KR100701201B1 - Mathod and apparatus for classification of fingerprint image quality and fingerprint image recognition system using the same - Google Patents

Abstract

The present invention relates to a method and apparatus for classifying a fingerprint image and a fingerprint image recognition system using the same, comprising: (a) dividing a fingerprint image into a plurality of blocks; (b) calculating and vectorizing a plurality of fingerprint image quality classification parameters for each of the plurality of divided blocks; (c) inputting the vectorized fingerprint image quality classification parameter into a previously learned quality classifier to obtain a quality determination value for each of the plurality of divided blocks; And (d) setting a representative value from the quality determination value for each of the plurality of blocks to determine the quality of the fingerprint image. Therefore, it is possible to more accurately determine the quality of the fingerprint image can provide a reliable fingerprint image recognition system.

Description

Method and apparatus for classifying a fingerprint image and a fingerprint image recognition system using the same {Mathod and apparatus for classification of fingerprint image quality and fingerprint image recognition system using the same}

1 is a block diagram of a fingerprint image recognition system including a device for classifying a fingerprint image according to the present invention.

FIG. 2 is a block diagram illustrating in detail the fingerprint image quality classifying unit of FIG. 1.

3 is a flowchart illustrating a quality classification method of a fingerprint image according to the present invention.

4 illustrates step S320 of FIG. 3 in more detail.

5A through 5D are diagrams illustrating an example of a quality determination value of a fingerprint image in FIGS. 2 and 3.

FIG. 6 is a diagram illustrating a quality determination value of a fingerprint image in FIGS. 2 and 3 in a three-dimensional space.

The present invention relates to a method and apparatus for classifying a fingerprint image and a fingerprint image recognition system using the same. More particularly, the quality of the fingerprint image is classified using a pre-learned quality classifier, and the feature information of the fingerprint image is classified. The present invention relates to a fingerprint image quality classification method and apparatus for providing a more reliable fingerprint image recognition system and a fingerprint image recognition system using the same.

Biometric technology has recently been emphasized as a technology for identifying a user by using the physical and behavioral characteristics of each individual. In particular, the fingerprint image recognition system using the fingerprint image has been positioned as the main user identification technology in the field of biometric technology due to the accuracy, convenience, light and small and low cost of the input means.

Fingerprint image recognition system has very high reliability for economical installation cost and security and has been proved with unique characteristics of unique person based on worldwide application cases for hundreds of years. Space utilization is very high. In particular, with the development of networks these days, with the growing interest in security and privacy, the fingerprint image recognition system is developing into a popular technology field among user identification technologies.

In the fingerprint image recognition system, in particular, the stability of the process of registering the fingerprint image of the user is required, and the stability is closely related to the performance of the feature information extraction algorithm for the fingerprint image. In addition, the feature information extraction algorithm for the fingerprint image is generally affected by the quality information of the fingerprint image. Therefore, the method of more accurately distinguishing the quality information of the fingerprint image will help to find a method of improving the algorithm for extracting feature information of the fingerprint image. As a result, the method of improving the feature information extraction algorithm of the fingerprint image may induce a result of improving the performance of the fingerprint image recognition system.

However, conventionally, only one preprocessing step is applied in the process of extracting feature information from a fingerprint image, and thus there is a problem in that the quality of the fingerprint image cannot be distinguished more accurately.

The present invention for solving the problems of the prior art is to analyze the correlation between the fingerprint image recognition and the fingerprint image quality in the fingerprint image recognition system and fingerprint image quality classification method and apparatus that can more accurately distinguish the quality of the fingerprint image The present invention also provides a fingerprint image recognition system having a feature information extraction algorithm for extracting fingerprint image feature information more reliably through the quality of the classified fingerprint image.

