to a profile of curvature. Then they used wavelet transform to convert the curvature into a multi-resolution signal (which included low- and high-frequency sub-bands). At last, they separated these sub-bands and detected corner points of palm’s boundary through local minimums of transformed profile (its high frequency sub-band). Although this method shows successful in extracting ROI precisely, it suffers from high demands of processing and therefore it is not appropriate for real-time applications. On the other hand, geometrical techniques used in other researches can not promise
high accuracy in extracting ROI and most of them
state that their approach is an approximate (not
absolute) solution [7]. A noticeable method which
uses geometrical techniques in extracting ROI is
discussed in [4]. It enjoys computing the center of
gravity of holes located between fingers and by this,
they extract ROI of the palm image. The main goal of
our paper is to improve the accuracy of ROI
extraction through this method and introduce new
techniques to eliminate the variations caused by
rotation and translation.
(a) (b)
Fig. 1. (a) A sample of palm image in PolyU DB; (b)
ROI on a palm image.
III METHODOLOGY
A. Pre-Processing:
The first step is preprocessing. The images
in the database we use, suffer from noise, shadows
and severe changes in illumination. So at the
beginning, we apply Gaussian Low Pass filter on
the images. This filter has a smoothing effect and
its significant vantage is that it does not make
images opaque.
The proposed method uses a low-pass
filter, g(x, y), on the original image , I(x,y), shown
in Figure 1(a). to obtain a blur version of the
image,M(x,y),