KR20170087755A - Method and apparatus for extracting finger vein image based on fuzzy inference - Google Patents

Abstract

The present invention relates to a fuzzy inference-based finger vein extraction method and an apparatus therefor, wherein finger vein images are enhanced using a filter bank and a Gaussian filter, and fuzzy inference is applied based on the augmented enhancement image, The present invention relates to a method and apparatus for precisely separating a background portion and a finger vein portion included in a Mac image.

Description

FIELD AND APPARATUS FOR EXTRACTING FINGER VEIN IMAGE BASED ON FUZZY INFERENCE FIELD OF THE INVENTION [0001]

The present invention relates to a method and apparatus for extracting finger vein images based on fuzzy inference, and more particularly, to a method and apparatus for extracting finger vein images based on fuzzy inference from an image sensor, The present invention relates to a finger vein image extracting method and apparatus that can accurately perform user authentication through a finger vein by reducing the detection rate and significantly improving the finger matching rate.

Recently, due to the development of information and communication technology, a more secure technology is required for authentication of users in order to protect information and property rights of individuals or companies as modern society enters into information society.

Therefore, the importance of biometrics technology for user authentication has been greatly emphasized, and biometrics technology has been actively applied in various fields such as access control, finance, communication, or ePassport.

Generally, biometrics technology is a security technology that identifies and authenticates a user based on unique and unique biometric information that is different for each user. That is, the biometric authentication technique extracts biometric information (e.g., fingerprint, iris, or fingerprint) unique to each user and stores the extracted biometric information in a database provided in the biometric authentication device. Then, And authenticates the identity of the user by comparing the biometric information with the stored biometric information to determine its authenticity.

However, fingerprints can be replicated using materials such as gelatin, and iris recognition can be counterfeited with high resolution pictures. For this reason, a finger vein recognition technique which can not be easily observed with the naked eye and which authenticates a user by utilizing a finger pattern that is a feature of a living body of a user whose copying or forgery is almost impossible is attracting attention.

Generally, the finger vein recognition technology captures the infrared LED light projected onto the user's finger with a CCD camera sensor, extracts a finger vein pattern from the taken finger vein image, compares the vein pattern with a finger vein pattern stored in advance, .

However, the finger vein recognition technology may cause a difference in recognition performance depending on a method of separating the finger vein and the background image from the noise included in the finger vein image captured from the camera sensor and the transparency of the infrared light according to the finger thickness of the user .

In order to solve these problems, a water shed method, a local adaptation threshold method, a vein profile method, and the like have been developed. However, the above methods are limited to internal and external factors such as a user's measurement posture, (Lingyu Wang, G. Leedham, Gray-Scale Skeletonization of Thermal Vein Patterns Using the Watershed Algorithm in Vein Pattern Biometrics, International Conference on Computational Intelligence and Ecology, Vol. 2, pp. 1597 -1602, 2006, Tong Liu, Jianbin Xie, Huanzhang Lu, Wei Yan, Peiqin Li, A Threshold Image Method for Finger-vein Segmentation, Applied Mechanics and Materials Vols. 263-266, pp. 2439-2442, 2013 and Naoto Miura , Akio Nagasaka, Extraction of finger-vein patterns using Maxim Curvature Points in Image Profiles, IAPR Conference on Machine Vision Applications, May 16-18, pp. 347-350, 2005).

In order to solve the above problems, a method using a directional filter and a treble filter has been developed. However, the method using the directional filter and the treble filter has a problem in that, due to image noise not removed from the pre- (Truc, P., Khan, MA, Lee, Y., Lee, S., Kim, T .: Vessel enhancement filter using directional filter bank. Vol. 113, pp. 101-112, 2009, Zhang J, Yang J, Finger-vein Image Enhancement Based Combination of Gray-level Grouping and Circular Gabor Filter, Proceedings of Information Engineering and Computer Science, 2009, Yang, JF, Yang, JL, Shi, YH: Finger-vein segmentation based on multi-channel even-symmetric Gabor filter. ICIS, Vol.4, Shin KY, Lee HC, Park KR, Kim SM, Kim HC, Enhancement of Finger-vein Image by Vein Line Tracking and Adaptive Gabor Filtering for Finger vein Recognition, Vol. 145, pp. 219-223, 2012, Naoto Miura, Akio Nagasaka, Takafumi Miyatake. tracking and its application to personal identification, Machine Vision and Applications, Vol. 15, No. 5, pp. 194 -203, 2004).