According to an aspect of the present invention, there is provided a quality classification method of a fingerprint image, the method comprising: dividing a fingerprint image into a plurality of blocks; (b) calculating and vectorizing a plurality of fingerprint image quality classification parameters for each of the plurality of divided blocks; (c) inputting the vectorized fingerprint image quality classification parameter into a previously learned quality classifier to obtain a quality determination value for each of the plurality of divided blocks; And (d) setting a representative value from the quality determination value for each of the plurality of blocks to determine the quality of the fingerprint image.

In addition, the apparatus for classifying a fingerprint image quality of the present invention for solving the above technical problem, the partition unit for dividing the input fingerprint image into a plurality of blocks; A quality classification parameter calculator for calculating a fingerprint image quality classification parameter for each of the plurality of divided blocks; A vector unit vectorizing fingerprint image quality parameters for each of the plurality of divided blocks; A quality determination value obtaining unit obtaining the quality determination value for each of the plurality of divided blocks by inputting the vectorized fingerprint image quality parameter to a predetermined quality classifier; And a quality determination unit that sets a representative value from the quality determination values for each of the plurality of divided blocks to determine the quality of the fingerprint image.

Fingerprint image recognition system of the present invention for solving the technical problem, the fingerprint image acquisition unit for obtaining a fingerprint image of the user; A fingerprint image quality classifier classifying a quality of the obtained fingerprint image; A fingerprint image feature information extraction unit for extracting feature information of the fingerprint image through the quality of the classified fingerprint images; And a fingerprint image comparator for authenticating the user through the feature information of the fingerprint image extracted by the fingerprint image feature information extractor.

Hereinafter, with reference to the accompanying drawings will be described in detail the operation of the preferred embodiment according to the present invention.

1 is a block diagram of a fingerprint image recognition system including a device for classifying a fingerprint image according to the present invention. Referring to FIG. 1, the fingerprint image recognition system includes a fingerprint image acquisition unit 100, a fingerprint image quality classification unit 110, a fingerprint image feature information extractor 120, a storage unit 130, and a fingerprint image comparison unit 140. ), A key input unit 150, and an interface unit 160.

The fingerprint image acquisition unit 100 includes a fingerprint input sensor 101 and an analog / digital converter 102 to obtain a fingerprint image by scanning a user's fingerprint. The fingerprint input sensor 101 scans a user's fingerprint, and the analog / digital converter 102 digitizes the fingerprint scanned by the fingerprint input sensor 101 to obtain a fingerprint image of the user. The fingerprint image obtaining unit 100 outputs the obtained fingerprint image of the user to the fingerprint image quality classifying unit 110 and the fingerprint image feature information extracting unit 120.

The fingerprint image quality classifying unit 110 classifies the quality of the fingerprint image of the user received from the fingerprint image obtaining unit 100 and outputs the quality information of the classified fingerprint image to the fingerprint image feature information extracting unit 120. .

The fingerprint image feature information extractor 120 receives a fingerprint image from the fingerprint image acquisition unit 100 and also receives quality information of the fingerprint image from the fingerprint image quality classifier 110. The fingerprint image feature information extractor 120 extracts feature information of the fingerprint image from the received fingerprint image and quality information of the fingerprint image. Here, the characteristic information of the fingerprint image refers to a unique characteristic to be used for user authentication with respect to the fingerprint image.

The key input unit 150 may input user identification information for performing user authentication through the fingerprint image recognition system to the fingerprint image comparison unit 140. That is, through the key input unit 150, the user may input user identification information, which is information indicating that the person currently attempting to perform user authentication, is his or her fingerprint to the fingerprint image comparison unit 140.

In addition, when the key input unit 150 registers the feature information of the user fingerprint image in the storage unit 130, the key input unit 150 may directly input the user identification information to the storage unit 130. That is, when the feature information of the fingerprint image of the user is stored in the storage unit 130 of the fingerprint image recognition system and is registered in the fingerprint image recognition system, the key input unit 150 stores the user identification information as the fingerprint image comparison unit ( It may be input to the storage 130 through the 140, or may be directly input to the storage 130.

