WO2025197125A1 - Pseudo blood vessel pattern generation device and pseudo blood vessel pattern generation method - Google Patents

Abstract

Provided is a pseudo blood vessel pattern generation device 10, wherein a processor 11 generates a first random number on the basis of a first random seed, generates a second random number on the basis of a second random seed different from the first random seed, adds first noise based on the first random number to a gray image to thereby generate a first image, adds second noise based on the second random number to the gray image to thereby generate a second image, synthesizes the first and second images to thereby generate a third image, and generates a pseudo blood vessel pattern image by using the third image. The first random seed is associated with a management ID, and the second random seed is associated with a personal ID. The processor 11 generates the third image by combining the first image and the second image, both of which are based on a normal distribution, and sets the weight of the normal distribution when generating the third image. The processor generates a plurality of clusters, each of which is defined by a combination of the management ID and the weight, the plurality of clusters each including a plurality of pseudo blood vessel pattern images generated under a plurality of mutually different personal IDs.

Description

Pseudo-vascular pattern generating device and pseudo-vascular pattern generating method

This disclosure relates to a pseudo-vascular pattern generation device and a pseudo-vascular pattern generation method.

The development of biometric authentication algorithms and systems based on vascular patterns requires large amounts of vascular pattern data. Traditionally, biometric authentication algorithms and systems have been developed using vascular patterns (hereinafter sometimes referred to as "real vascular patterns") extracted from images of actual living bodies (hereinafter sometimes referred to as "photographed images"). However, actually photographing living bodies to obtain a large number of real vascular patterns requires a great deal of time and expense to collect real vascular patterns. Furthermore, when photographing actual living bodies, the real vascular patterns extracted from the photographed images become data that identifies individuals, making it difficult to store real vascular patterns in storage due to contracts with the subjects and the laws of each country. Therefore, there is a need for pseudo-vascular patterns that can replace real vascular patterns.

In response to this, methods are known for generating pseudo-vascular patterns based on mathematical models such as reaction-diffusion equations that use, for example, Drosophila wing patterns or Turing patterns.

However, the shape of the pseudo-vascular patterns generated based on mathematical models is far from that of real vascular patterns. Furthermore, since it is difficult to generate diverse patterns when generating pseudo-vascular patterns based on mathematical models, it is not practical to use pseudo-vascular patterns generated based on mathematical models to evaluate biometric authentication algorithms.

In contrast, there is prior art that generates a first image by adding white noise to a gray image, generates a second image by diffusing the white noise in the first image, generates a third image by applying a blood vessel enhancement filter to the second image, generates a fourth image by setting a region of interest in the third image, and generates a fifth image containing a pseudo-vascular pattern based on the fourth image.

International Publication No. 2023/119653

M. Satoh, Mathematical Life Sciences: Getting Started Now, ISBN: 4339067628, pp. 176-193, Corona Publishing, 2020. (Published: January 8, 2021)

However, with the above prior art, it is easy to weaken the correlation between pseudo-vascular patterns but difficult to strengthen them, so even with the above prior art, it is difficult to approximate the distribution of pseudo-vascular patterns to the distribution of actual biological vascular patterns, which have a relatively strong correlation.

This disclosure therefore proposes a technology that can generate pseudo-vascular patterns with correlations of various strengths.

The pseudo-vascular pattern generating device of the present disclosure includes a processor. The processor generates a first random number based on a first random number seed, generates a second random number based on a second random number seed different from the first random number seed, generates a first image by adding first noise based on the first random number to a gray image, generates a second image by adding second noise based on the second random number to the gray image, generates a third image by combining the first image and the second image, and generates a pseudo-vascular pattern image using the third image, which is an image including a pseudo-vascular pattern. The first random number seed is associated with an administration ID, and the second random number seed is associated with a personal ID. The processor generates the third image by combining the first image and the second image, both of which are based on a normal distribution. When generating the third image, the processor sets a weight for the normal distribution and generates a plurality of clusters, each defined by a combination of the administration ID and the weight, each including a plurality of the pseudo-vascular pattern images generated under a plurality of different personal IDs.

The disclosed technology makes it possible to generate pseudo-vascular patterns with correlations of various strengths.

