Iris authentication device, iris authentication system, iris authentication method, and recording medium

Abstract

The present invention comprises: an iris image acquisition unit 211 for acquiring an iris image LI in which an iris of a living body is included; a calculation unit 212 for calculating a magnification for the iris image LI from an intended size and the size of an iris region included in the iris image LI; a generation unit 213 for generating a resolution conversion image RI obtained by converting the resolution of the iris image LI in accordance with the magnification; and a post-conversion feature amount extraction unit 214 for extracting a post-conversion feature amount OC that is a feature amount of the resolution conversion image RI.

Description

[Technical Field]

[0001] The present disclosure relates to the technical fields of an iris authentication device, an iris authentication system, an iris authentication method, and a recording medium. [Background technology]

[0002] Non-Patent Document 1 describes a technique for super-resolution of images using machine learning with a matching loss function to increase the amount of information available for matching. Non-Patent Document 2 also describes a technique for performing super-resolution corresponding to various magnification ratios with a single network by estimating a filter according to the upsampling magnification ratio. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Maneet Singh, Shruti Nagpal, Mayank Vatsa, Richa Singh, Angshul Majumdar; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 479-488 [Non-patent document 2] Xuecai Hu, Haoyuan Mu, Xiangyu Zhang, Zilei Wang, Tieniu Tan, Jian Sun; Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 1575-1584 Summary of the Invention [Problem to be solved by the invention]

[0004] An object of this disclosure is to provide an iris authentication device, an iris authentication system, an iris authentication method, and a recording medium that aim to improve upon the techniques described in prior art documents. [Means for solving the problem]

[0005] One aspect of the iris authentication device includes an iris image acquisition means for acquiring an iris image including the iris of a living body, a calculation means for calculating a magnification for the iris image from the size of the iris region included in the iris image and a desired size, a generation means for generating a resolution-converted image in which the resolution of the iris image is converted in accordance with the magnification, and a post-conversion feature extraction means for extracting post-conversion features that are features of the resolution-converted image.

[0006] One aspect of the iris authentication system includes an iris image acquisition means for acquiring an iris image including the iris of a living body, a calculation means for calculating a magnification for the iris image from the size of the iris region included in the iris image and a desired size, a generation means for generating a resolution-converted image in which the resolution of the iris image is converted in accordance with the magnification, and a post-conversion feature extraction means for extracting post-conversion features that are features of the resolution-converted image.

[0007] One aspect of the iris authentication method involves obtaining an iris image containing the iris of a living body, calculating a magnification for the iris image from the size of the iris region contained in the iris image and a desired size, generating a resolution-converted image in which the resolution of the iris image is converted according to the magnification, and extracting post-conversion features that are features of the resolution-converted image.

[0008] One aspect of the recording medium causes a computer to execute an iris authentication method that acquires an iris image that includes the iris of a living body, calculates a magnification for the iris image from the size of the iris region included in the iris image and a desired size, generates a resolution-converted image in which the resolution of the iris image is converted in accordance with the magnification, and extracts post-conversion features that are features of the resolution-converted image. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a block diagram showing the configuration of an iris authentication device according to the first embodiment. [Figure 2] FIG. 2 is a block diagram showing the configuration of an iris authentication device according to the second embodiment. [Figure 3] FIG. 3 is a flowchart showing the flow of the iris authentication operation performed by the iris authentication device in the second embodiment. [Figure 4] FIG. 4 is a block diagram showing the configuration of an iris authentication device according to the third embodiment. [Figure 5] FIG. 5 is a flowchart showing the flow of the learning operation performed by the iris authentication device in the third embodiment. [Figure 6] FIG. 6 is a block diagram showing the configuration of an iris authentication device according to the fifth embodiment. [Figure 7] FIG. 7 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device in the fifth embodiment. [Figure 8] FIG. 8 is a block diagram showing the configuration of an iris authentication device according to the sixth embodiment. [Figure 9] FIG. 9 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device in the sixth embodiment. [Figure 10] FIG. 10 is a block diagram showing the configuration of an iris authentication device according to the seventh embodiment. [Figure 11] FIG. 11 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device in the seventh embodiment. [Figure 12] FIG. 12 is a block diagram showing the configuration of an iris authentication device according to the eighth embodiment. [Figure 13] FIG. 13 is a flowchart showing the flow of the iris authentication operation performed by the iris authentication device in the eighth embodiment. [Figure 14] FIG. 14 is a block diagram showing the configuration of an iris authentication system according to the ninth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of an iris authentication device, an iris authentication system, an iris authentication method, and a recording medium will be described with reference to the drawings. [1: First embodiment]