Furthermore, when the key input unit 150 directly inputs user identification information to the storage unit 130, the storage unit 130 compares the fingerprint image by extracting feature information of the corresponding fingerprint image through the user identification information. Output to the unit 140. Also, in this case, the fingerprint image comparing unit 140 compares the feature information of the fingerprint image received from the storage unit 130 with the feature information of the fingerprint image received from the fingerprint image feature information extracting unit 120 to authenticate the user. Do this.

The storage unit 130 registers the feature information of the fingerprint image and the fingerprint image comparison unit 140 or the key received from the fingerprint image feature information extraction unit 120 when registering the feature information of the user fingerprint image in the storage unit 130. The user identification information received from the input unit 150 is stored together.

The fingerprint image comparing unit 140 compares the feature information of the fingerprint image inputted to the fingerprint image feature information extracting unit 120 with the feature information of the fingerprint image previously stored in the storage unit 130 to perform user authentication.

Here, the method for receiving the feature information of the pre-stored fingerprint image in the storage unit 130 in the fingerprint image comparison unit 140 is first, the feature information of the pre-stored fingerprint image directly from the storage unit 130 as described above There is a way to receive input. Second, there is a method of receiving user identification information from the key input unit 150 and loading feature information of the fingerprint image of the user corresponding to the user identification information from the storage unit 130. Here, the feature information of the fingerprint image is called a template.

In the matching method in the fingerprint image comparing unit 140, first, the fingerprint information received from the fingerprint image feature information extractor 120 and the feature information (or, also referred to as an “original template”) of the fingerprint image previously stored in the storage unit 130. Rotation, translation, and scaling of the feature information (also referred to as a "Comparison template") of the image are determined so that the two fingerprint images have a common coordinate system, origin, and magnification. Perform the alignment stage. Next, it is determined whether they are obtained from the same fingerprint image through a matching process of comparing the sorted original template and the comparative template.

The interface unit 160 is connected to the fingerprint image comparing unit 140 and plays a role of exchanging information generated in a fingerprint image matching process with an external device.

FIG. 2 is a block diagram illustrating in detail the fingerprint image quality classifying unit of FIG. 1. Referring to FIG. 2, the fingerprint image quality classification unit 110 includes a division unit 111, a quality classification parameter extraction unit 112, a normalization unit 113, a vector unit 114, and a quality determination value acquisition unit 115. And a quality determination unit 116.

The divider 111 divides the fingerprint image input from the fingerprint image acquirer 110 into a plurality of blocks. The divider 111 divides the block only for the fingerprint area when the fingerprint image inputted from the fingerprint image acquisition unit 100 knows the segmented image of the fingerprint area and the background area. However, since the background area is determined to be a block having poor quality even if the fingerprint area and the background area are not separated, the technical idea of the present invention is not significantly changed in determining the quality of the entire fingerprint image. It can be said that it belongs to a fact known to a person with ordinary knowledge.

The quality classification parameter calculator 112 calculates a fingerprint image quality classification parameter for each of the plurality of blocks divided by the divider 111. The fingerprint image quality classification parameter includes at least one of a directional control value, a ridge gradient distribution value, and a Fourier spectrum value for each of the plurality of divided blocks.

In the exemplary embodiment of the present invention, the fingerprint image quality classification parameter is described as including at least one of a directional control value, a ridge gradient distribution value, and a Fourier spectrum value for each of the plurality of divided blocks. In another embodiment of the present invention, various parameters that may classify the quality through the fingerprint image may be used.

More specifically, the calculation of the fingerprint image quality classification parameter will be described below.

The normalization unit 113 normalizes the fingerprint image quality classification parameter for each of the plurality of divided blocks received from the quality classification parameter calculator 112. As described above, the normalization is performed through the normalization unit 113 because the fingerprint image quality classification parameter values calculated for each of the plurality of divided blocks have different ranges.

