The identification of fingerprints in the security information system is investigated. Fingerprint scanning was performed, papillae thinning was performed, Sherlock and Monroe fields were orientated, Ateb-Gabor filtration was performed, and three-dimensional Ateb-Gabor properties were investigated for different rational parameters. Their effect on filtering and subsequently on identification is revealed. Ateb-Habor filtering for biometric prints is applied, which significantly extends classic filtering by implementing a broader set of filters and providing a comprehensive approach to identification. The skeleton was executed using the Hilditch algorithm. The next step is to create the direction fields, using the method proposed by Sherlock and Monroe when forming the direction field of the papillary lines. Uses an average spine / trough nine pixels period. This imitates a sensor with the discriminability of 500 dpi. Frequency in areas decreases depending on position. The Hilditch algorithm generates a vector image. This image is in the form of a loaded graph - that is, found endpoints, points of intersection of the top of the graph and the lines and arcs of the edges of the graph.
Fingerprint identification has become widely used lately. Various types of sensors — optical, capacities, ultrasonic and thermal — are used to obtain digital images of fingerprints. To date, the most common fingerprint imaging sensors are optical sensors.
There are two main categories of fingerprint comparison methods — the comparison of defined points and the comparison of the whole pattern of the fingerprint — a template-on. The pattern identification method compares two images to see its resemblance. This method is typically used in fingerprint identification systems (show Fig.1). More commonly used technology is recognition based on point-by-point comparisons.
Skeletonize algorithms are determined on consistent deletion of outline points.
These points get rid of and the line becomes thinned out. The thinning algorithm is the
removing of edge points where throughout the image moves mask 3 × 3. These set of
rules apply to every 3X3 window. Consider building a skeleton using the Hilditch
algorithm (show Fig.2). The algorithm is designed to work with binary images. The
algorithm is to scan the post-iterative pixel matrix of the image mask position and the
consistent change of black pixels on white. The algorithm was described in [
To create the direction field, we use the method proposed by Sherlock and Monroe in
forming the direction field of the papillary lines. As stated in [
The Sherlock and Monroe method was proposed in [
The orientation model proposed by Sherlock and Monroe [
… ,
- j-е nuclei end - i-delta in the complex plane [
The overall periodicity is selected consonant to the frequency distribution of the comb lines in actual fingerprints. Uses an average spine / trough period of nine pixels. This creates a sensor effect with a resolution of 500 dpi. Frequency in areas decreases depending on position.
The Sherlock and Monroe model is used as follows. The fingerprint is filtered by Ateb-Gabor. Further, thinning is carried out. The positions of nuclei and deltas are specified. A field of directions is built later. 2.1
Gabor wavelets are widely used for filtering fingerprints, and Gabor characteristics are recognized as the best representation for fingerprint recognition in terms of recognition factor. Moreover, it is demonstrated with resistance to changes in lighting and noise. When only one reference image is available per scanned image, they offer an adaptive weighted sub-Gabor's array for representing and recognizing fingerprints.
In [
Gabor filters are typically used for texture analysis, edge detection, feature extraction, unevenness estimation (in stereo), and more. Gabor filters include special classes of bandpass filters that pass a certain frequency band and reject others.
Images are filtered by Gabor filters in much the same way as conventional filters. We have a mask, the more precise term for it will be the "convolution core" that represents the filter. By mask we mean that we have a two-dimensional array. We use twodimensional images. In these images, each pixel is assigned a weight value. The above array moves on each pixel of the image and performs a convolution operation. When applied to an image, the Gabor filter gives the greatest response at the edges and in places where the texture changes. The following images show the test image and its transformation after applying the filter.
The Gabor filter responds to changes in edges and textures. The filter responds well to the spatial location of the function. This occurs when coagulation kernels are applied in the spatial domain, in the frequency domains.
The method based on the application of the Gabor filter is quite simple and effective in constructing images of papillary lines. An iterative change of an input image containing one or more isolated sources causes the image to grow with local orientation. As a result, a consistent and very realistic picture of the papillary lines gradually emerges. In this case, in the random positions of the fingerprint appear minute different types. 2.2
Gabor filtering [
In this work we propose to use a Ateb-Gabor filtering to filter biometric
fingerprints, which greatly expands classic filtering, implementing a wider set of filters and
providing a comprehensive approach. Filtration [
Skeletonize transforms objects in images into a type of skeleton. This is distorted to represent the topology of the object by performing sequential image passages. Each time, pixel boundaries are identified and deleted, provided that they do not disrupt the object. Many skeletonization algorithms have been developed as an iterative sequential deletion of points on contours. Thinning algorithms, in particular Hilditch, work to eliminate boundary points so that the 3 × 3 window moves across the pixels of the image and the contents of the window apply the rules. Skeletonization works when sequentially scaling images and then removing pixels at the edge of an object. This lasts as long as there are no more pixels at the edge of the image. It works like a mask. The search table is then used to write to pixels of values 0, 1, 2, or 3 that are selectively deleted during iterations.
Thinning algorithms are considered in the work of Rutovitz [
Another well-known algorithm is the Hilditch algorithm [
And work with thin vector elements was developed in [
Histogram alignment is a nonlinear process. Separating channels and aligning each channel individually is not a good way of leveling contrast. Alignment includes image intensity values, not color components.
This should be used to ensure that the intensity value is evenly balanced and does not in any way disturb the color balance. Therefore, the first step is to rebuild the color space from RGB into one of the color spaces, where the pixel intensity values are separated from the color components.The binarization results are shown in Figure 9. The initial image in grayscale is to the left Fig. 9a. In Figs. 9b shows an increase in countergrowth, 9c shows segmentation, which consists of splitting a digital image into several segments. This is done to simplify and change the presentation of the image to facilitate its analysis. In Figs. 9d shows local normalization, which is to bring the image back to the origin that is acceptable for recognition. In Figs. 9e shows a Gabor filter transceiver. In Fig.10 show is image filtering: a - adding a processed image, bbinarization; c - median filtration; d- function extrac, e -filtered image.
If there are many colors in the image, the splitting method will cause a color
imbalance. After the images were processed, they were subject to RSA encryption according
to the algorithms [
Biometric fingerprints were identified and recognition was based on new filtering techniques. Biometrics were scanned, and papillae were thinned using the Hilditch algorithm. This method produces a vector representation of the image in the form of downloaded graphs. This is done to separate the end points, that is, the points of intersection of the top of the graph and the lines and arcs that contain the edges of the graph.
Field orientation was performed using the Sherlock and Monroe method. AtebGabor-based filtering is implemented. Ateb-Gabor properties were investigated for various rational parameters and their effect on filtering was carried out. Fig. 10. Image filtering: a - adding a processed image, b-binarization; c - median filtration; dfunction extrac, e -filtered image