[0011] First, a first embodiment of an iris authentication device, an iris authentication method, and a recording medium will be described. Hereinafter, the first embodiment of the iris authentication device, the iris authentication method, and a recording medium will be described using an iris authentication device 1 to which the first embodiment of the iris authentication device, the iris authentication method, and a recording medium is applied. [1-1: Configuration of iris authentication device 1]

[0012] Fig. 1 is a block diagram showing the configuration of an iris authentication device 1 in the first embodiment. As shown in Fig. 1, the iris authentication device 1 includes an iris image acquisition unit 11, a calculation unit 12, a generation unit 13, and a post-conversion feature extraction unit 14.

[0013] The iris image acquisition unit 11 acquires an iris image LI containing the iris of a living subject. The iris refers to the ring-shaped area around the pupil inside the black part of the eye. Each individual has a unique iris pattern. Furthermore, the iris is covered by the cornea and is therefore less susceptible to damage, making it a suitable area for biometric authentication.

[0014] The calculation unit 12 calculates a magnification for the iris image LI from the size of the iris region included in the iris image LI and a desired size. The size of the iris region included in the iris image LI may be expressed, for example, by the number of pixels of the iris region in the iris image, the diameter of the iris region in the iris image, the area of the iris region in the iris image, etc. Below, an example will be described in which the size of the iris region included in the iris image LI is expressed by the number of pixels of the iris region in the iris image. Iris authentication uses the pattern of the iris for authentication. For this reason, it is necessary to use an image that includes an iris area with an appropriate number of pixels for iris authentication. The desired number of pixels may be the number of pixels appropriate for iris authentication. Since the iris is approximately circular, the number of pixels may be expressed as the radius of the area. An approximately circular iris area is also called an iris circle. The number of pixels may also correspond to the resolution. For example, the desired number of pixels may be 100 pixels or more, such as 100 pixels, 125 pixels, etc.

[0015] In general, it is preferable to use a relatively high-resolution iris image HI for iris authentication. In contrast, iris detection can also be performed on a low-resolution iris image LI because it detects the edges of the pupil and iris regions. Information obtained by this iris detection, such as the positions and pixel counts of the pupil and iris circles, can be used in super-resolution processing to increase the resolution of the low-resolution iris image LI. Here, super-resolution processing refers to processing that increases the resolution of a low-resolution image to generate a high-resolution image, and is capable of generating a relatively high-quality high-resolution image.

[0016] For example, the calculation unit 12 may calculate the magnification by finding the ratio between the radius of the detected iris circle and the radius of an area with a desired number of pixels. That is, the calculation unit 12 can calculate the magnification based on information obtained by iris detection. The magnification is not limited to 1 or more, and may be less than 1. For example, if the desired radius is 50 pixels and the radius of the iris circle included in the iris image LI is 25 pixels, the calculation unit 12 may calculate the magnification to be 2. Alternatively, if the desired radius is 50 pixels and the radius of the iris circle is 100 pixels, the calculation unit 12 may calculate the magnification to be 0.5.

[0017] The generation unit 13 generates a resolution-converted image RI by converting the resolution of the iris image LI according to the magnification. For example, if the magnification calculated by the calculation unit 12 from the radius of the iris circle is 2x, the generation unit 13 may generate a resolution-converted image RI by doubling the resolution of the iris image LI.

[0018] The post-conversion feature extraction unit 14 extracts post-conversion feature values OC, which are feature values of the resolution-converted image RI. The post-conversion feature extraction unit 14 may be configured to extract feature values from an image with a desired number of pixels. In order to enable the post-conversion feature extraction unit 14 to appropriately extract feature values, the calculation unit 12 may calculate a magnification, and the generation unit 13 may generate the resolution-converted image RI according to the magnification.