In addition, since the above-mentioned prior art extracts a finger vein by using an exponential function or a frequency conversion in order to extract a finger vein from the projected image of the vein, the embedded system, which is a finger vein terminal for performing user authentication, (M. Khalil-Hani, PC Eng, Personal Verification using Finger Vein Biometrics in FPGA-based System-on-Chip, ELECO International Conference on Electrical and Electronics Engineering, 1-4 December, Bursa, TURKEY, pp. 151-156, 2011).

Also, the above-mentioned prior arts have a problem that the processing speed for authentication is slow because the recognized finger vein is wide and the image processing for separating the background and the vein is sensitive to the image noise from the image capture device as compared with the fingerprint authentication technique There is a very complicated problem.

Therefore, in the present invention, the finger vein image inputted from the image sensor such as the CCD camera is accurately separated from the finger vein part and the background part by using the fuzzy inference, whereby the detection rate of the finger vein and the finger matching rate can be remarkably improved A method of extracting finger vein images and a device therefor.

Next, a brief description will be given of the prior arts that exist in the technical field of the present invention, and technical matters which the present invention intends to differentiate from the prior arts will be described.

Korean Patent Registration No. 1315646 (Oct. 20, 2013) discloses a method and apparatus for extracting finger vein patterns using guided Gabor filter, and sets a region of interest from a captured vein capture image, The present invention relates to a method and an apparatus for extracting a finger vein pattern using a guided finger filter capable of extracting a finger vein from the obtained finger vein capture image by outputting a guided image with an improved image quality and performing Gabor filtering.

The prior art extracts a finger vein pattern from a preprocessed vein capture image. The vein pattern is different from one user to another, and the environment at the time of taking a finger vein image (finger position, rotation, The user can not extract the correct finger vein pattern because it is hard to be applied to the finger vein images of all users.

On the other hand, according to the present invention, the obtained finger vein image is augmented using a filter bank and a Gaussian filter, a finger vein feature image is extracted from the enhancement processed enhancement image, and the extracted vein feature The finger vein detection rate is lowered and the finger vein matching rate is remarkably improved by removing the finger vein portion erroneously extracted from the image and accurately separating the acquired finger vein image of the user from the background portion and the finger vein portion, It is possible to accurately perform the user authentication through the < RTI ID = 0.0 >

Korean Patent Registration No. 1025666 (Mar. 30, 2011) relates to a method and apparatus for extracting a finger vein feature point, comprising: acquiring a finger vein image by removing background from an acquired finger image; And extracting a branch point of the finger vein as a feature point.

The prior art is similar to the present invention in that a finger vein image is acquired from a sensor for photographing a finger to remove a finger vein portion and a background portion. On the other hand, The finger vein detection rate and the finger matching rate can be remarkably improved by accurately separating the finger vein part and the background part based on the filter bank and the fuzzy inference of the user's finger vein image inputted from the sensor. However, the prior art does not describe or suggest such technical features of the present invention.

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide an apparatus and a method for accurately discriminating finger vein portions and background portions using fuzzy inference, And to provide a finger vein image extraction method and apparatus thereof that can perform user authentication accurately.

The method of extracting finger vein images according to an embodiment of the present invention includes a binary LLBP extracting step of extracting a finger vein feature image by acquiring a finger vein image and a finger LLBP extracting step of extracting a finger vein feature- And a fuzzy inference step of extracting only an image.

The method further includes a finger vein image enhancement processing step for enhancing the acquired finger vein image to extract a finger vein feature image.

Calculating a mean value of augmented image pixels according to pixel coordinates existing in the extracted binary LLBP, and calculating a difference image of the averaged value and the augmented image, And inputting the reasoning.

The enhancing process may further include extracting a high-frequency finger vein image and an estimated background image from the finger vein image, dividing the estimated background image into the high-frequency finger vein image, .

In addition, the apparatus for extracting finger vein images according to an embodiment of the present invention includes a binary LLBP extracting unit for extracting a finger vein feature image and extracting a finger vein feature image, and a finger vein feature extraction unit And extracting only the finger vein image by extracting the finger vein image.

The finger vein image enhancement processor may further include a finger vein enhancement processor for enhancing the acquired finger vein image to extract a finger vein feature image, Extracting a finger vein image and an estimated background image of the feature and dividing the estimated background image into the finger image of the extracted high frequency characteristic to enhance the pixels constituting the finger vein part.