The vector unit 114 vectorizes normalized fingerprint image quality parameters for each of the plurality of divided blocks received from the normalization unit 113.

The quality determination value obtaining unit 115 inputs a vectorized fingerprint image quality parameter for each of the plurality of divided blocks received from the vector unit 114 to a predetermined quality classifier to obtain a quality determination value. Here, the predetermined quality classifier previously learned is the K-means cluster classifier. The K-means clustering algorithm used in the K-means cluster classifier is introduced in the field of pattern recognition, and it is a representative example that learns in advance and presents the value of one of the learned categories for the unknown input based on the learning result. One type of supervised learning algorithm.

In an embodiment of the present invention, the K-means cluster classifier is described as a predetermined quality classifier. However, the exemplary embodiment is not limited thereto.

The quality determining unit 116 sets representative values for the plurality of quality determination values obtained by the quality determination value obtaining unit 115 to determine the quality of the fingerprint image input from the fingerprint image obtaining unit 100.

More specifically, the quality determination unit 116 may determine the quality determination value having the maximum frequency among the plurality of quality determination values as the quality of the fingerprint image input from the fingerprint image acquisition unit 100.

In the exemplary embodiment of the present invention, the quality determination value having the maximum frequency is determined as the quality of the fingerprint image input from the fingerprint image acquisition unit 100. However, in another embodiment of the present invention, the fingerprint image is determined in a different manner. The quality of the fingerprint image input from the acquirer 100 may be determined.

The fingerprint image classification unit 110 configured as described above will be described in more detail.

The dividing unit 111 divides the fingerprint image input from the fingerprint image obtaining unit 100 into H rectangular blocks each having N pixels horizontally and vertically. Here, the size of the block divided by the divider 111 is applied to the horizontal, vertical N pixels in the form of a square, but the present invention is not limited to this may have a free horizontal and vertical size and the square or It is not limited to the number of rectangular blocks.

The quality classification parameter calculator 112 calculates a directional control value, a ridge gradient distribution value, and a Fourier spectrum value for each of the H rectangular blocks divided by the divider 111.

Directional contrast is a parameter representing the pixel value contrast between the ridge and the valley according to the flow direction of the ridge in the fingerprint image. In general, in defining the quality of a fingerprint image, the fingerprint image with clear ridges and valleys and the difference is defined as a fingerprint image with good quality. The directional control value is calculated through the following equation.

여기서,

and

here,

는 융선의 흐름 방향에 따른 융선의 화소값이고

는 상기 융선에 대해 수직인 골의 화소값이다.

중에서 가장 큰 값이

이므로,

는 각 융선별 화소값을 나타낸다. 상기 (i,j)는 i행, j열의 픽셀을 나타내며, 상기 P(Pixel)는 픽셀의 화소값을 말한다. k는 1에서 8까지의 변수값이다.
융선 변화도 분포값은 화소값들의 변화도(Gradient)로써 융선의 방향과 그 방향으로의 분포 강도를 알 수 있는 파라미터이다. 융선 변화도 분포값은 아래 수학식 2와 같이 구하여진다. 수학식 2에서 구해진 융선 변화도 분포값은 지문영상의 융선 방향을 따라 변화도 분포가 얼마나 강하게 나타나는지를 알 수 있는 파라미터이다.D is a directional contrast,

Is the pixel value of the ridge according to the flow direction of the ridge

Is the pixel value of the valley perpendicular to the ridge.

The largest value

Because of,

Denotes a pixel value for each ridge. (I, j) represents pixels in rows i and j, and P (Pixel) denotes pixel values of pixels. k is a variable value from 1 to 8.
The ridge change distribution value is a parameter that shows the direction of the ridge and the distribution intensity in the direction as the gradient of pixel values. The ridge gradient distribution value is calculated as in Equation 2 below. The ridge gradient distribution value obtained in Equation 2 is a parameter that shows how strongly the gradient distribution appears along the ridge direction of the fingerprint image.