[0019] The feature amount here is a value that represents the iris feature required for iris authentication. The post-transformation feature amount extraction unit 14 may be configured with, for example, a convolutional neural network. [1-2: Technical Effects of Iris Recognition Device 1]

[0020] The iris authentication device 1 in the first embodiment can convert the iris image LI into an image with a desired number of pixels, regardless of the number of pixels of the iris image LI. The iris authentication device 1 in the first embodiment performs super-resolution processing on the low-resolution iris image LI to increase the resolution, and can obtain a high-resolution iris image HI. The resolution of the iris image LI on which the iris authentication device 1 in the first embodiment performs super-resolution processing may be any resolution and is not limited to a specific resolution. The iris authentication device 1 in the first embodiment can perform iris authentication using iris images LI with various resolutions.

[0021] In the iris authentication device 1 in the first embodiment, the calculation unit 12 calculates a magnification so that the post-conversion feature extraction unit 14 can appropriately extract features, and the generation unit 13 generates a resolution-converted image RI according to the magnification. That is, in the iris authentication device 1 in the first embodiment, there is no need to change the mechanism for iris authentication. Therefore, the iris authentication device 1 in the first embodiment can be applied to a mechanism that is constructed so that iris authentication can be performed using an iris image HI with a desired number of pixels. [2: Second embodiment]

[0022] Next, a second embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the second embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 2 to which the second embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied. [2-1: Configuration of Iris Authentication Device 2]

[0023] First, the configuration of the iris authentication device 2 in the second embodiment will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the configuration of the iris authentication device 2 in the second embodiment. In the following description, components that have already been described will be given the same reference numerals, and detailed description thereof will be omitted.

[0024] 2, the iris authentication device 2 includes a calculation device 21 and a storage device 22. Furthermore, the iris authentication device 2 may include a communication device 23, an input device 24, and an output device 25. However, the iris authentication device 2 does not have to include at least one of the communication device 23, the input device 24, and the output device 25. The calculation device 21, the storage device 22, the communication device 23, the input device 24, and the output device 25 may be connected via a data bus 26.

[0025] The arithmetic device 21 includes, for example, at least one of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and an FPGA (Field Programmable Gate Array). The arithmetic device 21 reads a computer program. For example, the arithmetic device 21 may read a computer program stored in the storage device 22. For example, the arithmetic device 21 may read a computer program stored in a computer-readable and non-transitory recording medium using a recording medium reading device (e.g., an input device 24 described later) not shown that is provided in the iris authentication device 2. The arithmetic device 21 may acquire (i.e., download or read) the computer program from a device (not shown) arranged outside the iris authentication device 2 via the communication device 23 (or another communication device). The arithmetic device 21 executes the read computer program. As a result, logical functional blocks for executing operations to be performed by the iris authentication device 2 are realized within the arithmetic device 21. In other words, the arithmetic device 21 can function as a controller for realizing logical function blocks for executing the operations (in other words, processing) that the iris authentication device 2 should perform.

[0026] FIG. 2 shows an example of logical functional blocks implemented within the computing device 21 for performing iris authentication operations. As shown in FIG. 2, the computing device 21 includes an iris image acquisition unit 211, which is a specific example of "iris image acquisition means," a calculation unit 212, which is a specific example of "calculation means," a generation unit 213, which is a specific example of "generation means," a transformed feature extraction unit 214, which is a specific example of "transformed feature extraction means," and an authentication unit 215, which is a specific example of "determination means" and "authentication means." The calculation unit 212 may include an iris circle detection unit 2121 and a magnification ratio calculation unit 2122. Details of the operations of the iris image acquisition unit 211, the calculation unit 212, the generation unit 213, the transformed feature extraction unit 214, and the authentication unit 215 will be described later with reference to FIG. 3. The computing device 21 does not necessarily have to include the authentication unit 215.

[0027] The storage device 22 can store desired data. For example, the storage device 22 may temporarily store a computer program executed by the arithmetic device 21. The storage device 22 may temporarily store data that the arithmetic device 21 temporarily uses when the arithmetic device 21 is executing a computer program. The storage device 22 may store data that the iris authentication device 2 stores long-term. The storage device 22 may include at least one of a RAM (Random Access Memory), a ROM (Read Only Memory), a hard disk device, a magneto-optical disk device, an SSD (Solid State Drive), and a disk array device. In other words, the storage device 22 may include a non-transitory recording medium.