And a mean value calculation unit for calculating an average value of pixels for an augmented image according to pixel coordinates existing in the extracted binary LLBP and calculating a difference image between the averaged value and the augmented image, And the fuzzy inference is input.

The present invention relates to a fuzzy inference-based finger vein image extraction method and apparatus thereof, and more particularly, to a fuzzy inference-based finger vein image extraction method and apparatus that, when authenticating a user via a finger vein, The finger vein detection and the finger matching rate can be remarkably improved by accurately separating the finger vein portion and the background portion.

FIG. 1 is a block diagram schematically illustrating a process of extracting a finger vein image in which a background portion and a finger vein portion are separated from an input user's vein image, according to an embodiment of the present invention. to be.
FIG. 2 is a block diagram illustrating a process of enhancing a finger vein image of an input user in a finger vein image extraction apparatus according to an exemplary embodiment of the present invention. Referring to FIG.
3 is a block diagram illustrating a procedure of extracting finger vein images using binary LLBP extraction and fuzzy inference in a finger vein image extracting apparatus according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating an enhancement image and LLBP of an input user's finger vein image enhanced according to an exemplary embodiment of the present invention. Referring to FIG.
5 is a diagram illustrating an input / output fuzzy membership function according to an embodiment of the present invention.
FIG. 6 is a flowchart illustrating a method of extracting a finger vein image according to an exemplary embodiment of the present invention. Referring to FIG. 6, FIG. 2 is an illustration showing an example of a finger vein image.
7 is a block diagram illustrating a configuration of a finger vein image extraction apparatus according to an embodiment of the present invention.
8 is a flowchart illustrating a procedure for extracting a finger vein image from a user's vein image according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Like reference symbols in the drawings denote like elements.

FIG. 1 is a block diagram schematically illustrating a process of extracting a finger vein image in which a background portion and a finger vein portion are separated from an input user's vein image, according to an embodiment of the present invention. to be.

Generally, a finger vein authentication system is a system in which a finger of a user projected with a near-infrared ray is photographed with a CCD camera to acquire a finger vein image including a background to extract a finger vein feature, Authenticate the user.

In order to authenticate a user as described above, the most important steps in the process of recognizing and processing finger vein images are finger vein extraction and feature extraction. Particularly, the finger vein image is an image obtained by infrared ray projection, and the image contrast is low according to the finger thickness of the user, so that there is a large difference in the performance of the finger vein extraction according to the performance of the preprocessing step. .

Therefore, in a system for authenticating a user through a finger vein, it is very important to separate the background portion and the finger vein portion from the finger vein image and extract only the correct finger vein image.

In order to solve the above problem, many techniques for extracting only the finger vein portion from the finger vein image have been developed. However, the finger vein recognition rate is remarkably reduced according to the posture of the user, external light, and complexity of blood vessels There is a problem.

Further, since the above techniques use techniques such as an exponential function and a frequency conversion, a large processing speed is required in implementing the finger vein authentication system to which the above-described techniques are applied, and there is a problem that it takes a long time to recognize the finger vein .

Accordingly, in the present invention, based on the filter bank and the fuzzy inference, only the finger vein portion is extracted from the input finger vein image, and the finger vein detection rate and the finger matching rate are significantly improved, And to provide a device for extracting a finger vein.

As shown in FIG. 1, in order to extract a finger vein portion from a finger vein image, the finger vein image extraction device 100 first irradiates an infrared LED light source onto a finger of a user, irradiates the projected infrared ray to a CCD camera sensor And acquires the finger vein image of the user.

Next, the finger vein image extraction apparatus 100 enhances the acquired finger vein image by using a filter bank and a Gaussian filter.

The finger vein image extraction apparatus 100 processes the obtained finger vein image to enhance a pixel corresponding to a finger vein area in the finger vein image to extract an image having a low frequency characteristic, A subtraction image between the frequency image and the input finger vein image is obtained to acquire a high-frequency finger vein image. Further, the finger vein image extraction apparatus 100 acquires the estimated background image by performing the Gaussian filtering and root processing on the acquired high-frequency characteristic finger vein image, and acquires the acquired finger vein image Thereby enhancing the pixels corresponding to the finger vein portions.

As described above, a process for enhancing the pixels of the finger vein portion from the inputted finger vein image will be described in detail with reference to FIG.

