JP6075084B2 - Image conversion apparatus, image conversion method, biometric authentication apparatus, biometric authentication method, and computer program for image conversion - Google Patents

Description

  The present invention relates to, for example, an image conversion apparatus and image conversion method for converting information of a subject in an image into non-image information, a computer program for image conversion, and a biometric authentication apparatus and biometric authentication using such an image conversion method. Regarding the method.

  In recent years, biometric authentication technology for determining whether to authenticate an individual using biometric information such as fingerprints or palm prints has been developed. Biometric authentication technology is a device used by a specific individual such as a computer or mobile terminal from a large-scale system with a large number of registered users, such as an entrance / exit management system, a border control system, or a system using a national identification number. Until widely used.

  For example, when a fingerprint of any finger is used as biometric information, the biometric authentication device acquires a biometric image representing the fingerprint as an input biometric image. Then, the biometric authentication device collates the input biometric information that is the fingerprint of the user represented in the input biometric image with the registered biometric information that is the fingerprint represented in the biometric image of the registered user registered in advance. If the biometric authentication device determines that the input biometric information and the registered biometric information match based on the result of the verification process, the biometric authentication device authenticates the user as a registered user having a legitimate authority. The biometric authentication device permits an authenticated user to use a device in which the biometric authentication device is incorporated or another device connected to the biometric authentication device.

  In addition, a technique has been proposed in which feature information extracted from a biological image is vectorized and the vectorized feature information is used for collation (see, for example, Patent Document 1 and Non-Patent Document 1).

Japanese Patent Laid-Open No. 10-177650

Haiyun Xu et al., “Spectral Minutiae: A Fixed-length Representation of a Minutiae Set”, Computer Vision and Pattern Recognition Workshops, CVPRW '08. IEEE Computer Society Conference, 2008

  By calculating the feature information as a vector and using the feature information vector for matching, the calculation processing at the time of matching is simplified. Therefore, even when a large number of collation processes are executed as in the so-called 1: N authentication method, or even when collation processes are performed using a processor with low calculation capability, the time required for collation is reduced. In addition, since it is difficult to restore the original biometric image from the feature information vector, the biometric authentication device does not store the biometric image of the registered user, but stores only the feature information vector. Risk of leaking biometric information is reduced. In particular, by applying irreversible encryption processing to the feature information vector, even if a third party illegally obtains the encrypted feature information vector of the registered user, the third party is the registered user. It is difficult to reproduce the feature information vector.

  However, since the feature information vector has a relatively small amount of information, information unique to each individual included in the original biological information may not be sufficiently reflected in the feature information vector. For this reason, in the biometric authentication device that uses the feature information vector for the collation processing, the authentication accuracy may be insufficient. On the other hand, as the number of elements included in the feature information vector increases, it is possible to sufficiently reflect information unique to each individual. However, as the number of elements increases, the amount of calculation required for matching increases. .

  In view of this, in one aspect, an object of the present specification is to provide an image conversion apparatus capable of generating a vector that appropriately reflects the characteristics of a subject captured on an image while suppressing the number of elements.

  According to one embodiment, an image conversion device is provided. This image conversion apparatus includes a frequency converter that calculates a frequency spectrum image by converting an image of a subject from a spatial domain to a frequency domain, and a band selection that selects a frequency band that represents the characteristics of the subject from the frequency spectrum image. And a vector generation unit that generates a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band.

The objects and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims.
It should be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention as claimed.

  The image conversion apparatus disclosed in the present specification can generate a vector that appropriately reflects the characteristics of a subject captured on an image while suppressing the number of elements.

It is a schematic block diagram of the biometrics apparatus which is one Embodiment of an image conversion apparatus. It is a functional block diagram of a processing part. (A) is a figure which shows an example of the pattern pattern of a fingerprint, (b) is a figure which shows an example of the edge pattern of a fingerprint. It is a figure which shows the flow of the process for selecting the frequency band showing the characteristic of a fingerprint in an image conversion process. It is a figure which shows an example of the feature information vector. It is a figure which shows the operation | movement flowchart of an image conversion process. It is a figure which shows the operation | movement flowchart of a biometrics authentication process. It is a schematic structure figure of an example of a biometrics system by which an image conversion device by an embodiment or its modification was mounted.

Hereinafter, image conversion devices according to various embodiments will be described with reference to the drawings.
This image conversion apparatus extracts a subject region in which a subject is captured from a biological image representing biological information that is a subject acquired at the time of registration or verification of the biological information of the user, and represents frequency information of the subject. A frequency spectrum image is calculated. Then, the image conversion apparatus approximates the frequency spectrum of the subject in a specific direction extracted from the frequency spectrum image, and then selects a frequency band representing the characteristics of the subject, and the selected frequency band and the frequency band within the frequency band. Vectorize the frequency spectrum.

In the present embodiment, the image conversion apparatus uses the fingerprint of any finger as biometric information about the subject. However, the biological information as the subject may be other biological information that can be represented on the biological image, such as a palm print or a palm vein pattern.
Further, in this specification, the term “collation process” is used to indicate a process of calculating an index representing the degree of difference or the degree of similarity between the biometric information of the user and the biometric information of the registered user. In addition, the term “biometric authentication process” is used to indicate the entire authentication process including not only the verification process but also the process of determining whether to authenticate the user using the index obtained by the verification process. used.

  FIG. 1 is a schematic configuration diagram of a biometric authentication apparatus which is one embodiment of an image conversion apparatus. As illustrated in FIG. 1, the biometric authentication device 1 includes a display unit 2, an input unit 3, a biometric information acquisition unit 4, a storage unit 5, and a processing unit 6. The display unit 2, the input unit 3, and the biological information acquisition unit 4 may be provided separately from the housing in which the storage unit 5 and the processing unit 6 are accommodated. Or the display part 2, the input part 3, the biometric information acquisition part 4, the memory | storage part 5, and the process part 6 may be accommodated in one housing like what is called a notebook personal computer or a tablet-type terminal. The biometric authentication device 1 may further include a storage medium access device (not shown) that accesses a storage medium such as a magnetic disk, a semiconductor memory card, and an optical storage medium. The biometric authentication device 1 reads, for example, a computer program for biometric authentication processing executed on the processing unit 6 stored in the storage medium via the storage medium access device, and performs biometric authentication processing according to the computer program. May be executed.