, From here,

,

,

,

L is a ridge change degree distribution value to be calculated, P _ij (Pixel) is a pixel value of pixels i, j columns, and dx, dy represents a change value of the pixel value.
Fourier spectral values are calculated using Fourier Spectrum and Fourier transform techniques belong to known techniques. If the fingerprint image ridge is looked at the pixel values along this axis along the orthogonal axis of the ridge direction, it can be estimated as a sine wave form. If the energy of this waveform is large, it can be judged that the quality of the fingerprint image, which is close to the sine wave, is a good quality with a clear distinction between the ridge and the valley. The Fourier spectral value can be calculated using Equation 3. In Equation 3, the ideal Fourier Domain (RFD) is defined as a stable region that does not include the highest frequency band and the lowest frequency band.

는

에서의 푸리에 스펙트럼(Fourier spectrum)값을 뜻하며, 여기에서

와

는 주파수 도메인(frequency domain)에서의 좌표값을 나타낸다.

에서의 Re는 복소수의 실수부(Real part)를, Im은 허수부(Imaginary part)를 나타낸다. 수학식 3에서 "Reasonable Fourier domain"은 푸리에 스펙트럼 결과에 적용하는 밴드패스 필터링 영역을 의미하는데, 그 값은 해당 지문영상 블록 내에 존재하는 지문 융선들의 융선간 거리에 따라 다르게 나타난다.In equation (3)

Is

Fourier spectrum at, where

Wow

Denotes a coordinate value in the frequency domain.

Re denotes a real part of a complex number, and Im denotes an imaginary part. In Equation 3, "Reasonable Fourier domain" means a bandpass filtering region applied to a Fourier spectrum result, and the value is different depending on the ridge distance of the fingerprint ridges present in the fingerprint image block.

The normalization unit 113 has three different quality classification parameters calculated for each of the H rectangular blocks, and normalizes them to a constant value. More specifically, the normalizer 113 normalizes three fingerprint image quality classification parameters from 0 to 100.

The vector unit 114 vectorizes the quality classification parameters normalized from 0 to 100 in the normalization unit 113 into a 1 * 3 vector.

The quality classification value obtaining unit 115 inputs a fingerprint image quality parameter vectorized as a 1 * 3 vector into each K-means cluster classifier for each of the H rectangular blocks input from the vector unit 114 to the previously learned K-means cluster classifier. Acquire. As described above, in one embodiment of the present invention, the K-means cluster classifier is described, but in another embodiment of the present invention, the present invention is not limited thereto.

More specifically, an example of the quality determination value in the previously learned K-means cluster classifier is as follows. The method proposed in the present invention divides the quality of the fingerprint image into four types, and the K-means cluster classifier divides the distribution of the experimental samples for these four quality characteristics into a mean value and a variance value. Expressed in statistical figures. That is, in the present invention, a cluster is generated using three quality classification parameters including a directional control value, a ridge gradient distribution value, and a Fourier spectrum value. Then, a mean vector and a covariance matrix are obtained for each cluster of four different qualities. When the sample image to be measured is input, the directional control value, the ridge distribution value, and the Fourier spectral value obtained from the input image are judged based on the mean and variance values of the four quality clusters. The high quality cluster is determined as a fingerprint image quality determination value of the input image. Here, there may be various ways of determining similarity with the cluster, and typically, equations based on Euclidean distance may be used.

The quality determination unit 116 converts the quality determination value of the maximum frequency by the majority vote among the plurality of quality determination values received from the quality determination value acquisition unit 115 as the quality of the fingerprint image input from the fingerprint image acquisition unit 100. Decide

In the exemplary embodiment of the present invention, three fingerprint image quality classification parameters and four fingerprint image quality determination values are shown. However, the number of fingerprint image quality classification parameters may use various measures in addition to the above parameters, and the fingerprint image quality determination value may be further subdivided unlike the above.