Next, the finger vein image extraction apparatus 100 extracts the vein feature image (binary LLBP) from the enhancement image obtained by enhancing the finger vein image using the LLBP extraction method.

The LLBP extraction method is a method developed by Petpon and Srisuk for face recognition, and shows excellent performance in the feature extraction of the finger vein image.

Also, the LLBP extraction method calculates an index value by coding a relative brightness change of a peripheral region (e.g., an area of 3 x 3) of each pixel in a binary number, and calculates a histogram of the index value calculated for each pixel And then, the histogram is used as a texture model for the corresponding finger vein regions.

That is, the LLBP extracting method divides the finger vein image into cells of a predetermined size, calculates the index value for each cell, obtains a histogram, and extracts a vector connecting the histograms as a final image.

However, since the LLBP extraction method does not limit only the finger vein portion from the finger vein image but also extracts features for the background portion, the performance of the finger vein recognition deteriorates. Therefore, the present invention uses the fuzzy inference method to extract the LLBP extraction method So that only the accurate finger vein image is extracted. Thus, the finger vein recognition performance can be improved.

Next, the finger vein image extraction apparatus 100 extracts the binary LLBP from the LLBP with the pixels having high brightness as threshold values.

That is, the values of the pixels constituting the extracted binary LLBP have values of 0 and 1, the pixels having the value of 0 (black) constitute the background portion, and the pixels having the value of 1 (bright color) Constitute the vein section.

Next, the finger vein image extraction apparatus 100 calculates the difference image of the augmented image and the average value of the augmented image pixels according to the coordinates of the pixel having the value 1 based on the extracted binary LLBP. Further, the finger vein image extraction apparatus 100 extracts only the correct finger vein image using the calculated difference image and the binary LLBP as fuzzy inference inputs.

Hereinafter, a reinforcement process performed in order to separate the background portion and the finger vein portion from the input finger vein image will be described in detail with reference to FIG.

FIG. 2 is a block diagram illustrating a process of enhancing a finger vein image of an input user in a finger vein image extraction apparatus according to an exemplary embodiment of the present invention. Referring to FIG.

As shown in FIG. 2, the finger vein image extraction apparatus 100 preferentially enhances the user's finger vein image in order to accurately separate the finger vein portion and the background portion from the input finger vein image.

Further, the enhancement processing is performed by using the filter bank and the Gaussian filter, and comparing the background pixels of the finger vein image with the pixels of the finger vein portion other than the background.

Also, the enhancement processing step first down-samples the input finger vein image I _N (n, m) according to the following equation (1) to reduce the capacity of the finger vein image.

[Equation 1]

Next, the finger vein image extracting apparatus 100 up-samples the down-sampled finger vein image according to the following formula (2), and calculates a 7 x 7 size Sampled finger vein image is subjected to two-dimensional convolution ( ^* ) processing to extract a finger vein image I _G (n, m) having a low frequency characteristic by using the Gaussian filter G mask of FIG.

&Quot; (2) "

&Quot; (3) "

Next, the finger vein image extraction apparatus 100 obtains a substraction between the input finger vein image I _N (n, m) and the extracted finger vein image I _G (n, m) obtains the specified image having a high-frequency characteristic _{vein, I diff (n, m)} .

Specifies obtained above in the following Mac image, I _diff (n, m) for the background image estimated by performing a square processing with a Gaussian filter, and muscle (root) processed according to the following Equation 4], I _S (n, m) and divides the estimated background image I _S (n, m) by the high-frequency characteristic _finger vein I _diff (n, m) to enhance the pixels of the finger vein .

&Quot; (4) "

On the other hand, the enhancement is performed in order to enhance the fingerprint recognition performance by enhancing the pixels constituting the vein part which is a pixel other than the background.

Hereinafter, a process of extracting the finger vein portion using the binary LLBP extraction and the fuzzy inference, which is the next step for extracting the finger vein portion from the enhanced finger vein image, will be described in detail with reference to FIGS. 3 and 4 .

FIG. 3 is a block diagram illustrating a process of extracting finger vein portions using LLBP extraction and fuzzy inference in a finger vein image extraction apparatus according to an embodiment of the present invention. Referring to FIG.

3, the finger vein image extracting apparatus 100 extracts a finger vein image using the LLBP extraction method from the enhancement image I _E (n, m) obtained by enhancing the input finger vein image of the user, The feature image LLBP (n, m) is extracted.