[0028] The storage device 22 may store the super-resolution model SM, the feature generation model GM, and the matching feature CC. However, the storage device 22 does not have to store at least one of the super-resolution model SM, the feature generation model GM, and the matching feature CC. The super-resolution model SM, the feature generation model GM, and the matching feature CC will be described in detail later.

[0029] The communication device 23 is capable of communicating with devices external to the iris authentication device 2 via a communication network (not shown).

[0030] The input device 24 is a device that accepts input of information to the iris authentication device 2 from outside the iris authentication device 2. For example, the input device 24 may include an operation device (for example, at least one of a keyboard, a mouse, and a touch panel) that can be operated by an operator of the iris authentication device 2. For example, the input device 24 may include a reading device that can read information recorded as data on a recording medium that can be externally attached to the iris authentication device 2.

[0031] The output device 25 is a device that outputs information to the outside of the iris authentication device 2. For example, the output device 25 may output information as an image. That is, the output device 25 may include a display device (a so-called display) that can display an image showing the information to be output. For example, the output device 25 may output information as sound. That is, the output device 25 may include an audio device (a so-called speaker) that can output sound. For example, the output device 25 may output information on paper. That is, the output device 25 may include a printing device (a so-called printer) that can print desired information on paper. [2-2: Iris authentication operation performed by iris authentication device 2]

[0032] Next, the iris authentication operation performed by the iris authentication device 2 in the second embodiment will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of the iris authentication operation performed by the iris authentication device 2 in the second embodiment. Below, an example will be described in which the number of pixels in the iris area is smaller than the desired number of pixels and the magnification is an enlargement ratio.

[0033] As shown in FIG. 3, the iris image acquisition unit 211 acquires an iris image including the iris of a living body (step S21). The iris circle detection unit 2121 detects an iris circle from the iris image (step S22). The iris circle detection unit 2121 may calculate a vector representing the center position and radius of the iris circle from the input iris image. The iris circle detection unit 2121 may be configured, for example, as a recurrent neural network. The recurrent neural network includes multiple convolutional layers and multiple activation layers, extracts features of the input image, and converts the extracted features into a vector representing the center position and radius of the corresponding region using a linear layer. The iris image LI input to the iris circle detection unit 2121 and the vector output from the iris circle detection unit 2121 may be normalized. When the iris circle detection unit 2121 is configured as a neural network, any neural network structure may be used as long as it meets the requirements. Examples of the neural network structure include those similar to those of VGG, Residual neural network (ResNet), and other models trained on large-scale image datasets, but other structures may also be used. A normalization layer such as batch normalization may be used as an intermediate layer of the neural network. ReLU (Rectified Linear Unit) is often used as an activation layer, but other activation functions may also be used. Furthermore, the iris circle detection unit 2121 may be an image processing mechanism that is not configured as a neural network.

[0034] The magnification ratio calculation unit 2122 calculates a magnification ratio for the iris image LI from the radius of the iris circle included in the iris image LI detected by the iris circle detection unit 2121 and the desired radius (step S23). The magnification ratio may be the ratio between the radius of the iris circle included in the iris image LI and the radius of an iris circle of the desired size. Furthermore, the magnification ratio may not be a simple ratio between the radius of the iris circle included in the iris image LI and the radius of an iris circle of the desired size, but may be a value converted into, for example, a logarithm or a power of the ratio. Note that in the second embodiment, as with the calculation unit 12 in the first embodiment, the magnification ratio calculation unit 2122 may calculate a magnification of less than 1 as a parameter corresponding to the magnification ratio in addition to or instead of the magnification ratio.

[0035] Furthermore, the iris circle detection unit 2121 may calculate the diameter of the iris circle from the input iris image. In this case, the magnification ratio calculation unit 2122 calculates the magnification ratio for the iris image LI from the diameter of the iris circle included in the iris image LI detected by the iris circle detection unit 2121 and the desired diameter. Furthermore, the iris circle detection unit 2121 may calculate the area of the iris circle from the input iris image. In this case, the magnification ratio calculation unit 2122 calculates the magnification ratio for the iris image LI from the area of the iris circle included in the iris image LI detected by the iris circle detection unit 2121 and the desired area.