  The biometric authentication device 1 performs biometric authentication processing by using the biometric image representing the fingerprint of the user's finger generated by the biometric information acquisition unit 4 and collating the fingerprint with the registered user's fingerprint. When the biometric authentication device 1 authenticates the user as one of the registered users as a result of the biometric authentication process, the biometric authentication device 1 permits the user to use the device on which the biometric authentication device 1 is mounted. Alternatively, the biometric authentication device 1 transmits a signal indicating that the user has been authenticated to another device (not shown), thereby permitting the user to use the other device.

  The display unit 2 includes a display device such as a liquid crystal display, for example. The display unit 2 then displays, for example, a message indicating a part (any finger) used for collation, or a guidance message for placing the part at a position where the biological information acquisition unit 4 can acquire a proper biological image. Display to users. Further, the display unit 2 displays a message representing the result of the biometric authentication process executed by the processing unit 6 or various information related to the application.

  The input unit 3 has a user interface such as a keyboard, a mouse, or a touch pad, for example. Then, the user name or command or data input by the user via the input unit 3 is passed to the processing unit 6. However, the input unit 3 may be omitted when the user does not need to input information other than the biometric information to the biometric authentication device 1.

The biological information acquisition unit 4 includes, for example, a fingerprint sensor using an area sensor. This fingerprint sensor can be, for example, a sensor that employs any one of an optical type, a capacitance type, an electric field type, and a thermal type. The biometric information acquisition unit 4 generates a biometric image representing a fingerprint by photographing the surface of the finger while the user places the finger on the sensor surface of the fingerprint sensor.
The biometric information acquisition unit 4 may have a slide type fingerprint sensor. In this case, the biometric information acquisition unit 4 sequentially generates partial images at predetermined time intervals while sliding a finger on the fingerprint sensor. In the partial image, a part of the fingerprint on the surface of the finger is copied. By connecting a plurality of partial images in order of generation time, a biometric image in which the entire fingerprint of the finger is captured is synthesized.

  The biometric information acquisition unit 4 passes the biometric image to the processing unit 6 every time a biometric image is generated.

The storage unit 5 includes, for example, a nonvolatile semiconductor memory and a volatile semiconductor memory. And the memory | storage part 5 memorize | stores the application program used with the biometrics authentication apparatus 1, the user name of at least one registered user, a user identification number, personal setting information, various data, etc. The storage unit 5 stores a program for executing biometric authentication processing. Further, for each registered user, the storage unit 5 stores a feature information vector representing the characteristics of a fingerprint of a specific finger, which is registered biometric information of the registered user, and a registered user such as the user name and user identification number of the registered user. Together with the identification information. Details of the feature information vector will be described later.
Further, the storage unit 5 may temporarily store the biological image received from the biological information acquisition unit 4.

  The processing unit 6 includes one or a plurality of processors and their peripheral circuits. Then, the processing unit 6 calculates a feature information vector representing the feature of the fingerprint shown in the biological image from the biological image acquired from the biological information acquisition unit 4. Then, the processing unit 6 executes biometric authentication processing or registration processing using the feature information vector.

  FIG. 2 is a functional block diagram of the processing unit 6. As illustrated in FIG. 2, the processing unit 6 includes a frequency conversion unit 11, a curve approximation unit 12, a band selection unit 13, a vector generation unit 14, a matching unit 15, an authentication determination unit 16, and a registration unit. 17. Each of these units included in the processing unit 6 is a functional module implemented by a computer program executed on a processor included in the processing unit 6. Alternatively, these units included in the processing unit 6 may be implemented in the biometric authentication device 1 as firmware.

  The frequency conversion unit 11, the curve approximation unit 12, the band selection unit 13, and the vector generation unit 14 are used in both the biometric authentication process and the registration process. The verification unit 15 and the authentication determination unit 16 are used in the biometric authentication process. On the other hand, the registration unit 17 is used in the registration process. Therefore, in the following, first, processing of each unit common to biometric authentication processing and registration processing will be described.

The frequency conversion unit 11 calculates a frequency spectrum image representing the frequency information of the subject shown in the biological image.
In the present embodiment, the frequency conversion unit 11 first extracts a subject area in which a fingerprint, which is a subject, appears on a biological image. The brightness in the subject area is different from the brightness in the background area where the subject is not shown. Therefore, the frequency conversion unit 11 divides the biological image into, for example, blocks of m in the horizontal direction × n in the vertical direction (where m and n are integers of 2 or more). Each block is preferably set to a size including a plurality of ridges. The frequency conversion unit 11 calculates, for each block, a statistical representative value such as an average value or a median value of pixel values as a threshold value. Then, for each block, the frequency conversion unit 11 compares the value of each pixel in the block with a threshold value to identify the pixel included in the subject area and the pixel included in the background area in the block. For example, when the pixel value included in the background region is higher than the pixel value included in the subject region, the frequency conversion unit 11 assigns, for each block, a pixel having a pixel value lower than the threshold value for the block to the subject region. And a pixel having a pixel value equal to or greater than the threshold value is included in the background area.

  Alternatively, when the luminance unevenness of the biological image is negligible, the frequency conversion unit 11 may set one threshold for the entire biological image. For example, the threshold value is determined by performing discriminant analysis on the average value or median value of the pixel values of the entire biological image, a preset fixed value (for example, 128), or the pixel value of the entire biological image. Can be a value. Similar to the above embodiment, the frequency conversion unit 11 determines whether or not each pixel is included in the subject region by comparing the value of each pixel of the biological image with the threshold value. For example, when the pixel value included in the background region is higher than the pixel value included in the subject region, the frequency converter 11 includes a pixel having a pixel value lower than the threshold value in the subject region, and a pixel equal to or higher than the threshold value. Include pixels with values in the background area.