FIG. 1 is a diagram showing an example of the configuration of a pseudo blood vessel pattern generation system according to the present disclosure. FIG. 2 is a diagram showing an example of a processing procedure in the pseudo blood vessel pattern generating device of the present disclosure. FIG. 3 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 4 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 5 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 6 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 7 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 8 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 9 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 10 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 11 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 12 is a diagram showing an example of an image in the process of generating a pseudo blood vessel pattern according to the present disclosure. FIG. 13 is a diagram showing an example of a pseudo blood vessel pattern image according to the present disclosure. FIG. 14 is a diagram illustrating an example of a cluster according to the present disclosure. FIG. 15 is a diagram showing the measurement results of FRR and FAR for a pseudo blood vessel pattern generated by the pseudo blood vessel pattern generating device of the present disclosure. FIG. 16 is a diagram showing the measurement results of FRR for the pseudo blood vessel pattern generated by the pseudo blood vessel pattern generating device of the present disclosure. FIG. 17 is a diagram showing the FAR measurement results for the pseudo blood vessel pattern generated by the pseudo blood vessel pattern generating device of the present disclosure.

Embodiments of the present disclosure will be described below with reference to the drawings. In the following embodiments, identical components will be designated by the same reference numerals.

[Example]
<Configuration of pseudo-vascular pattern generation system>
FIG. 1 is a diagram showing an example of the configuration of a pseudo blood vessel pattern generation system according to the present disclosure.

In FIG. 1, the pseudo vascular pattern generation system 1 includes a pseudo vascular pattern generation device 10, an input device 20, and a display 30. The input device 20 and the display 30 are connected to the pseudo vascular pattern generation device 10. Examples of the input device 20 include a pointing device such as a mouse, and a keyboard. An example of the display 30 is an LCD (Liquid Crystal Display).

The pseudo-vascular pattern generating device 10 has a processor 11 and a memory unit 12. Examples of the processor 11 include a CPU (Central Processing Unit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and an ASIC (Application Specific Integrated Circuit). Examples of the memory unit 12 include memory and storage. The pseudo-vascular pattern generating device 10 is realized, for example, by a computer.

<Processing of the pseudo blood vessel pattern generating device>
2 is a diagram showing an example of a processing procedure in the pseudo blood vessel pattern generating device of the present disclosure, and FIGS. 3 to 12 are diagrams showing examples of images in the pseudo blood vessel pattern generating process of the present disclosure.

In FIG. 2, in step S100, the processor 11 obtains a management ID (hereinafter sometimes referred to as "MID") from the storage unit 12. The management ID is input by the operator to the pseudo-vascular pattern generation device 10 using the input device 20 and is pre-stored in the storage unit 12.

Next, in step S105, the processor 11 initializes the value of the personal ID (hereinafter sometimes referred to as "PID") to "1".

Next, in step S110, the processor 11 retrieves from the storage unit 12 a gray image Ia (Figure 3) that has been previously stored in the storage unit 12. When the gradation value of each pixel forming the image is any value between 0 and 255, an image in which the gradation values of all pixels are, for example, 128, the intermediate value between 0 and 255, is previously stored in the storage unit 12 as the gray image Ia.

Next, in step S115, the processor 11 sets a first random number seed and a second random number seed. The first random number seed and the second random number seed are different from each other, and the first random number seed is associated with the management ID, and the second random number seed is associated with the personal ID.

For example, a plurality of mutually different first random number seeds corresponding one-to-one to each of a plurality of mutually different management IDs and a plurality of mutually different second random number seeds corresponding one-to-one to each of a plurality of mutually different personal IDs are pre-stored in the memory unit 12, and the processor 11 acquires from the memory unit 12 the first random number seed corresponding to the management ID acquired in step S100 and the second random number seed corresponding to the current value of the personal ID.

Also, for example, the processor 11 may use the management ID acquired in step S100 as the first random number seed, and the current value of the personal ID as the second random number seed.

Next, in step S120, the processor 11 generates random numbers that follow a first normal distribution N(μ ₁ , σ ₁ ² ) based on the first random number seed, and generates random numbers that follow a second normal distribution N(μ ₂ , σ ₂ ² ) based on the second random number seed. Hereinafter, random numbers generated based on the first random number seed may be referred to as "first random numbers," and random numbers generated based on the second random number seed may be referred to as "second random numbers." Also, in step S120, the processor 11 adds white noise based on the first random number (hereinafter may be referred to as "first noise") to all pixels of the grayscale image Ia, and adds white noise based on the second random number (hereinafter may be referred to as "second noise") to all pixels of the grayscale image Ia. Because the first random number seed and the second random number seed are different random number seeds, the first noise and the second noise are white noises with different noise states. This generates an image Ib1 (Figure 4) in which the first noise has been added to the gray image Ia (hereinafter referred to as the "first noise image"), and an image Ib2 (Figure 5) in which the second noise has been added to the gray image Ia (hereinafter referred to as the "second noise image").