In order to extract the LLBP, the finger vein image extraction apparatus 100 extracts the horizontal component of the LLBP using the threshold value function of Equation (5) and the following Equations (6) and (7) The image LLBP _v and the vertical component image LLBP _h are extracted.

&Quot; (5) "

&Quot; (6) "

&Quot; (7) "

Next, the finger vein image extraction apparatus 100 extracts the LLBP from the augmented image by extracting the extracted horizontal component image and the vertical component image according to the following equation (8).

&Quot; (8) "

Where h _n and v _n denote the values of the pixels on the vertical and horizontal lines, respectively, and h _c and v _c denote the center of the N-bit, c = N / 2 < / RTI >

Next, the finger vein image extraction apparatus 100 extracts a binary LLBP (L _B (n, m)) by thresholding pixels constituting the extracted LLBP according to the following equation (9).

&Quot; (9) "

Next, using the extracted binary LLBP and the following Equation (10), the finger vein image extraction apparatus 100 extracts pixels of the enhancement image according to pixel coordinates for pixels having a value of 1 in the binary LLBP (L _M ).

&Quot; (10) "

Next, the finger vein image extraction apparatus 100 calculates a difference image of the augmented image and an average value calculated according to the following Equation (11).

&Quot; (11) "

Next, the finger vein image extracting apparatus 100 calculates the membership function shown in Fig. 5 composed of the function according to the following equation (12) with the calculated difference image and the binary LLBP as the inputs of the fuzzy inference Fuzzy.

&Quot; (12) "

Where a and c represent the left / right position parameter of the triangle and b represents the position parameter of the maximum value of the triangle.

Next, the finger vein image extraction apparatus 100 extracts a finger vein portion according to the following Equation (13) using Mandini fuzzy inference and a weighted center non-fuzzy method based on the rule of the following [Table 1] To extract the finger vein image y (n, m).

[Table 1]

In Table 1, if the pixel of the augmented image corresponding to the same coordinate as the pixel coordinate having the value of 1 in the binary LLBP is dark, it is regarded as a finger vein. If it is bright, it is regarded as a background, It is.

&Quot; (13) "

Next, the finger vein authentication system 100 adapts the threshold value to the classified finger vein images according to the following equation (14) to extract a more accurate binary finger vein image. That is, when the pixels constituting the classified finger vein image exceed a specific threshold value, they are binarized to 1 and when they do not exceed the threshold value, they are binarized to extract clearer finger vein images.

&Quot; (14) "

4 (b) shows the LLBP output according to Equation (8), FIG. 4 shows the LLBP output according to Equation (8) (FIG. 4 (a)) and the LLBP extracted from the augmented image (FIG. 4 (b)). As shown in FIG. 4, if the LLBP is compared with the augmented image, it can be seen that the LLBP extraction method is effective for extracting the feature of the input FG image. However, It can be seen that it is extracted. This has the problem that the performance of the finger vein recognition can be deteriorated.

Accordingly, as described in detail with reference to FIG. 3, in the present invention, by using fuzzy inference to improve the above problem, only the finger vein portion can be accurately extracted from the LLBP.

FIG. 6 is a flowchart illustrating a method of extracting a finger vein image according to an exemplary embodiment of the present invention. Referring to FIG. 6, FIG. 2 is an illustration showing an example of a finger vein image.

Meanwhile, in order to verify the performance of the finger vein enhancement using the filter bank proposed in the present invention and extraction of the finger vein, wavelengths of 850 nm infrared LEDs were used. Adult male and female (age: 20 to 50 years) The images were photographed, and the images were augmented.

In addition, the performance of the finger vein feature image (LLBP) and the fuzzy inference extracted based on the enhancement processed finger vein image was evaluated.

In order to analyze the fingerprint image extraction performance, the adaptive threshold method, which is the most commonly used image segment technique, is compared with the present invention.

6 (a) shows a finger vein image taken by a CCD camera and FIG. 6 (b) shows a vein image finally extracted using an adaptive threshold method, which is a conventional method Fig. 6 (c) is a view showing a finger vein image finally extracted using the fuzzy inference-based finger vein image extraction method of the present invention.

6, if the finger vein image finally extracted by the adaptive threshold value method and the fuzzy inference-based finger vein image extraction method is visually analyzed, the finger vein image extracted by the adaptive threshold method is used as the finger vein image It can be seen that a portion for a similar pixel value is detected by mistaking it as a finger vein portion.