The frequency conversion unit 11 replaces the value of each pixel included in the background area with a predetermined value (for example, 0, 128, 255, or an average value of pixel values in the subject area). Thereby, the frequency converter 11 prevents the frequency spectrum of the background region from being included in the frequency spectrum image.
The frequency conversion unit 11 calculates a frequency spectrum image representing the frequency information of the subject by converting a biological image in which each pixel in the background region is replaced with a predetermined value from the spatial region to the frequency region. Note that the frequency transform unit 11 can use, for example, discrete cosine transform or fast Fourier transform as the transform process from the spatial domain to the frequency domain. In the frequency spectrum image, the value of each pixel represents the power spectrum of the frequency corresponding to the pixel or the absolute value of the amplitude of the spectrum of the frequency corresponding to the pixel.

  In addition, when the luminance is substantially constant in a region other than the region in which the subject is captured on the biological image, the frequency conversion unit 11 directly performs discrete cosine transformation or fast Fourier transformation on the entire biological image, and the frequency A spectral image may be calculated. Thereby, the biometric authentication apparatus can reduce the amount of calculation required for calculating the frequency spectrum image.

  The frequency conversion unit 11 converts the frequency spectrum image into an image represented by a polar coordinate system with the zero frequency as the center and the horizontal direction as 0 degrees. Then, the frequency conversion unit 11 passes the frequency spectrum image expressed in the polar coordinate system to the curve approximation unit 12.

  The curve approximation unit 12 extracts a frequency spectrum in one or more directions from a frequency spectrum image expressed in a polar coordinate system. For example, the curve approximation unit 12 extracts frequency spectra in the horizontal direction, the direction that forms 45 ° with the horizontal direction, and the vertical direction. The curve approximating unit 12 approximates the frequency spectrum in each direction with a polynomial. By performing such approximation, the biometric authentication device 1 reduces the influence of the deformation of the frequency spectrum due to the state of the finger surface and the like when reading the biometric information, and obtains characteristic information of the subject from the frequency spectrum. It can be extracted.

In the present embodiment, a fingerprint that is a subject is represented as a pattern of ridges and valleys on a biological image.
FIG. 3A is a diagram illustrating an example of a fingerprint pattern, and FIG. 3B is a diagram illustrating an example of an edge pattern of a fingerprint. In the image 300 shown in FIG. 3A, the ridge 301 corresponds to a bright portion. In the image 310 shown in FIG. 3B, the edge 311 between the ridge line and the valley line is represented by a bright line.
The inventor has primarily contributed to the frequency spectrum of the fingerprint, mainly the ridge and valley pattern, as shown in image 300, and the edge between the ridge and valley, as shown in image 310. I got the knowledge that it was a pattern.

With respect to the pattern of ridges and valleys, as shown in the enlarged view 302a of the partial region 302 in the image 300, the ridges and valleys have a certain width, and between the pixels in the ridges. And the difference in brightness of each pixel in the valley are relatively small. Therefore, in the frequency spectrum, the intensity of the frequency spectrum increases in a relatively low frequency band corresponding to the ridge interval.
On the other hand, as shown in the enlarged view 312a of the partial region 312 in the image 310, the edge 311 between the ridge line and the valley line is narrower than the ridge line and the valley line, and in the direction intersecting the edge near the edge. The pixel value changes rapidly along. Therefore, on the frequency spectrum, the frequency band corresponding to the edge pattern between the ridge and the valley is higher than the frequency band corresponding to the pattern of the ridge and the valley.

Thus, the frequency spectrum of the fingerprint is mainly in two frequency bands, a frequency band corresponding to the pattern pattern of ridges and valleys, and a frequency band corresponding to the edge pattern between the ridges and valleys. It has a bimodal shape that increases its strength. Therefore, the biometric authentication device 1 converts only the information representing the frequency spectra of these two frequency bands into a feature information vector, thereby suppressing the lack of information to the extent that sufficient authentication accuracy can be obtained, and the feature information vector. The size can be reduced.
The curve approximation unit 12 can express such a bimodal shape by using, for example, a fourth or higher order polynomial, for example, a fourth or sixth order polynomial, for example, the frequency spectrum in each direction by the least square method. Approximate the curve.

The curve approximation unit 12 may approximate the frequency spectrum in each direction using a Fourier series or spline interpolation instead of using a polynomial.
Or the curve approximation part 12 calculates | requires the envelope which connects the maximum point of a frequency spectrum, and may curve approximate a frequency spectrum with the envelope.
The curve approximation unit 12 passes the frequency spectrum in each direction approximated to the curve to the band selection unit 13.

  The band selection unit 13 selects a frequency band representing the characteristics of the subject from the frequency spectrum approximated by a curve in each direction. In the frequency band representing the characteristics of the subject, generally, frequency components corresponding to the pattern of the subject are included, so the intensity of the spectrum is relatively high. Therefore, in this embodiment, the inflection point of the approximate curve is detected by calculating the second derivative of the approximate curve of the frequency spectrum for each direction. And when the second derivative between two adjacent inflection points becomes negative, that is, when the approximate curve is convex upward, the band selection unit 13 determines the frequency band between the two inflection points, The frequency band represents the characteristics of the subject. In the present embodiment, since the subject is a fingerprint, the approximate curve of the frequency spectrum has bimodality. Therefore, two frequency bands are selected. Of these two frequency bands, the lower frequency band corresponds to the pattern of ridges and valleys, while the higher frequency band corresponds to the edge pattern between the ridges and valleys.

