Detecting and Counting Minutiae in Human Fingerprint

Human fingerprint is unique to every person and can be used easily for identification purposes. People leave fingerprints almost everywhere and that's why using fingerprint databases for investigation is so common. Fingerprints can be used in many important and vital applications: providing vital security, and sometimes fingerprint may be distorted (deformed) because it was taken from somewhere where the place was wood plank, wall or door hand, so we need to enhance the image, in this paper we will introduce  LBP enhancement for better minutiae detection and counting, and it will be showed how to construct a fingerprint identifier (features) to be used later as a key to retrieve or recognize the fingerprint.


Introduction
From the beginning of time to the current present day, data security systems have been a matter of interest and concern to everyone looking to protect vital and important data. Walls, gates, and watchtowers were some of the safety system used at that time and these days are still to secure our home, office, or workplace that required security from an unwanted hacker [1], [2]. These secure and safety systems have changed by the time, and they guarantee the latest technology to make this safety safer. Advanced safety and classic safety are always the same [3], [4].
Human fingerprint is unique to every person and can be used easily for identification purposes. People leave fingerprints almost everywhere and that's why using fingerprint databases for investigation is so common. Fingerprints can Human fingerprint is unique to every person and can be used easily for identification purposes. People leave fingerprints almost everywhere and that's why using fingerprint databases for investigation is so common. Fingerprints can be used in many important and vital applications: providing vital security, and sometimes fingerprint may be distorted (deformed) because it was taken from somewhere where the place was wood plank, wall or door hand, so we need to enhance the image, in this paper we will introduce LBP enhancement for better minutiae detection and counting, and it will be showed how to construct a fingerprint identifier (features) to be used later as a key to retrieve or recognize the fingerprint.. be used in many important and vital applications: providing vital security (for example, controlling access to areas or safe systems) ... conducting background checks (including government job applications, defensive security clearance, hidden weapons permits, etc.).
Fingerprint, papillary impression on the fingerprint ridges tips and thumb which are called minutiae. Fingerprints provide an impeccable way to personal identity, because minutiae arrangement on each finger of every human being is unique and does not change with growth or age.
Fingerprint image can be represented by a 2D matrix (for gray fingerprint image), or by a 3d matrix (for a color fingerprint image), the capture fingerprint image is subjective to some preprocessing operations such as: - Converting image to binary image.

Fingerprint structure
Human fingerprint image contains several unique objects each of them is called minutiae as shown in figure 1, the number and types and the locations of these objects are differ from one person to another [5].  Each minutiae in the fingerprint image can be easily detected depending on the 8 neighbors values and using the calculated classifier number (CN) as shown in figure 3 [3], [4], [5]: For better minutiae detection we need to enhance the fingerprint image, one of the popular methods used for image enhancement is histogram equalization [6], [7].
Image histogram is a one column array of 256 elements, each element value points to the repetition of a gray value (0 to 255) [8], [9], [10], the histogram may give us a true picture about the image, if the contents of histogram are normally distributed, then the image is clear, otherwise it requires equalization, and to do this we have to follow the following steps [11], [12]: Step 1: for images with discrete gray values, compute(formulas 1 and 2) [23], [24]: L: Total number of gray levels n k : Number of pixels with gray value n: Total number of pixels in the image Step 2: Based on CDF, compute the discrete version of the previous transformation: , (2) Table 1 shows an example of histogram equalization (here for simplicity we use a maximum gray value of 7) [13][14][15][16]. Here 0 will become 1, 1 well become 2 and so on. Sometime the histogram equalization fails when the fingerprint image has a large area of low-intensity background. In this case, the histogram will have a spike component corresponding to the background gray level. After histogram equalization, the output image will have a severe washed-out appearance while its dynamic range actually becomes smaller (see figure 4).
Considering the disadvantages of histogram equalization, we can use a LBP histogram method to enhance the fingerprint image [21], [22]. Creating LBP histogram Local binary pattern (LBP) histogram of the fingerprint image is a histogram of an output image after applying LBP operator calculations for each pixel in the input image [17], [18].
LBP image [19] can be obtained applying the steps shown in figure 5, the resulting image then will be used as an input image for fingerprint minutiae detection, this will give us a better enhancement, and this will be reflected in the accuracy of the process of detecting and counting minutiae in the fingerprints [25], [26], figure 6 show a sample example of new image pixel calculation based on LBP method: Fingerprint capturing.