Next, in step S125, the processor 11 combines the first noise image Ib1 and the second noise image Ib2. As a result, an image Ic (FIG. 6) in which the first noise image Ib1 and the second noise image Ib2 are combined (hereinafter, may be referred to as the "composite noise image") is generated. The composite noise image Ic is an image that follows a normal distribution N(μ,σ ² ).

Next, in step S130, the processor 11 diffuses the noise in the composite noise image Ic. For example, the processor 11 applies a Gaussian filter to the composite noise image Ic to generate an image Id (FIG. 7) in which the noise in the composite noise image Ic has been diffused (hereinafter, sometimes referred to as a "noise-diffused image").

Next, in step S135, the processor 11 emphasizes blood vessels in the noise diffusion image Id. The processor 11 applies a blood vessel emphasis filter, such as a Frangi filter, to the noise diffusion image Id to generate an image Ie (Figure 8) in which blood vessels are emphasized in the noise diffusion image Id (hereinafter sometimes referred to as a "blood vessel emphasis image").

Next, in step S140, processor 11 smooths vessel-enhanced image Ie. For example, processor 11 smooths vessel-enhanced image Ie using a Gaussian filter to generate an image If (Figure 9) in which vessel-enhanced image Ie has been smoothed (hereinafter, sometimes referred to as a "smoothed image"). Note that it is possible to omit the processing of step S140.

Next, in step S145, the processor 11 inverts the color of the smoothed image If. This generates an image Ig (Figure 10) in which the smoothed image If is inverted from negative to positive (hereinafter sometimes referred to as a "color-inverted image"). Note that if the processing of step S140 is omitted, in step S145 the processor 11 inverts the color of the blood vessel-enhanced image Ie to generate a color-inverted image Ig. Note that the processing of step S145 can also be omitted.

Next, in step S150, processor 11 sets a region of interest (ROI) in the color-inverted image Ig. This generates an image Ih (Figure 11) in which a region of interest has been set in the color-inverted image Ig (hereinafter sometimes referred to as the "image after region of interest setting"). Note that if the processing of step S145 is omitted, processor 11 sets a region of interest in the smoothed image If in step S150. Also, if the processing of steps S140 and S145 is omitted, processor 11 sets a region of interest in the blood vessel-enhanced image Ie in step S150.

Next, in step S155, the processor 11 performs a geometric transformation on the post-region-of-interest setting image Ih. This generates an image Ii (Figure 12) in which the post-region-of-interest setting image Ih has been geometrically transformed (hereinafter may be referred to as the "post-geometric transformation image"). In the post-geometric transformation image Ii, the image within the region of interest becomes an image BV containing a pseudo blood vessel pattern (hereinafter may be referred to as the "pseudo blood vessel pattern image"). Examples of geometric transformations performed on the post-region-of-interest setting image Ih include affine transformation and projective transformation. Note that the processing of step S155 can be omitted. If the processing of step S155 is omitted, the image within the region of interest in the post-region-of-interest setting image Ih becomes the pseudo blood vessel pattern image BV.

Next, in step S160, the processor 11 stores the geometrically transformed image Ii generated in step S155 in the storage unit 12.

Next, in step S165, processor 11 determines whether the PID value has reached a predetermined value N. If the PID value has not reached the predetermined value N (step S165: No), processing proceeds to step S170; if the PID value has reached the predetermined value N (step S165: Yes), the processing procedure ends.

In step S170, processor 11 increments the PID value by "1". After processing step S170, processing returns to step S110.

The operator can set the predetermined value N using the input device 20. The operator can also visually recognize the pseudo blood vessel pattern image BV using the display 30.

Here, the processor 11 changes the values of the first random number seed and second random number seed set in step S115 depending on the values of the MID and PID.