  In the present embodiment, the contrast between the ridges and the valleys varies depending on the dry state of the finger surface during fingerprint reading. Therefore, the spectrum intensity of each frequency in the frequency spectrum of the fingerprint also varies. However, the position of the inflection point in the approximate curve of the frequency spectrum is not significantly affected by the absolute value of the spectrum intensity. In this example, since the frequency spectrum of the fingerprint can be approximated by a polynomial of a relatively low order such as 4th order or 6th order, the position of the inflection point of the approximate curve is less affected by the accuracy of the approximation. Therefore, by setting the inflection point of the approximate curve as the upper limit or lower limit of the frequency band representing the characteristics of the subject, the band selection unit 13 can appropriately select the frequency band representing the characteristics of the subject.

  Note that the band selection unit 13 assigns a frequency band of a certain ratio (for example, 30% to 50%) in order from the highest spectrum intensity to the entire frequency spectrum for each direction according to the P tile method. You may select as a frequency band showing a characteristic. Alternatively, the band selection unit 13 may set a statistical representative value such as an average value or median value of spectrum intensity for each frequency, or a predetermined fixed value as a threshold value for the spectrum intensity. Then, the band selecting unit 13 may select a frequency band having a spectral intensity equal to or higher than the threshold as a frequency band representing the characteristics of the subject.

FIG. 4 is a diagram showing a flow of processing for selecting a frequency band representing the characteristics of a fingerprint as a subject in the image conversion processing.
When the biological image 400 is obtained, the processing unit 6 extracts a subject region from the biological image 400 and then performs a conversion process from the spatial region to the frequency region to obtain the frequency spectrum image 410. And the process part 6 calculates the frequency spectrum image 420 represented by the polar coordinate system by carrying out polar coordinate conversion of the frequency spectrum image 410. FIG. Then, the processing unit 6 extracts a frequency spectrum 430 corresponding to one or more specific directions specified in advance as indicated by an arrow 421 from the frequency spectrum image 420 expressed in the polar coordinate system. The processing unit 6 obtains an approximate curve 440 of the frequency spectrum 430 by performing curve approximation on the extracted frequency spectrum 430. And the process part 6 calculates | requires the inflection points 441-444 of the approximate curve 440, and based on the inflection point, the frequency band 451 equivalent to the pattern pattern of a ridge and a valley line, and between a ridge and a valley line A frequency band 452 corresponding to the edge pattern is selected.

  The band selection unit 13 notifies the vector generation unit 14 of the selected frequency band and the frequency spectrum within the frequency band for each preset direction.

  The vector generation unit 14 generates a feature information vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the frequency band for each preset direction.

  FIG. 5 is a diagram illustrating an example of the feature information vector. In this example, a feature information vector is generated for each selected frequency band. The feature information vector 500 is a frequency band feature information vector corresponding to the ridge and valley line pattern patterns, and the feature information vector 510 is a frequency band feature information vector corresponding to the edge pattern between the ridge and valley lines. It is. Each of the feature information vectors 500 and 510 includes, for example, a selected frequency band width 501, an offset 502 from the zero frequency to the selected frequency band, and a flag 503 indicating a pattern pattern or an edge pattern. including. Further, the feature information vectors 500 and 510 include, as elements, a normalized value 504 of the intensity of the frequency spectrum at each frequency obtained by sampling the selected frequency band equally at a predetermined sampling number, for example, 5 to 20. Including. Alternatively, the feature information vectors 500 and 510 may include normalized values of the intensity of the frequency spectrum at each frequency within the selected frequency band.

The width 501 of the selected frequency band is expressed by an integer part of a value obtained by multiplying a ratio of the width of the selected frequency band to the sampling frequency of the frequency spectrum image by a predetermined constant, for example, 255, for example. Is done. The selected frequency band offset 502 is, for example, a value obtained by dividing the lower limit frequency, the upper limit frequency, or the center frequency of the selected frequency band by the sampling frequency of the frequency spectrum image, and a predetermined constant, for example, 255. It is represented by the integer part of the value obtained by multiplication.
The flag 503 is represented by a 1-bit value. For example, the flag 503 is '0' for the frequency band corresponding to the ridge and valley line pattern, and the flag 503 is '1' for the frequency band corresponding to the edge pattern between the ridge and valley lines.
The normalized value 504 of the frequency spectrum intensity is a ratio of the difference between the spectral intensity of the sampled frequency and the minimum value of the approximate curve to the difference between the maximum value and the minimum value of the approximate curve of the frequency spectrum. For example, it is represented by the integer part of the value obtained by multiplying by 255.

FIG. 6 is an operation flowchart of image conversion processing executed by the processing unit 6.
When the processing unit 6 acquires a biometric image representing the user's fingerprint from the biometric information acquisition unit 4 (step S101), the frequency conversion unit 11 of the processing unit 6 captures the subject that has the fingerprint as a subject from the biometric image. An area is extracted (step S102). The frequency conversion unit 11 replaces the value of each pixel in the background area other than the subject area with a predetermined value, and then calculates a frequency spectrum image from the biological image (step S103). Furthermore, the frequency conversion unit 11 converts the frequency spectrum image into an image represented by a polar coordinate system (step S104).

  The curve approximation unit 12 of the processing unit 6 selects one or more frequency spectra in a specific direction from the frequency spectrum image expressed in the polar coordinate system, and performs curve approximation on the selected frequency spectrum (step S105).

  The band selection unit 13 of the processing unit 6 selects, as the frequency band representing the characteristics of the subject, the frequency band in which the second derivative is negative from the frequency spectrum approximated by the curve for each of the selected directions (step). S106). Then, the vector generation unit 14 of the processing unit 6 generates a feature information vector that includes an element that identifies the selected frequency band and an element that represents the frequency spectrum within the frequency band for each selected direction (step). S107). Then, the processing unit 6 ends the image conversion process.

(Biometric authentication process)
FIG. 7 is an operation flowchart of the biometric authentication process executed by the processing unit 6.
The processing unit 6 acquires a user name or a user identification number for specifying a registered user to be authenticated via the input unit 3 (step S201). Further, the processing unit 6 receives an input biometric image in which input biometric information that is biometric information of a user who wants to receive authentication is received from the biometric information acquisition unit 4. And the process part 6 calculates the input feature information vector which is a feature information vector showing the user's biometric information reflected in the input biometric image by performing an image conversion process with respect to the input biometric image (step S202). ).