On the other hand, it can be seen that the present invention is significantly smaller in terms of false positives detected than the adaptive threshold method using an adaptive filter.

7 is a block diagram illustrating a configuration of a finger vein image extraction apparatus according to an embodiment of the present invention.

As shown in FIG. 7, the fuzzy inference-based finger vein image extraction apparatus 100 includes a photographing unit 110 for photographing a user's finger to acquire a finger vein image of a user, A binary LLBP extracting unit for extracting a binary LLBP from the enhancement processed finger vein image, an average value calculating unit for calculating an average value of pixels constituting the enhancement processed finger vein image based on the extracted binary LLBP, And a finger vein extraction unit 150 extracting a vein image based on the calculated average value and the extracted binary LLBP.

Meanwhile, the finger vein image of the user photographed by the photographing unit 110 means an original image or a row image including noise, finger vein part and background part of the CCD camera, The finger vein image extracted by the extracting unit 150 means an image obtained by extracting only the finger vein part by separating the background part and the finger vein part from the original image.

In addition, the photographing unit 110 includes an infrared LED (not shown) for illuminating infrared rays on the user's finger, an LED driver (not shown) for controlling the infrared LED, and a finger vein image from the user's finger And a CCD camera (not shown) for photographing.

The infrared LED is a near-infrared LED using a wavelength of 850 nm.

Further, the LED driver controls the direction or intensity of the infrared ray illuminating the infrared ray.

In addition, the CCD camera acquires a finger vein image of a user by photographing infrared rays illuminated on the finger of the user, and the finger vein image finally output by the camera is composed of a finger vein part and a background part do.

Further, the finger vein image enhancement processor 120 compares the obtained finger vein image of the user with pixels constituting the finger vein part and pixels constituting the background part using a filter bank and a Gaussian filter, The finger vein portion of one finger vein image is strengthened.

In order to perform the enhancement processing, the finger vein image enhancement processor 120 down-samples the obtained finger vein image to 1/2 and simultaneously up-samples the acquired vein image, and the Gaussian filter G A vein image having a low frequency characteristic is extracted from the acquired vein image by performing a two-dimensional convolution process on the up-sampled vein image using a mask.

Further, the finger vein image enhancement processor 120 calculates a difference between the extracted finger vein image of the low frequency characteristic and the finger vein image acquired from the photographing unit 110 to obtain a finger vein image having a high frequency characteristic .

Thereafter, the finger vein image enhancement processor 120 sequentially performs the square root processing, the Gaussian filtering, and the root processing on the acquired finger vein image of the high frequency characteristic according to Equation (4) And acquires the enhancement image in which the acquired background image is divided into the finger vein images of the high frequency characteristics to enhance the pixels of the finger vein portions other than the background.

The binary LLBP extractor 130 extracts the finger vein feature image LLBP from the augmented image.

Meanwhile, the LLBP is extracted by the LLBP extraction method, and the method extracts a finger vein portion from the augmented image, and can extract features including a portion of a background portion without limiting the finger vein portion.

Also, the binary LLBP extractor 130 extracts the binary LLBPs from the extracted LLBPs by thresholding pixels having high brightness. This makes it possible to acquire clearer finger vein feature images by binarizing the extracted finger vein feature images LLBP.

Also, the average value extracting unit 140 calculates an average value of pixels constituting the augmented image corresponding to the coordinates of pixels having a value of 1 in the extracted LLBP.

Further, the average value extracting unit 140 may calculate the difference image between the averaged value and the augmented image, and extract the accurate finger vein image by using the calculated difference image and the binary LLBP as the inputs of the fuzzy inference do.

Further, the finger vein extracting unit 150 receives the calculated difference image and the binary LLBP as input and fuzzifies it with the fuzzy function shown in FIG. 5, and based on the rule set in the table 1, .

On the other hand, extraction is performed using the Mandani fuzzy inference and the weighted center non-fuzzy method.

Further, the finger vein extracting unit 150 adapts the threshold according to Equation (14) to the extracted vein image to extract a more accurate binary finger vein image.

The finger vein image extraction apparatus 100 may further include a feature extraction unit 160 that extracts finger vein features based on the extracted binary finger vein images. By comparing with the finger feature, user authentication can be performed.