For example, when the MID value is "A" (MID=A), processor 11 sets the value of the first random number seed to "SA", when the MID value is "B" (MID=B), the value of the first random number seed to "SB", and when the MID value is "C" (MID=C), the value of the first random number seed to "SC". On the other hand, when the PID value is "1" (PID=1), processor 11 sets the value of the second random number seed to "S1", when the PID value is "2" (PID=2), the value of the second random number seed to "S2", and when the PID value is "3" (PID=3), the value of the second random number seed to "S3".

In this way, the value of the first random number seed changes depending on the value of MID, and the value of the second random number seed changes depending on the value of PID.

<Combining First Noise Image and Second Noise Image>
In order to generate a composite noise image Ic conforming to ^the normal distribution N( _μ , _σ2 ) by combining a first noise image Ib1 generated based on a first normal distribution N( _μ1 , _σ12 ) with a second noise image Ib2 generated based on a _second normal distribution N(μ2, ^σ22 ) (where _μ1 ≦ _μ2 and ^{σ12 ≦} _σ22 ), the processor 11 sets weights for the mean and variance of the normal distributions that can be generated from each of ^the first random number seed and the second random number seed as follows: Below, a method for setting a weight for the mean of the normal distribution (hereinafter sometimes referred to as the "mean weight") _wa (where 0≦ _wa ≦1) and a method for setting a weight for the variance of the normal distribution (hereinafter sometimes referred to as the "variance weight") _wb (where 0≦ _wb ≦1) will be explained separately.

<How to set the average weight>
If μ ₁ =0, μ ₂ =0, and μ≠0, it is difficult for the processor 11 to set the average weight w _a , so the processor 11 resets either μ ₁ or μ ₂ .

Furthermore, if _μ1 ≦μ≦ _μ2 , processor 11 sets an average weight _wa for _μ1 and _μ2 that satisfies the weighted average ( _{wa μ1} + (1- _wa ) _μ2 = μ). In particular, if _μ1 = μ = _μ2 , the weighted average ( _{wa μ1} + (1- _wa ) _μ2 = μ) holds for any average weight _wa , so processor 11 sets the average weight wa taking into account the influence of the first normal distribution N( _μ1 , _σ12 ) and the second normal distribution N( _μ2 , _σ22 ) generated from each of ^the first random number seed and ^the second random number _seed .

Furthermore, if μ< _μ1 or _μ2 <μ, there is no value within the domain of the average weight _wa that satisfies the weighted average ( _{wa μ1} + (1- _wa ) _μ2 = μ), so processor 11 resets either _μ1 or _μ2 to a value that satisfies the conditions μ1 _≦ μ and μ≦ _μ2 .

<How to set the distribution weight>
If σ ₁ ² ≦σ ² ≦σ ₂ ² , the processor 11 sets a variance _{weight w b that satisfies the weighted average (w b σ 1} ² ₊ ⁽ ₁ ⁻ w _b )σ ₂ ² =σ ² ) for σ _{1 2} and σ 2 2. In particular, if σ ₁ ² =σ ² =σ ₂ ² , the weighted average (w _b σ ₁ ² +(1−w _b )σ ₂ ² =σ ² ) holds for any variance weight w _b , so the processor 11 sets the variance weight w b taking into account the influence of the first normal distribution N(μ ₁ ,σ ₁ ² ) and the second normal distribution N(μ ₂ ,σ ₂ ² ) generated from the first random number seed and the second random number _seed , respectively.

Furthermore, if σ ² < σ ₁ ² or σ ₂ ² < σ ² , there is no value within the domain of the variance weight w _b that satisfies the weighted average (w _b σ ₁ ² + (1 - w _b )σ ₂ ² = σ ² ), so the processor 11 resets either σ ₁ ² or σ ₂ ² to a value that satisfies the conditions σ ₁ ² ≦ σ ² and σ ² ≦ σ ₂ ² .

In the following, the mean weight w _a and the variance weight w _b may be collectively referred to as the "normal distribution weight w."

<Pseudo-vascular pattern image>
FIG. 13 is a diagram showing an example of a pseudo blood vessel pattern image according to the present disclosure.