  The collation unit 15 of the processing unit 6 receives a registered feature information vector, which is a feature information vector representing features of biometric information of a registered user specified by a user name or the like input via the input unit 3, from the storage unit 5. Read. Then, the collation unit 15 collates the input feature information vector with the registered feature information vector to calculate the similarity between the user's biometric information and the registered user's biometric information (step S203).

The matching unit 15 calculates the distance between the input feature information vector and the registered feature information vector for each direction and for each selected frequency band. The distance between these two vectors can be, for example, the L1 norm or L2 norm of each element of the two feature information vectors. Alternatively, the matching unit 15 identifies frequencies that match each other between the feature information vectors based on the offset and the width of the selected frequency band, and sets the L1 norm or L2 norm for the spectrum intensities of the matching frequencies to two. It may be a distance between feature information vectors. In the present embodiment, the matching unit 15 corresponds to the distance between two feature information vectors for the frequency band corresponding to the ridge and valley line pattern patterns and the edge pattern between the ridge and valley lines for each direction. The distance between two feature information vectors for the frequency band to be calculated is calculated. And the collation part 15 makes the sum total of the distance between each feature information vector a similarity. In this case, as the similarity value is smaller, the biometric information of the user and the biometric information of the registered user are more similar.
If the direction of the finger with respect to the sensor surface of the biometric information acquisition unit 4 may be different every time biometric information is read, the collation unit 15 determines the direction about the input feature information vector and the registered feature information vector. The total distance is calculated while changing the combination with the direction. And the collation part 15 is good also considering the minimum value of the sum total of the distance as a similarity.

  The collation unit 15 passes the similarity to the authentication determination unit 16 along with the user name or user identification number of the registered user.

  The authentication determination unit 16 determines whether or not the similarity is equal to or lower than an authentication determination threshold value (step S204). When the similarity is equal to or lower than the authentication determination threshold (step S204—Yes), the authentication determination unit 16 determines that the biometric information of the user matches the biometric information of the registered user. And the authentication determination part 16 authenticates a user as the registered user (step S205). When authenticating the user, the authentication determination unit 16 notifies the processing unit 6 of the authentication result. Then, the processing unit 6 permits the authenticated user to use the device on which the biometric authentication device 1 is mounted or the device to which the biometric authentication device 1 is connected.

On the other hand, when the similarity is larger than the authentication determination threshold value (step S204-No), it is determined that the biometric information of the user does not match the biometric information of the registered user. Therefore, the authentication determination unit 16 does not authenticate the user (step S206). Then, the authentication determination unit 16 notifies the processing unit 6 that the user is not authenticated. In this case, the processing unit 6 refuses that a user who has not been authenticated uses a device in which the biometric authentication device 1 is mounted or a device to which the biometric authentication device 1 is connected. The processing unit 6 may cause the display unit 2 to display a message indicating that the authentication has failed.
After step S205 or S206, the processing unit 6 ends the biometric authentication process.

  The authentication determination threshold is preferably set to a value such that the authentication determination unit 16 succeeds in authentication only when the registered user is a user. The authentication determination threshold is preferably set to a value that causes the authentication determination unit 16 to fail authentication when another person different from the registered user is the user.

  In addition, when the so-called 1: N authentication method is adopted, that is, when the registered user identification information such as the registered user's user name is not input, the matching unit 15 determines the similarity for each registered user. Ask. Then, the matching unit 15 selects a registered user with the minimum similarity. The verification unit 15 passes the minimum value of the similarity and the user name or user identification number of the registered user corresponding to the minimum value to the authentication determination unit 16. The authentication determination unit 16 authenticates the user as a registered user corresponding to the minimum value of the similarity if the minimum value of the similarity is equal to or less than the authentication determination threshold.

(registration process)
Also in the registration process, the processing unit 6 receives a biometric image in which the biometric information of the user to be registered is shown from the biometric information acquisition unit 4. And the process part 6 calculates the characteristic information vector showing the user's biometric information reflected in the biometric image by performing an image conversion process with respect to the biometric image.

  The registration unit 17 acquires the user name of the registration target user from the input unit 3. The registration unit 17 sets a user identification number that is uniquely set for the user. Then, the registration unit 17 stores the user name and user identification number of the user in the storage unit 5 together with the feature information vector output from the vector generation unit 14. As a result, the user is registered as a registered user who is permitted to use the device on which the biometric authentication device 1 is mounted.

  As described above, the biometric authentication device which is an example of the image conversion device generates a feature information vector from a frequency band having a high spectrum intensity and a frequency spectrum included in the frequency band. Therefore, the feature information vector can appropriately represent the feature of the subject. In addition, since this biometric authentication device includes only information related to a frequency band that is assumed to include subject information in the feature information vector, the size of the feature information vector can be suppressed.

  According to the modified example, the processing unit 6 may divide the frequency spectrum in each selected direction into a plurality of frequency bands and linearly approximate each of the frequency spectra, thereby approximating the frequency spectrum by a polygonal line. In this case, according to the P tile method, the band selection unit 13 may select a frequency band having a certain ratio in order from the higher spectrum intensity as the frequency band representing the characteristics of the subject with respect to the entire frequency spectrum. Good. Alternatively, the band selection unit 13 may set a statistical representative value such as an average value or a median value of the spectrum intensity for each frequency or a predetermined fixed value as a threshold for the spectrum intensity. Then, the band selecting unit 13 may select a frequency band having a spectral intensity equal to or higher than the threshold as a frequency band representing the characteristics of the subject.

  According to another modification, when the collation unit 15 performs collation processing between the user and each registered user according to the 1: N authentication method, the feature information vector is used to reduce the amount of computation at the time of collation. The registered user may be selected based on the simple similarity calculated using only a part of the registered user. Then, the collation unit 15 may collate the selected registered user with the user by using another part of the feature information vector.