[0036] The generation unit 213 generates a resolution-converted image RI, which is a super-resolution image obtained by increasing the resolution of the iris image LI, in accordance with the magnification ratio (step S24). The generation unit 213 may use the magnification ratio calculated by the magnification ratio calculation unit 2122 as is, or may normalize the magnification ratio calculated by the magnification ratio calculation unit 2122 before using it. The generation unit 213 may generate the resolution-converted image RI, which is a super-resolution image, using a super-resolution model SM. The super-resolution model SM is a model constructed by machine learning so as to output a resolution-converted image RI in response to an input of an iris image LI. Specific examples of a method for constructing the super-resolution model SM will be described in detail in the third and fourth embodiments. Specific examples of the constructed super-resolution model SM will be described in detail in the fifth to seventh embodiments.

[0037] The post-conversion feature extraction unit 214 extracts post-conversion feature values OC that are feature values of the resolution-converted image RI (step S25). The post-conversion feature extraction unit 214 may extract the post-conversion feature values OC from the resolution-converted image RI using the feature generation model GM.

[0038] The feature generation model GM is a model capable of generating features of the iris image HI when the iris image HI, which includes an iris region with a desired number of pixels and has a resolution suitable for authentication, is input by the post-conversion feature extraction unit 214. The feature generation model GM may be constructed by machine learning so that, when the iris image HI is input, it can output features suitable for iris authentication. Specifically, the feature generation model GM may be constructed by adjusting learning parameters included in the feature generation model GM so that a loss function set based on the error between multiple features generated from the iris image HI of the same individual is reduced (preferably minimized). The feature generation model GM may be constructed, for example, as a convolutional neural network that generates features by convolution processing. The feature generation model GM may be any model capable of generating features with high accuracy, and may also be another trained neural network.

[0039] The constructed feature generation model GM may receive matching data as input and generate matching features CC, which are features of the matching data. The generated matching features CC may be registered in the storage device 22.

[0040] The authentication unit 215 performs authentication using a score indicating the similarity between the transformed feature OC and a pre-prepared feature (step S26). Here, authentication refers to at least one of identifying the person and determining that the person is the person. The authentication unit 215 may determine that the person is the person if the matching score indicating the similarity between the transformed feature OC and a pre-prepared matching feature CC is equal to or greater than a threshold. The authentication unit 215 may calculate the matching score, for example, using the cosine similarity between the transformed feature OC and the matching feature CC. The authentication unit 215 may determine whether the respective feature values are similar by utilizing the property that feature values of data related to the same individual tend to be similar and point in the same direction, i.e., tend to have large cosine similarity. Alternatively, the authentication unit 215 may calculate the matching score, for example, using an L2 distance function or an L1 distance function between the transformed feature OC and the matching feature CC. The authentication unit 215 may determine whether the feature amounts are similar by utilizing the property that the feature amounts of data relating to the same individual tend to be close to each other, such as the L2 distance function or the L1 distance function.

[0041] The output device 25 outputs the authentication result by the authentication unit 215, the magnification calculated by the magnification calculation unit 2122, and the resolution-converted image RI generated by the generation unit 213, to the outside of the iris authentication device 2 (step S27). The output from the output device 25 may be confirmed by the person to be authenticated, a manager, a security guard, etc. The output device 25 may also output an alert when the magnification is a predetermined size or larger. If the generation unit 213 enlarges the image at a magnification larger than a predetermined size, there is a possibility that the authentication accuracy may decrease, but by the output device 25 outputting an alert, the manager, security guard, etc. can pay attention to the corresponding authentication. [2-3: Technical Effects of Iris Recognition Device 2]

[0042] Iris authentication often requires a relatively high-resolution iris image HI with an iris radius of 100 pixels or more. On the other hand, even if a low-resolution image LI of less than 100 pixels is used for iris authentication, as long as a certain level of accuracy is achieved, it is possible to perform authentication simultaneously with other biometric authentication methods using, for example, a single relatively low-resolution camera.

[0043] The iris authentication device 2 in the second embodiment can perform iris authentication with high accuracy, even when a low-resolution iris image LI is input, because it can convert the image into a high-resolution converted image RI, which is a super-resolution image, regardless of the resolution of the iris image LI. Therefore, by applying the iris authentication device 2 in the second embodiment, it is possible to realize authentication that performs both other biometric authentication and iris authentication, for example, using an image captured using a single, relatively inexpensive camera.