When the MID value obtained in step S100 is "A" (MID=A) and the predetermined value N in step S165 is set to "2" (N=2), as shown in FIG. 13, the first pseudo blood vessel pattern image BV1 to the tenth pseudo blood vessel pattern image BV10 are generated according to the magnitude of the normal distribution weight w (w=0.0, 0.2, 0.4, 0.6, 0.8). The first pseudo blood vessel pattern image BV1 to the tenth pseudo blood vessel pattern image BV10 are generated under MID=A. Furthermore, the first pseudo blood vessel pattern image BV1 to the fifth pseudo blood vessel pattern image BV5 are generated under PID=1, and the sixth pseudo blood vessel pattern image BV6 to the tenth pseudo blood vessel pattern image BV10 are generated under PID=2. Furthermore, the first pseudo blood vessel pattern image BV1 and the sixth pseudo blood vessel pattern image BV6 are generated under w = 0.0, the second pseudo blood vessel pattern image BV2 and the seventh pseudo blood vessel pattern image BV7 are generated under w = 0.2, the third pseudo blood vessel pattern image BV3 and the eighth pseudo blood vessel pattern image BV8 are generated under w = 0.4, the fourth pseudo blood vessel pattern image BV4 and the ninth pseudo blood vessel pattern image BV9 are generated under w = 0.6, and the fifth pseudo blood vessel pattern image BV5 and the tenth pseudo blood vessel pattern image BV10 are generated under w = 0.8.

According to the first pseudo blood vessel pattern image BV1 to the tenth pseudo blood vessel pattern image BV10 shown in Figure 13, the larger the normal distribution weight w, the more similar the pseudo blood vessel pattern images BV become between PID=1 and PID=2, that is, the greater the similarity (hereinafter sometimes referred to as "pattern image similarity") of the pseudo blood vessel pattern images BV between PID=1 and PID=2. For example, it can be seen that the pattern image similarity between the fifth pseudo blood vessel pattern image BV5 and the tenth pseudo blood vessel pattern image BV10 is greater than the pattern image similarity between the fourth pseudo blood vessel pattern image BV4 and the ninth pseudo blood vessel pattern image BV9; the pattern image similarity between the fourth pseudo blood vessel pattern image BV4 and the ninth pseudo blood vessel pattern image BV9 is greater than the pattern image similarity between the third pseudo blood vessel pattern image BV3 and the eighth pseudo blood vessel pattern image BV8; the pattern image similarity between the third pseudo blood vessel pattern image BV3 and the eighth pseudo blood vessel pattern image BV8 is greater than the pattern image similarity between the second pseudo blood vessel pattern image BV2 and the seventh pseudo blood vessel pattern image BV7; and the pattern image similarity between the second pseudo blood vessel pattern image BV2 and the seventh pseudo blood vessel pattern image BV7 is greater than the pattern image similarity between the first pseudo blood vessel pattern image BV1 and the sixth pseudo blood vessel pattern image BV6.

<Cluster generation>
FIG. 14 is a diagram illustrating an example of a cluster according to the present disclosure.

For example, when MID=A, the processor 11 sets the normal distribution weight w to w1 (w=w1) and w2 (w=w2) greater than w1, and generates six pseudo blood vessel pattern images BV by changing the PID to "1", "2", and "3" when w=w1 and changing the PID to "4", "5", and "6" when w=w2. This generates a first cluster CL1 defined by the combination of MID=A and w=w1, and a second cluster CL2 defined by the combination of MID=A and w=w2. The first cluster CL1 includes three pseudo blood vessel pattern images BV generated when MID=A, w=w1, and PID=1, 2, and 3, and the second cluster CL2 includes three pseudo blood vessel pattern images BV generated when MID=A, w=w2, and PID=4, 5, and 6. Because w2 > w1, the pattern image similarity between the three pseudo blood vessel pattern images BV belonging to the second cluster CL2 is greater than the pattern image similarity between the three pseudo blood vessel pattern images BV belonging to the first cluster CL1.

Furthermore, for example, the processor 11 generates three pseudo blood vessel pattern images BV by setting the normal distribution weight w to w1 (w=w1) when MID=B and changing the PID to "7", "8", and "9" when w=w1. This generates a third cluster CL3 defined by the combination of MID=B and w=w1. The three pseudo blood vessel pattern images BV generated when MID=B, w=w1, and PID=7, 8, and 9 belong to the third cluster CL3.