  For example, the matching unit 15 calculates the distance between the feature information vector corresponding to the pattern of the ridge and valley lines of the user and the feature information vector corresponding to the pattern of the ridge and valley lines of the registered user. Alternatively, it may be calculated as a simple similarity. And the collation part 15 selects a predetermined number of registered users in an order from the one where a value of simple similarity is small. The predetermined number is set to, for example, 1/100 to 1/10 of the total number of registered users, or a fixed number (for example, 10). Alternatively, the collation unit 15 may select all registered users whose simple similarity values are equal to or less than a predetermined selection threshold. Then, for each of the selected registered users, the matching unit 15 includes a feature information vector corresponding to an edge pattern between the ridges and valleys of the user, and between the ridges and valleys of the selected registered user. The distance between the feature information vector corresponding to the edge pattern is calculated as the similarity. If the minimum value of the similarities calculated for each of the selected registered users is equal to or less than the authentication determination threshold value, the authentication determining unit 16 sets the user as a registered user corresponding to the minimum value of the similarity. Certify. On the other hand, if the minimum value of the similarity is larger than the authentication determination threshold, the authentication determination unit 16 does not authenticate the user.

  The matching unit 15 calculates the distance between the feature information vector corresponding to the edge pattern between the ridge and valley lines of the user and the feature information vector corresponding to the edge pattern between the ridge and valley lines of each registered user. It may be calculated as a simple similarity and a registered user may be selected based on the simple similarity. And the collation part 15 may collate with a user about the selected registration user using the characteristic information vector corresponded to the pattern pattern of a ridge and a valley.

  According to another modification, when the biometric authentication device performs the biometric authentication process by the 1: N authentication method, the biometric authentication apparatus performs the collation process based on the feature information vector and the collation process based on the feature point extracted from the biometric information. It may be used to select a registered user to perform.

In this case, for each registered user, minutiae such as ridge branch points and end points are extracted from the registered user's biometric image, and the position and type of each minutiae identify the registered user. It is stored in advance in the storage unit 5 together with information.
For this purpose, for example, when the registration process is executed, the processing unit 6 binarizes the value of each pixel of the registered user's biological image, and distinguishes between a pixel representing a ridge and a pixel representing a valley. The threshold value for binarization can be, for example, an average value of pixel values in the subject area. Next, the processing unit 6 performs a thinning process on a pixel having a pixel value corresponding to a ridge with respect to a binarized biological image, so that a line connecting pixels representing a ridge is, for example, 1 Thinning into lines with pixel width. Then, the processing unit 6 scans the thinned biological image using a plurality of mask patterns having a binary pattern corresponding to a branch point or an end point of the ridge, thereby matching with any mask pattern. The position on the biological image is detected. Then, the processing unit 6 sets the center pixel at the detected position as the minutiae, and sets the type of minutia (that is, the branch point or the end point) represented by the matched mask pattern as the type of the detected minutiae.

  In this modification, the matching unit 15 selects a predetermined number of registered users in order from the smallest similarity value calculated based on the feature information vector. Moreover, the collation part 15 detects a minutiar from a user's biometric image at the time of execution of a biometrics authentication process. Then, the collation unit 15 collates the biometric information of the selected registered user with the biometric information of the user by minutia matching.

  For example, the collation unit 15 aligns a noticed minutia among minutiae extracted from the user's biological image with any of the minutiae extracted from the selected registered user's biological image. And the collation part 15 calculates | requires the number of the minutia which matches the minutia extracted from the biometric image of the selected registration user among the minutia extracted from the biometric image of the user. The collation unit 15 determines that the two minutiae match if the distance between the two minutiae is, for example, equal to or less than the ridge interval. The collation unit 15 may determine that the two minutiae match only when the two minutiae types match.

The matching unit 15 obtains the number of matching minutiae while changing the set of minutiae to be aligned. Then, the matching unit 15 sets the ratio of the number of matching minutiae to the total number of minutiae extracted from the user's biological image as the similarity.
The collation unit 15 calculates the similarity by minutia matching for each selected registered user, and obtains the maximum value of the similarities.
The authentication determination unit 16 authenticates the user as a registered user corresponding to the maximum value if the maximum value of the similarity is greater than or equal to the authentication determination threshold, and if the maximum value of the similarity is less than the authentication determination threshold, Do not authenticate users.

  Furthermore, the image conversion apparatus and the image conversion method disclosed in the present specification execute a biometric authentication process between the user's biometric information and pre-registered biometric information in order for the user to perform some operation. Applicable to various devices or systems.

FIG. 8 is a schematic configuration diagram of an example of a biometric authentication system in which the image conversion apparatus according to each of the above embodiments or modifications thereof is mounted.
For example, the biometric authentication system 100 includes at least one terminal 110 and a server 120. The terminal 110 and the server 120 are connected via a wired or wireless communication network 130. In FIG. 6, among the constituent elements of the biometric authentication system 100, the constituent elements corresponding to any of the constituent elements of the biometric authentication apparatus 1 shown in FIG. 1 are referred to the constituent elements of the biometric authentication apparatus 1. The same reference numbers as the numbers are attached.

In this system, the terminal 110 is a mobile terminal such as a mobile phone or a tablet terminal, or a terminal that is fixedly installed, and includes a display unit 2, an input unit 3, and a biological information acquisition unit 4. Further, the terminal 110 includes a storage unit 21, an image acquisition control unit 22, and an interface unit 23.
The storage unit 21 includes, for example, a semiconductor memory circuit, and temporarily stores a biological image generated by the biological information acquisition unit 4. The image acquisition control unit 22 includes one or a plurality of processors and peripheral circuits thereof, controls each unit of the terminal 110, and executes various programs that operate on the terminal 110. Then, the image acquisition control unit 22 transmits the biometric image generated by the biometric information acquisition unit 4 to the server 120 via the interface unit 23 having an interface circuit for connecting the terminal 110 to the communication network 130. Further, the image acquisition control unit 22 may also transmit user identification information input via the input unit 3 to the server 120.