3 is a flowchart illustrating a quality classification method of a fingerprint image according to the present invention. Referring to FIG. 3, first, a fingerprint image is obtained from the fingerprint image acquisition unit 100 (S300).

Next, the fingerprint image obtained in step S300 is divided into a plurality of blocks (S310).

Next, a plurality of fingerprint image quality classification parameters are calculated for each of the plurality of blocks divided in step S310 (S320). The fingerprint image quality classification parameter may include at least one of a directional control value, a ridge gradient distribution value, and a Fourier spectrum value for each of the plurality of divided blocks. More specifically, referring to FIG. 2 for calculating a specific fingerprint image quality classification parameter.

In the exemplary embodiment of the present invention, the fingerprint image quality classification parameter is described as including at least one of a directional control value, a ridge gradient distribution value, and a Fourier spectrum value for each of the plurality of divided blocks. In another embodiment of the present invention, various parameters that may classify the quality through the fingerprint image may be used.

Next, the fingerprint image quality classification parameter for each of the plurality of blocks calculated in step S320 is normalized (S330). As described above, the reason for performing normalization is to match the fingerprint image quality classification parameter values calculated for each of the plurality of divided blocks because they have different ranges.

Next, the normalized fingerprint image quality parameter is vectorized for each of the plurality of blocks divided in step S330 (S340).

Next, in step S340, the vectorized fingerprint image quality classification parameter is input to a previously learned quality classifier to obtain a quality determination value for each of the plurality of divided blocks (S350). Here, the predetermined quality classifier previously learned is the K-means cluster classifier. The K-means clustering algorithm used in the K-means cluster classifier is introduced in the field of pattern recognition, and it is a pattern that learns in advance and presents the value of one of the learned categories for an unknown input based on the learning result. One kind of recognition algorithm. In an embodiment of the present invention, the K-means cluster classifier is described as a predetermined quality classifier, but is not limited thereto in another embodiment of the present invention. Reference will be made to FIG. 2 regarding the K-means cluster classifier to obtain a quality judgment value.

Next, a representative value for the plurality of quality determination values obtained in step S350 is set to determine the quality of the fingerprint image acquired in step S300 (S360). More specifically, step S360 may determine the quality determination value of the maximum frequency by majority vote among the plurality of input quality determination values as the quality of the fingerprint image obtained in step S300.

Parts not described in FIG. 3 will be described with reference to FIGS. 1 and 2.

4 illustrates step S320 of FIG. 3 in more detail. 4 illustrates step S320 of FIG. 3 in more detail, calculating a direction contrast value for each of the plurality of blocks divided by the fingerprint image quality classification parameter (S321), and calculating a ridge gradient distribution value ( S322) and the step of calculating the Fourier spectral value (S323) can be seen.

In one embodiment of the present invention, the step of calculating the fingerprint image quality classification parameter may include calculating a directional contrast value for each of the plurality of divided blocks (S321), calculating a ridge gradient distribution value (S322), and a Fourier. It is described as including at least one of the step of calculating the spectral value (S323), in another embodiment of the present invention is not limited to this by calculating various parameters that can classify the quality through the fingerprint image Can also be used.

In addition, with reference to the calculation of the directional control value, the calculation of the ridge gradient distribution value, and the Fourier spectral value, the reference to FIG.

5A to 5D illustrate four types of quality determined by the K-means cluster classifier using three quality classification parameters of the fingerprint images inputted in FIGS. 2 and 3. It was classified as good, dry, moist and bad. FIG. 5A illustrates a good quality classification value of a fingerprint image, in which the ridge is clear and the pixel value difference between the ridge and the valley is clear. In FIG. 5B, since the quality classification value of the fingerprint image is dry, most ridge structures can be identified, but the image is not clear. 5C is an image in which the quality classification value of the fingerprint image is moist, so that two parallel ridges are not easily separated and viewed. 5D is an image in which the quality classification value of the fingerprint image is bad and the ridge structure is completely damaged. This quality classification may determine whether it can be used in the fingerprint recognition step by using the input fingerprint image. In addition, even if the image is of usable quality as a whole, the entire image may have various quality blocks in a non-homogeneous state, and in this case, the image processing and feature extraction technique may be selectively applied in consideration of the quality state. .