Here, the pseudo blood vessel pattern image belonging to the first cluster CL1 and the pseudo blood vessel pattern image belonging to the second cluster CL2 are generated from the same MID. Therefore, the first cluster CL1 and the second cluster CL2 have similar properties, resulting in a strong correlation between the first cluster CL1 and the second cluster CL2. On the other hand, the pseudo blood vessel pattern image belonging to the first cluster CL1 and the pseudo blood vessel pattern image belonging to the second cluster CL2 are generated from normal distribution weights w of different magnitudes, and therefore, as described above, the similarity between the pseudo blood vessel pattern images belonging to the first cluster CL1 differs from the similarity between the pseudo blood vessel pattern images belonging to the second cluster CL2. Therefore, by forming the first cluster CL1 and the second cluster CL2, it is possible to generate a wide variety of pseudo blood vessel patterns, including pseudo blood vessel patterns that are highly correlated with each other.

Furthermore, the pseudo blood vessel pattern image belonging to the first cluster CL1 and the pseudo blood vessel pattern image belonging to the third cluster CL3 are generated from different MIDs. As a result, the first cluster CL1 and the third cluster CL3 have different properties, and the correlation between the first cluster CL1 and the third cluster CL3 is weak. Therefore, by forming the third cluster CL3 in addition to the first cluster CL1 and the second cluster CL2, it is possible to generate even more diverse pseudo blood vessel patterns.

<False Acceptance Rate (FAR) and False Rejection Rate (FRR)>
Fig. 15 is a diagram showing the measurement results of FRR and FAR for a pseudo blood vessel pattern generated by the pseudo blood vessel pattern generation device of the present disclosure, Fig. 16 is a diagram showing the measurement results of FRR for a pseudo blood vessel pattern generated by the pseudo blood vessel pattern generation device of the present disclosure, and Fig. 17 is a diagram showing the measurement results of FAR for a pseudo blood vessel pattern generated by the pseudo blood vessel pattern generation device of the present disclosure.

When measuring the FRR and FAR shown in Figures 15, 16, and 17, the MID value was set to one value, the predetermined value N was set to "1000," and the PID value was changed in 1000 ways to generate 1000 pseudo blood vessel pattern images BV using the pseudo blood vessel pattern generating device 10. Three different geometric transformations were then performed on each pseudo blood vessel pattern image BV. Two sets of data (hereinafter sometimes referred to as "registration data") were used to register the pseudo blood vessel pattern images BV generated from each pseudo blood vessel pattern image BV after the three geometric transformations in a database, and one set of data (hereinafter sometimes referred to as "matching data") was used to compare the data registered in the database. This total of three data sets was used to generate 1000 pairs of pseudo blood vessel pattern images BV after the geometric transformations.

Figure 15 shows the FRR measurement results for 1,000 pseudo-vascular pattern image BV pairs after geometric transformation, where the registered data and matching data have the same MID and PID (i.e., 1,000 pairs without rearrangement), and the FAR measurement results for 999,000 pairs of 1,000 pseudo-vascular pattern image BV pairs after geometric transformation, where the registered data and matching data have the same MID and different PID. Figure 15 also shows the FRR and FAR measurement results when the normal distribution weight w is changed to 0.00, 0.20, 0.40, 0.60, and 0.80. In the measurement results shown in Figure 15, the FRR is 0% at or below approximately -0.18, regardless of the magnitude of the normal distribution weight w. On the other hand, when the normal distribution weight w is 0.00 or 0.20, the FAR is 0% at approximately -0.28 or greater, when the normal distribution weight w is 0.40, it is 0% at approximately -0.27 or greater, when the normal distribution weight w is 0.60, it is 0% at approximately -0.22 or greater, and when the normal distribution weight w is 0.80, it is 0% at approximately -0.16 or greater. Thus, the measurement results shown in Figure 15 show that while the FRR distribution hardly changes when the normal distribution weight w is changed, the FAR distribution shifts to the right in the figure as the normal distribution weight w is increased.

Furthermore, Figure 16 shows the measurement results of the mean and median of each FRR distribution when the normal distribution weight w is changed in increments of 0.02 from 0.00 to 0.96 under the same conditions as Figure 15. Furthermore, Figure 17 shows the measurement results of the mean and median of each FAR distribution when the normal distribution weight w is changed in increments of 0.02 from 0.00 to 0.96 under the same conditions as Figure 15. The measurement results shown in Figure 16 show that the mean and median of FRR change very little even when the normal distribution weight w is changed. On the other hand, the measurement results shown in Figure 17 show that the mean and median of FAR increase as the normal distribution weight w is increased.