2)
If the fingerprint is color, then convert the color image to gray one.

3)
The fingerprint may be distorted (deformed) because it was taken from somewhere where the place was wood plank, wall or door hand, so we need to enhance the image, and here we recommend using LBP enhancement because of the reasons mentioned previously, and the reasons that will be discussed from the obtained experimental results later.

4)
Convert the enhanced image to binary image.

5)
Apply image thinning using morphological thin operation.

6)
For each pixel in the thinned binary image do the following: a) Calculate CN. b) Add 1 to the type of minutiae denoted by CN value. c) Save the minutiae coordinates. d) Save the minutiae orientation angle.
These steps were implemented using matlab, several fingerprints were taken and below we will show the experiment results.
Although the first method of image enhancement (histogram equalization) takes less time, we recommend that you use the second method (LBP equalization) for the aforementioned reasons (see table 2). a) Without image enhancement Several fingerprints were selected, a matlab code was written and implemented, and table4 shows the results of detecting the number of minutiae and the counts for the most appearing in the image minutiae:  1  3556800  12  192  49795  2615  9  2  50625  0  0  0  0  0  3  296400  0  11  7380  343  5  4  3878400  0  172  57484  696  5  5  90000  0  42  7593  302  5  6  3747900  0  127  41803  1302  8  7  151200  1  156  6207  752  7  8  4757340  0  168  46849  528  5  9  3878400  0  7172  101339  21203  7  10  3292596  1  219  77182  1505  5  11  4164000  2  157  28708  187  5  12  262144  0  0  0  0  0  13  153450  0  0  0  0  0  14  5760000  1  338  44590  432  5 From Several fingerprints were selected, a matlab code was written and implemented, and table 5 shows the results of detecting the number of minutiae and the counts for the most appearing in the image minutiae: Several fingerprints were selected, a matlab code was written and implemented, and table 6 shows the results of detecting the number of minutiae and the counts for the most appearing in the image minutiae: Experiment 2: Using a segment from the fingerprint To reduce the minutiae detection and extraction time we can use a selected segment with a smaller size from the fingerprint, this segment can be used as an input image to detect and count minutiae. a)Without image enhancement Several fingerprints were selected, a segment from the image was defined, a matlab code was written and implemented, and table 7 shows the results of detecting the number of minutiae and the counts for the most appearing in the image minutiae:  implemented, and table 9 shows the results of detecting the number of minutiae and the counts for the most appearing in the image minutiae: is not a unique, so we cannot use each of them as a features (identifier) to retrieve or recognize the fingerprint, and this will lead us to seek a better method, one of these methods is to enhanced LBP fingerprint image.
As we said earlier in this paper, each fingerprint contains a unique number of minutiae and a unique count of each of them, so it is suitable to use them as a fingerprint identifier. From the above obtained results we can see that ridge ending and bifurcation minutiae have the most bigger counts, so we can focus on the to use them as a fingerprint identifier, by adding the Euclidian distant between coordinates [20] , these parameters are fixed for each fingerprint, thus we can easily use them as a fingerprint features, table 10 shows the features for the fingerprint 1, by taking a 100 by 100 pixels segment.  86  23  86  26  81  31  86  27  71  32  87  27  90  32  79  30  82  38  84  34  63  39  84  35  67  39  87  36  76  40  73  37  82  40  74  37 Conclusion Several methods of preparing a fingerprint as an input image to detect and count minutiae, histogram equalization gave better performance with a speedup of 3.5651 times comparing with LBP method(0.1369/0.0384), but he base method to be used is an LBP based histogram equalization, because some time using histogram equalization will lead to fault fingerprint features. From the obtained results we can conclude the following: -Each fingerprint has a fixed number of deferent types of minutiae.
-The counts of minutiae are fixed for each fingerprint and are unique.
-We can use a segment from the fingerprint to detect and count minutiae, this will reduce minutiae extraction time, For each fingerprint we can form the features (identifier) of each fingerprint by using ridge ending minutiae and bifurcation minutiae plus the coordinates Euclidian distant