FIG. 6 is a diagram illustrating a quality determination value of a fingerprint image in FIGS. 2 and 3 in a three-dimensional space. Referring to FIG. 6, in FIG. 2 and FIG. 3, the quality determination value of the fingerprint image is represented by the fingerprint image quality classification parameters as the respective axes of the 3D space.

The first quality judgment value, the second quality judgment value, the third quality judgment value and the fourth quality judgment value are divided and displayed in three-dimensional space.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD_ROM, magnetic tape, floppy disks, and optical data storage, and may also include those implemented in the form of carrier waves (eg, transmission over the Internet). . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, optimal embodiments have been disclosed in the drawings and the specification. Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not intended to limit the scope of the invention as defined in the claims or the claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, the present invention has the effect of more accurately determining the quality of the fingerprint image through the fingerprint image quality classification parameter for the quality classification of the fingerprint image that has been recognized as an auxiliary means for extracting the fingerprint image feature information. You can get it.

In addition, it is possible to obtain the characteristic information of the desired fingerprint image by repeatedly and stably utilizing the results similar to those classified by the expert, thereby obtaining performance optimization in the fingerprint image recognition system.

Claims (20)

(a) dividing the fingerprint image into a plurality of blocks; (b) calculating and vectorizing a plurality of fingerprint image quality classification parameters for each of the plurality of divided blocks; (c) inputting the vectorized fingerprint image quality classification parameter into a previously learned quality classifier to obtain a quality determination value for each of the plurality of divided blocks; And and setting a representative value from the quality determination value for each of the plurality of blocks to determine the quality of the fingerprint image. The method of claim 1, The pre-learned quality classifier is a K-means cluster classifier. The method of claim 1, The quality determination value is a quality classification method of a fingerprint image, characterized in that the value is divided into a plurality by the pre-learned quality classifier to classify the quality for the plurality of divided blocks. According to claim 1, wherein step (b), Calculating the plurality of fingerprint image quality parameters for each of the divided plurality of blocks; Normalizing the plurality of fingerprint image quality parameters; And Vectorizing the normalized plurality of fingerprint image quality parameters. The method of claim 1, Setting the representative value from the quality determination value for each of the plurality of blocks sets the quality determination value having the maximum frequency among the quality determination values for each of the plurality of blocks as the representative value. Quality classification method. The fingerprint image quality classification parameter of claim 1, A directional contrast value for the pixel value difference between the ridge and the valley according to the flow direction of the fingerprint image for each of the plurality of divided blocks; Fourier spectral values of pixel values with an orthogonal axis orthogonal to the ridge for each of the plurality of divided blocks; And And a ridge change degree distribution value for a change degree distribution of pixel values along the ridge for each of the plurality of divided blocks. The method of claim 6, wherein the quality determination value for each of the plurality of blocks, Define a plurality of quality clusters for learning the quality of the fingerprint image from the distribution pattern of the heat vector using the fingerprint image quality classification parameter, and obtain statistical characteristics of the plurality of quality clusters to determine the quality of each of the plurality of blocks. And determining the determination value as one of the plurality of quality clusters. A division unit for dividing an input fingerprint image into a plurality of blocks; A quality classification parameter calculator for calculating a fingerprint image quality classification parameter for each of the divided plurality of blocks; A vector unit vectorizing fingerprint image quality parameters for each of the plurality of divided blocks; A quality determination value obtaining unit obtaining the quality determination value for each of the plurality of divided blocks by inputting the vectorized fingerprint image quality parameter to a predetermined quality classifier; And And a quality determination unit that sets a representative value from the quality determination values for each of the plurality of divided blocks to determine the quality of the fingerprint image. The method of claim 8, The pre-learned quality classifier is a K-means cluster classifier. The method of claim 8, The quality