The measurement results shown in Figures 15, 16, and 17 show that the pseudo blood vessel patterns contained in the pseudo blood vessel pattern image generated by the pseudo blood vessel pattern generation device 10 are diverse and meet the ideal conditions for FAR and FRR, which are key indices used in evaluating biometric authentication algorithms.

Furthermore, the measurement results shown in Figures 15, 16, and 17 show that the FAR of the pseudo blood vessel patterns contained in the pseudo blood vessel pattern image generated by the pseudo blood vessel pattern generation device 10 can be easily adjusted by changing the normal distribution weight w.

The above explains the examples.

As described above, the pseudo vascular pattern generation device (pseudo vascular pattern generation device 10 in the embodiment) of the present disclosure includes a processor (processor 11 in the embodiment). The processor generates a first random number based on a first random number seed, generates a second random number based on a second random number seed different from the first random number seed, generates a first image (first noise image Ib1 in the embodiment) by adding first noise based on the first random number to a gray image (gray image Ia in the embodiment), and generates a second image (second noise image Ib2 in the embodiment) by adding second noise based on the second random number to the gray image. The processor also generates a third image (composite noise image Ic in the embodiment) by combining the first and second images, and generates a pseudo vascular pattern image (pseudo vascular pattern image BV in the embodiment) using the third image.

Here, the first random number seed is associated with the management ID, and the second random number seed is associated with the personal ID. The processor also generates a third image by combining the first image and the second image, both of which are based on normal distributions. The processor also sets the weight of the normal distribution when generating the third image. The processor then generates a plurality of clusters, each defined by a combination of the management ID and the weight, each containing a plurality of pseudo blood vessel pattern images generated under a plurality of mutually different personal IDs.

By generating multiple clusters like this, each cluster can contain multiple pseudo blood vessel patterns that are highly correlated with each other. Furthermore, while the correlation between pseudo blood vessel patterns can be strengthened between clusters with the same management ID but different weights, the correlation between pseudo blood vessel patterns can be weakened between clusters with different management IDs. Therefore, by generating multiple clusters like this, pseudo blood vessel patterns with correlations of various strengths can be generated.

1 pseudo-vascular pattern generation system 10 pseudo-vascular pattern generation device 11 processor 12 storage unit

Claims (4)

a processor that generates a first random number based on a first random number seed, generates a second random number based on a second random number seed different from the first random number seed, generates a first image by adding a first noise based on the first random number to a gray image, generates a second image by adding a second noise based on the second random number to the gray image, generates a third image by combining the first image and the second image, and generates a pseudo blood vessel pattern image that is an image including a pseudo blood vessel pattern using the third image;
Equipped with
the first random number seed is associated with an administration ID, and the second random number seed is associated with a personal ID;
The processor:
generating the third image by combining the first image and the second image, both of which are based on a normal distribution;
When generating the third image, a weight of the normal distribution is set;
generating a plurality of clusters each defined by a combination of the management ID and the weight, the plurality of clusters including a plurality of the pseudo blood vessel pattern images generated under a plurality of personal IDs different from one another;
A pseudo-vascular pattern generator. The processor:
generating a fourth image by diffusing noise contained in the third image;
generating a fifth image by applying a vessel enhancement filter to the fourth image;
generating the pseudo blood vessel pattern image by setting a region of interest in the fifth image;
The pseudo-vascular pattern generating device according to claim 1 . the processor, when generating the third image, sets a first weight for the mean of the normal distribution and a second weight for the variance of the normal distribution.
The pseudo-vascular pattern generating device according to claim 1 . generating a first random number based on a first random number seed;
generating a second random number based on a second random number seed different from the first random number seed;
generating a first image by adding a first noise based on the first random number to a gray image;
generating a second image by adding second noise based on the second random number to the gray image;
generating a third image by combining the first image and the second image;
generating a pseudo blood vessel pattern image, which is an image including a pseudo blood vessel pattern, using the third image;
the first random number seed is associated with an administration ID, and the second random number seed is associated with a personal ID;
generating the third image by combining the first image and the second image, both of which are based on a normal distribution;
When generating the third image, a weight of the normal distribution is set;
generating a plurality of clusters each defined by a combination of the management ID and the weight, the plurality of clusters including a plurality of the pseudo blood vessel pattern images generated under a plurality of personal IDs different from one another;
A method for generating pseudo-vascular patterns.