  The server 120 includes a storage unit 5, a processing unit 6, and an interface unit 24 having an interface circuit for connecting the server 120 to the communication network 130. The processing unit 6 of the server 120 uses the biometric image received via the interface unit 24 to realize the function of each unit included in the processing unit according to any one of the above embodiments or a modification thereof, thereby biometric authentication. Execute the process. Then, the server 120 returns a determination result as to whether or not the authentication is successful to the terminal 110 via the interface unit 24.

  Alternatively, the image acquisition control unit 22 of the terminal 110 performs processing related to the image conversion processing among the functions of the processing units according to the above embodiments, that is, a frequency conversion unit, a curve approximation unit, a band selection unit, and a vector generation unit. The process may be executed. In this case, the feature information vector extracted from the user's biological image and the user identification information may be transmitted from the terminal 110 to the server 120 to the server 120. On the other hand, the processing unit 6 of the server 120 executes only the processes of the collation unit, the authentication determination unit, and the registration unit among the functions of the processing units according to the above embodiments. Thereby, since the load of the server 120 is reduced, the biometric authentication system 100 can suppress the waiting time for the user even if a large number of biometric authentication processes are executed at the same time.

  A computer program having instructions for causing a computer to realize the functions of the processing units according to the above embodiments is provided in a form recorded in a recording medium such as a magnetic recording medium, an optical recording medium, or a nonvolatile semiconductor memory. Also good. Note that the computer-readable recording medium does not include a carrier wave.

  All examples and specific terms listed herein are intended for instructional purposes to help the reader understand the concepts contributed by the inventor to the present invention and the promotion of the technology. It should be construed that it is not limited to the construction of any example herein, such specific examples and conditions, with respect to showing the superiority and inferiority of the present invention. Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions and modifications can be made thereto without departing from the spirit and scope of the present invention.

The following supplementary notes are further disclosed regarding the embodiment described above and its modifications.
(Appendix 1)
A frequency converter that calculates a frequency spectrum image by converting an image of a subject from a spatial domain to a frequency domain;
A band selection unit that selects a frequency band representing the characteristics of the subject from the frequency spectrum image;
A vector generation unit that generates a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
An image conversion apparatus.
(Appendix 2)
Selecting a frequency spectrum along a specific direction from the frequency spectrum image, and further comprising a curve approximation unit that approximates the frequency spectrum of the selected direction by a curve;
The image conversion apparatus according to appendix 1, wherein the band selection unit selects a frequency band representing the characteristics of the subject from the frequency spectrum approximated by the curve.
(Appendix 3)
The image conversion apparatus according to appendix 2, wherein the band selection unit selects at least one frequency band in which a second derivative of the frequency spectrum approximated by the curve is negative as a frequency band representing the characteristics of the subject.
(Appendix 4)
The subject is a fingerprint;
The band selection unit selects the first and second frequency bands in which the second derivative of the frequency spectrum approximated by the curve is negative as frequency bands representing the characteristics of the subject,
The vector generation unit includes a first vector including an element that specifies the first frequency band and an element that represents a frequency spectrum in the first frequency band, and an element that specifies the second frequency band. The image conversion apparatus according to appendix 3, wherein a second vector including an element representing a frequency spectrum in the second frequency band is generated.
(Appendix 5)
The image conversion apparatus according to appendix 1, wherein the band selection unit selects a frequency band including a predetermined number of frequencies in descending order of the intensity of the frequency spectrum as a frequency band representing the characteristics of the subject.
(Appendix 6)
The frequency conversion unit identifies a subject area where the subject appears in the image and a background area where the subject does not appear, and after replacing the value of each pixel in the background area with a predetermined value, The image conversion device according to any one of appendices 1 to 5, wherein the image conversion device converts the region into the frequency region.
(Appendix 7)
A frequency spectrum image is calculated by converting the image of the subject from the spatial domain to the frequency domain,
From the frequency spectrum image, select a frequency band representing the characteristics of the subject,
Generating a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
An image conversion method.
(Appendix 8)
A frequency spectrum image is calculated by converting the image of the subject from the spatial domain to the frequency domain,
From the frequency spectrum image, select a frequency band representing the characteristics of the subject,
Generating a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
An image conversion computer program for causing a computer to execute the above.
(Appendix 9)
A storage unit that stores a registration vector including an element that specifies a frequency band that represents the characteristics of biometric information of a registered user, and an element that represents a frequency spectrum within the frequency band;
A frequency converter that calculates a frequency spectrum image by converting an image in which a user's biological information is reflected from a spatial domain to a frequency domain;
A band selection unit that selects a frequency band representing characteristics of the user's biological information from the frequency spectrum image;
A vector generation unit that generates a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
A collation unit for obtaining a similarity between the biometric information of the user and the biometric information of the registered user by collating the registered vector and the vector;
An authentication determination unit for determining whether to authenticate the user according to the similarity;
A biometric authentication device.
(Appendix 10)
The storage unit includes, for each of a plurality of registered users, an element that identifies a first frequency band representing characteristics of biometric information of the registered user as the registration vector, and a frequency within the first frequency band. A first registration vector including an element representing a spectrum; an element specifying a second frequency band higher than the first frequency band and representing characteristics of biometric information of the registered user; and the second A second registration vector including an element representing a frequency spectrum in the frequency band;
Selecting a frequency spectrum along a specific direction from the frequency spectrum image, and further comprising a curve approximation unit that approximates the frequency spectrum of the selected direction by a curve;
The band selection unit includes a third frequency band in which a second derivative of the frequency spectrum approximated by the curve is negative, and a second order of the frequency spectrum higher than the third frequency band and approximated by the curve. Selecting a fourth frequency band having a negative derivative as a frequency band representing the characteristics of the biological information,
The vector generation unit includes a third vector that includes an element that specifies the third frequency band, an element that represents a frequency spectrum in the third frequency band, and an element that specifies the fourth frequency band. And a fourth vector including an element representing a frequency spectrum in the fourth frequency band,
The verification unit selects a predetermined number of registered users in order from the one in which the first registered vector is similar to the third vector among the plurality of registered users,
The authentication determining unit uses the second registration vector of the selected registered user based on the similarity with respect to a second registration vector that is most similar to the fourth registration vector. The biometric authentication device according to appendix 9, wherein it is determined whether to authenticate a person.
(Appendix 11)
The frequency spectrum image is calculated by converting the image showing the user's biological information from the spatial domain to the frequency domain,
From the frequency spectrum image, select a frequency band representing the characteristics of the user's biological information,
Generating a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
By collating a registered vector including an element for specifying a frequency band representing characteristics of biometric information of a registered user and an element representing a frequency spectrum in the frequency band with the vector, the biometric information of the user Obtaining the similarity of biometric information of the registered user;
It is determined whether to authenticate the user according to the similarity.
A biometric authentication method.

