Hybrid Fingerprint Matcher

ABSTRACT:
We describe a hybrid fingerprint matching scheme that uses both energy & variance information to represent & match fingerprints. A set of 8 Gabor filters whose spatial frequencies correspond to the average interridge spacing in fingerprints is used to capture the ridge strength at equally spaced orientation. A square tessellation of the filtered images is then is used to construct an eight dimensional feature map called the ridge feature map. The ridge feature map along with energy feature set of a fingerprint image is used for matching purposes. Verification rate of the hybrid matcher is observed to be  92% and identification rate is found to be 90%. Fingerprints verification using the hybrid matcher is tested in MATLAB 6.5 environment.

INTRODUCTION:
Fingerprint matching techniques can be broadly classified as being minutiae- based or correlation-based. Minutiae- based technique attempt to align two sets of minutiae points and determine the total number of matched minutiae. Correlation-based technique, on the other hand, compare the global pattern of the ridges and furrows to see if the ridges in the two fingerprints align. The performance of the minutiae-based technique relies on the accurate detection of minutiae points and the use of sophisticated matching technique to compare two minutiae fields which undergo non-rigid transformations. The performance of the correlation-based technique is affected by non-linear distortion and noise present in the image. Jainetal.  have proposed a novel representation scheme that captures global and local feature of a fingerprint in a compact fixed length feature vector, called the FINGERCODE. This technique views a fingerprint as an oriented texture and their generic representation of oriented texture relies on extracting a core point in the fingerprint. Their technique , however, suffers from the following shortcomings:
(i) The frame of the reference is based on a global singular point i.e. the core point .Detection of the core point is non-trival; the core point may not even be present in small-sized images obtained using solid-state sensors.
(ii) The fingerprint alignment is based on a single reference point and is therefore , not very robust with respect to errors in the location of the reference point.
(iii) The tessellation does not cover the entire image . Furthermore, if the core were to be detected close to the boundary of the image , the tessellation will include an extremely small portion of the image.

We present a fingerprint representation scheme that constructs a ridge feature map by observing the local ridge orientation. The local ridge characteristics are extracted via a set of Gabor filter whose frequency corresponds to the inter-ridge spacing in the fingerprints. Unlike in the filtering is done on the enhanced images rather than the raw input images. Instead of using By tuning a Gabor filter to a specific frequency and direction, texture information from images can be extracted. An even symmetric Gabor filter has the following general form in the spatial domain: circular tessellation, a square tessellationis used . The tessellation includes the entire image and all the tessellated cells are of the same size. The tessellation is not based on detecting any landmark points. The fingerprint images are aligned using the overall minutiae information; this is more robust than using only the core point for aligning image pairs as done in .

Figure 1 Original fingerprint with core point.
Figure 2 Tessellated Fingerprint

RIDGE FEATURE MAPS:
where f is the frequency of the sinusoidal plane wave at an angle with the x-axis and dx and dy are the standard deviations of the Gaussian envelope along the x and y axes, respectively. For the 256 x 256 images taken from the AFIS Database, the Gabor parameters are set to the following values:
(i) f =0.1 (corresponds to an inter- ridge distance of 10);
(ii) dx = dy = d =4 (based on empirical data);
(iii) ?={0,22.5,45,67.5,90,112.5,135, 157.5} ,resulting in eight Gabor filters.
We enhanced the input image prior to filtering (in order to improve the clarity of the ridges and furrows ). Filtering requires convolving the enhanced image with each of the 8 Gabor filters in the spatial domain. In order to speed-up the filtering , the convolution is performed in the frequency domain. Eight filtered images are obtained as a result of this filtering(fig 4).
While a filtered image in its entirely can be used as a representation scheme, the presence of local distortion would affect the matching process drastically. Moreover it is the local variation in ridge structure (combined with the global ridge configuration) that will provide a better representation of the fingerprint. To examine local variations, the image is tessellated into square cells, and features from each of the cells as computed. The size of a cell is chosen (8 x 8). A 8-pixel wide border of the image is not included in the tessellation. This results in nc=15 cells in each row and column of the square grid, with a total of 225 cells. The variance of the pixel intensities in each cell of all 8 filtered images is used as a feature vector. The variance corresponds to the energy of the filter response and is therefore a useful measure of ridge orientation in a local neighborhood. Those tessellated cells that contain a certain proportion of background pixels are labeled as background cells and the corresponding feature value is set to zero.

HYBRID FINGERPRINT MATCHER:

The hybrid fingerprint matcher proposed here utilizes two distinct sets of fingerprint information: energy features and variance feature maps. When a query image is presented, the matching proceeds as follows:
Figure 3 (a) 00 variance feature image
Figure 3 (b) 450 variance feature image
(i) The query image is filtered using 8 Gabor filters.
(ii) The ridge feature map is extracted from these filtered images.
(iii) The query and template ridge feature maps are matched.
(iv) The energy and the variance feature map matching scores are combined to generate a single matching score.
Figure 3 (c) 900 variance feature image
For comparing the ridge feature maps to two images, it is necessary that the images themselves are aligned appropriately to ensure an overlap of common region in the two fingerprint images. The energy matching score is a measure of the similarity of the feature sets of the query and template images. The similarity score is normalized. The ridge feature maps of the query and the template images are compared by computing the sun of Euclidean distance of the 8-dimensional feature vectors in the corresponding tessellated cells.(cells that are marked as background are not used in the matching process). The distance score is normalized range and converted to a similarity score by simply subtracting it from 100. Let Se and Sv indicate the similarity scores obtained using minutiae matching and ridge feature map matching, respectively . Then, the final matching score S is computed as ,
s = a * Se + (1 -a ) * Sv
Where a = [0,1]. For the experimental results in this paper, was set to 0.5.

 Experiments and Results
The fingerprint database used in our experiments consist of fingerprint impressions obtained from 160 fingerprints, taken from theAFIS database which are of 20 persons each has given 8 images of same fingerprint. Each fingerprint is 256 x 256. The hybrid fingerprint matcher’s verification rate was found to be 92%. Whereas the identification rate was 90%. The results are tested in MATLAB 6.5 environment.

Download this Seminar Report

Password: be

Share on Google Plus

About Unknown

This is a short description in the author block about the author. You edit it by entering text in the "Biographical Info" field in the user admin panel.