determination value is a quality classification device of a fingerprint image, characterized in that the value is divided into a plurality by the pre-learned quality classifier to classify the quality for the plurality of divided blocks. The method of claim 8, And receiving the fingerprint image quality classification parameter calculated by the quality classification parameter calculator and normalizing the fingerprint image quality classification parameter and outputting the normalized fingerprint image quality classification parameter to the vector unit. The method of claim 8, Setting the representative value from the quality determination value for each of the plurality of blocks sets the quality determination value having the maximum frequency among the quality determination values for each of the plurality of blocks as the representative value. Quality sorting device. The method of claim 8, wherein the fingerprint image quality classification parameter, A directional contrast value for the pixel value difference between the ridge and the valley according to the flow direction of the fingerprint image for each of the plurality of divided blocks; Fourier spectral values of pixel values with an orthogonal axis orthogonal to the ridge for each of the plurality of divided blocks; And A ridge change distribution value for a change distribution of pixel values along the ridge for each of the divided plurality of blocks; Quality classification device of the fingerprint image, characterized in that it comprises at least one of. The method of claim 13, wherein the quality determination value for each of the plurality of blocks is: Define a plurality of quality clusters for learning the quality of the fingerprint image from the distribution pattern of the heat vector using the fingerprint image quality classification parameter, and obtain statistical characteristics of the plurality of quality clusters to determine the quality of each of the plurality of blocks. And a determination value is determined as one of the plurality of quality clusters. Fingerprint image acquisition unit for obtaining a fingerprint image of the user; A fingerprint image quality classifier classifying a quality of the obtained fingerprint image; A fingerprint image feature information extraction unit for extracting feature information of the fingerprint image which is a unique feature to be used for user authentication through an algorithm for extracting feature information of a fingerprint image based on the quality of the classified fingerprint images; And And a fingerprint image comparator for authenticating the user through the feature information of the fingerprint image extracted by the fingerprint image feature information extracting unit. The method of claim 15, wherein the fingerprint image quality classification unit A divider for dividing a fingerprint image input through the fingerprint acquirer into a plurality of blocks; A quality classification parameter calculator for calculating a fingerprint image quality classification parameter for each of the divided plurality of blocks; A vector unit vectorizing fingerprint image quality parameters for each of the plurality of divided blocks; A quality determination value obtaining unit obtaining the quality determination value for each of the plurality of divided blocks by inputting the vectorized fingerprint image quality parameter to a predetermined quality classifier; And And a quality determination unit that sets a representative value from the quality determination values for each of the plurality of divided blocks to determine the quality of the fingerprint image. The method of claim 15, wherein the fingerprint image acquisition unit A fingerprint input sensor scanning the fingerprint of the user; And And an analog / digital converter for converting the scanned user's fingerprint into a fingerprint image. The method of claim 15, And a storage unit which prestores feature information of a fingerprint image for users registered in the fingerprint image recognition system. The method of claim 18, And a key input unit for inputting the user identification information to the fingerprint image comparison unit. The fingerprint image comparing unit loads feature information of the fingerprint image stored in the storage unit through user identification information input from the key input unit, and extracts the feature information of the loaded fingerprint image and the fingerprint image feature information extracting unit. Fingerprint image recognition system, characterized in that to authenticate the user by comparing the feature information of the fingerprint image. The method of claim 18, And a key input unit for inputting the user identification information to the storage unit. The storage unit outputs user identification information input from the key input unit to the fingerprint image comparison unit, and the fingerprint image comparison unit is configured to extract the fingerprint image extracted from the fingerprint image comparison unit and the fingerprint image feature information extraction unit output from the storage unit. Fingerprint image recognition system characterized in that the authentication of the user by comparing the feature information.

Images