1 Biometric authentication device (image conversion device)
DESCRIPTION OF SYMBOLS 2 Display part 3 Input part 4 Biometric information acquisition part 5 Memory | storage part 6 Processing part 11 Frequency conversion part 12 Curve approximation part 13 Band selection part 14 Vector generation part 15 Collation part 16 Authentication determination part 17 Registration part 100 Biometric authentication system 110 Terminal 120 Server 130 Communication network 21 Storage unit 22 Image acquisition control unit 23, 24 Interface unit

Claims (9)

A frequency converter that calculates a frequency spectrum image by converting an image of a subject from a spatial domain to a frequency domain;
A band selection unit that selects a frequency band representing the characteristics of the subject according to the intensity of each frequency band of the frequency spectrum along a specific direction from the frequency spectrum image;
A vector generation unit that generates a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
An image conversion apparatus. Wherein selecting a frequency spectrum along the specific direction from the frequency spectral images, further comprising a curve approximation unit for curve approximation a frequency spectrum of the selected direction,
The image conversion apparatus according to claim 1, wherein the band selection unit selects a frequency band representing the characteristics of the subject from the frequency spectrum approximated by the curve.   The image conversion apparatus according to claim 2, wherein the band selection unit selects at least one frequency band in which a second derivative of the frequency spectrum approximated by the curve is negative as a frequency band representing the characteristics of the subject. .   The frequency conversion unit identifies a subject area where the subject appears in the image and a background area where the subject does not appear, and after replacing the value of each pixel in the background area with a predetermined value, The image conversion apparatus according to any one of claims 1 to 3, wherein the image conversion device performs conversion from a region to a frequency region. A frequency spectrum image is calculated by converting the image of the subject from the spatial domain to the frequency domain,
From the frequency spectrum image, select a frequency band representing the characteristics of the subject according to the intensity for each frequency band of the frequency spectrum along a specific direction ,
Generating a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
An image conversion method. A frequency spectrum image is calculated by converting the image of the subject from the spatial domain to the frequency domain,
From the frequency spectrum image, select a frequency band representing the characteristics of the subject according to the intensity for each frequency band of the frequency spectrum along a specific direction ,
Generating a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
An image conversion computer program for causing a computer to execute the above. A storage unit that stores a registration vector including an element that specifies a frequency band that represents the characteristics of biometric information of a registered user, and an element that represents a frequency spectrum within the frequency band;
A frequency converter that calculates a frequency spectrum image by converting an image in which a user's biological information is reflected from a spatial domain to a frequency domain;
From the frequency spectrum image, a band selection unit that selects a frequency band representing the characteristics of the user's biological information according to the intensity of each frequency band of the frequency spectrum along a specific direction ;
A vector generation unit that generates a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
A collation unit for obtaining a similarity between the biometric information of the user and the biometric information of the registered user by collating the registered vector and the vector;
An authentication determination unit for determining whether to authenticate the user according to the similarity;
A biometric authentication device. The storage unit includes, for each of a plurality of registered users, an element that identifies a first frequency band representing characteristics of biometric information of the registered user as the registration vector, and a frequency within the first frequency band. A first registration vector including an element representing a spectrum; an element specifying a second frequency band higher than the first frequency band and representing characteristics of biometric information of the registered user; and the second A second registration vector including an element representing a frequency spectrum in the frequency band;
Wherein selecting a frequency spectrum along the specific direction from the frequency spectral images, further comprising a curve approximation unit for curve approximation a frequency spectrum of the selected direction,
The band selection unit includes a third frequency band in which a second derivative of the frequency spectrum approximated by the curve is negative, and a second order of the frequency spectrum higher than the third frequency band and approximated by the curve. Selecting a fourth frequency band having a negative derivative as a frequency band representing the characteristics of the biological information,
The vector generation unit includes a third vector that includes an element that specifies the third frequency band, an element that represents a frequency spectrum in the third frequency band, and an element that specifies the fourth frequency band. And a fourth vector including an element representing a frequency spectrum in the fourth frequency band,
The verification unit selects a predetermined number of registered users in order from the one in which the first registered vector is similar to the third vector among the plurality of registered users,
The authentication determining unit uses the second registration vector of the selected registered user based on the similarity with respect to a second registration vector that is most similar to the fourth registration vector. The biometric authentication device according to claim 7, wherein it is determined whether to authenticate a person. The frequency spectrum image is calculated by converting the image showing the user's biological information from the spatial domain to the frequency domain,
From the frequency spectrum image, select a frequency band representing the characteristics of the user's biological information according to the intensity of each frequency band of the frequency spectrum along a specific direction ,
Generating a vector including an element that identifies the selected frequency band and an element that represents a frequency spectrum within the selected frequency band;
By collating a registered vector including an element for specifying a frequency band representing characteristics of biometric information of a registered user and an element representing a frequency spectrum in the frequency band with the vector, the biometric information of the user Obtaining the similarity of biometric information of the registered user;
It is determined whether to authenticate the user according to the similarity.
A biometric authentication method.