In the meantime, in explaining the present invention, a method of extracting a vein placed on a finger of a user has been described as an example. However, it is obvious that the present invention can be applied to extract veins located on the back of a hand or a wrist.

8 is a flowchart illustrating a procedure for extracting a finger vein from a finger vein image according to an embodiment of the present invention.

As shown in FIG. 8, the procedure of extracting a finger vein from a user's finger vein image is performed by first capturing a finger of a user through the photographing unit 110 and acquiring a finger vein image (S110).

The finger vein image may be obtained through the photographing unit 110, acquired through the Internet, or captured by a smart phone.

That is, the apparatus for extracting finger vein images 100 may include a separate wired / wireless interface (not shown) to acquire the user's finger vein images through various routes.

Next, the obtained finger vein image is augmented through the finger vein enhancement processing unit 120 (S120).

Wherein the enhancement processing includes extracting a high-frequency characteristic finger vein image from the finger vein image, acquiring a background image estimated from the finger vein image using a filter bank and a Gaussian filter, Is performed by dividing the image.

Next, the binarized LLBP is extracted from the augmented enhancement image through the binary LLBP extracting unit 130 (S130).

On the other hand, the extraction is performed by extracting LLBP from the augmented image using an LLBP extraction method and extracting pixels having high brightness values with threshold values from the brightness values of the pixels constituting the extracted LLBP.

Next, the mean value extracting unit 140 calculates an average value of the pixels of the augmented image according to the pixel coordinates of the binary LLBP, and calculates a difference image between the averaged value and the augmented image (S150).

On the other hand, the difference image calculates an average value of pixels constituting the augmented images corresponding to the coordinates of pixels having a value of 1 in the binary LLBP (i.e., finger vein portions), and subtracts the augmented image from the averaged value .

Next, the finger vein extraction unit 150 extracts the finger vein image from the binary LLBP (S160).

On the other hand, the extraction is performed using a fuzzy inference method and a weighted center non-fuzzy method.

As described above, the fuzzy inference-based finger vein image extracting method and apparatus thereof according to the present invention is a method for extracting a finger vein image based on a finger vein image of a user input from an image sensor such as a CCD camera by fuzzy inference , The finger vein detection rate and the finger matching rate are remarkably improved, and the user authentication through the finger vein can be accurately performed.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. .

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the present invention.

100: finger vein image extraction device 110:
120: finger vein image enhancement processing unit 130: binary LLBP extraction unit
140: average value calculation unit 150: finger vein extraction unit
160: Feature extraction unit

Claims (7)

A binary LLBP extraction step of acquiring a finger vein image and extracting a finger vein feature image; And
And a fuzzy inference step of extracting only a finger vein image by removing a finger vein portion erroneously extracted from the extracted finger vein feature image. The method according to claim 1,
The method of extracting finger vein images according to claim 1,
And a finger vein image enhancement processing step for enhancing the obtained finger vein image to extract a finger vein feature image. The method of claim 2,
Calculating an average value of augmented image pixels according to pixel coordinates existing in the extracted binary LLBP; And
And calculating a difference image between the averaged value and the augmented image,
Wherein the binary LLBP and the difference image are input to the fuzzy inference. The method of claim 2,
The reinforcement processing step includes:
Extracting a high-frequency characteristic finger vein image and an estimated background image from the vein image,
Dividing the estimated background image into the finger vein images of the extracted high frequency characteristics to enhance the pixels constituting the finger vein portions. A binary LLBP extraction unit for acquiring a finger vein image and extracting a finger vein feature image; And
And a finger vein extracting unit for extracting only the finger vein image by removing a finger vein portion erroneously extracted from the extracted finger vein feature image using fuzzy inference. The method of claim 5,
Wherein the finger vein image extracting device comprises:
And a finger vein image enhancement processor for enhancing the acquired finger vein image to extract a finger vein feature image,
Wherein the finger vein image enhancement processor comprises:
Extracting a high-frequency characteristic finger vein image and an estimated background image from the vein image,
Dividing the estimated background image into the finger vein image of the extracted high frequency characteristic to enhance the pixels constituting the finger vein portion. The method of claim 6,
And an average value calculation unit for calculating an average value of the augmented image pixels according to pixel coordinates existing in the extracted binary LLBP and calculating a difference image between the averaged value and the augmented image,
Wherein the binary LLBP and the difference image are input to the fuzzy inference.

Images