[0044] In addition, when the magnification is 1 or less, resolution conversion processing may be performed using a general bilinear method, bicubic method, area average method, nearest neighbor method, etc., instead of the resolution conversion processing by the generation unit 213 in the second embodiment. [3: Third embodiment]

[0045] Next, a third embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the third embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 3 to which the third embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied. [3-1: Configuration of Iris Recognition Device 3]

[0046] 4 is a block diagram showing the configuration of the iris authentication device 3 in the third embodiment. In the following description, components that have already been described are given the same reference numerals, and detailed description thereof will be omitted.

[0047] As shown in Fig. 4, the iris authentication device 3 includes a calculation device 21 and a storage device 22. As shown in Fig. 4, the calculation device 21 includes a training image acquisition unit 316 which is a specific example of "training image acquisition means", an input image generation unit 317 which is a specific example of "input image acquisition means", a learning unit 318 which is a specific example of "learning means", and an iris information estimation unit 300 which includes an iris image acquisition unit 211, a calculation unit 212, a generation unit 213, and a converted feature extraction unit 214. The input image generation unit 317 includes a batch data extraction unit 3171 and a resolution conversion unit 3172. The learning unit 318 includes a loss function calculation unit 3181, a gradient calculation unit 3182, and a parameter update unit 3183.

[0048] The storage device 22 may store a training image TI. However, the storage device 22 does not have to store a training image TI. If the storage device 22 does not store a training image TI, the communication device 23 may acquire the training image TI from a device external to the iris authentication device 2, or the input device 24 may accept input of the training image TI from outside the iris authentication device 2. The training image TI may be an iris image including an iris region with a desired number of pixels.

[0049] In the third embodiment, the training image acquisition unit 316, the input image generation unit 317, the training unit 318, and the iris information estimation unit 300 perform machine learning using the training images TI to construct a super-resolution model SM used by the generation unit 213. Details of the operations of the training image acquisition unit 316, the input image generation unit 317, the training unit 318, and the iris information estimation unit 300 will be described with reference to FIG. 5 . [3-2: Learning operation performed by iris authentication device 3]

[0050] Next, the learning operation performed by the iris authentication device 3 in the third embodiment will be described with reference to Fig. 5. Fig. 5 is a flowchart showing the flow of the learning operation performed by the iris authentication device 3 in the third embodiment.

[0051] 5, the training image acquisition unit 316 acquires a data set of training images TI including an iris region with a desired number of pixels, which is stored in, for example, the storage device 22 (step S31). The training images TI may be images with the same resolution as the iris images HI, which have a resolution suitable for authentication by the iris authentication device 3.

[0052] The batch data extraction unit 3171 randomly extracts batch data of training images TI of a batch size from the data set of training images TI acquired by the training image acquisition unit 316 (step S32). For example, if the batch size is 32, the batch data extraction unit 3171 extracts 32 training images TI. The batch size may be a value of 32, 64, 128, etc. There is no particular limitation on the value of the batch size, and any value can be used.

[0053] The resolution conversion unit 3172 generates an input image II by converting the resolution of the training image TI according to the reciprocal of an arbitrary enlargement ratio (step S33). That is, the resolution conversion unit 3172 generates a low-resolution input image II from the high-resolution training image TI. The input image generation unit 317 can prepare an image obtained by reducing the resolution of the training image TI as the input image II.

[0054] The resolution conversion unit 3172 may reduce the training image TI to generate a low-resolution input image II. The resolution conversion unit 3172 may reduce the training image TI by thinning out pixels of the training image TI. In other words, by reducing the training image TI by the resolution conversion unit 3172, it is possible to generate an input image II with a lower resolution than the training image TI.

[0055] The resolution conversion unit 3172 may generate input images II by lowering the resolution of each of the training images TI extracted by the batch data extraction unit 3171, for example, by using the reciprocal of an arbitrary enlargement ratio selected according to a uniform random number distribution. In this case, the resolution conversion unit 3172 can generate batch data that uniformly includes input images II of various resolutions.

[0056] Alternatively, the resolution conversion unit 3172 may generate input images II by reducing the resolution of all batch data of training images TI of a batch size extracted by the batch data extraction unit 3171 at the same time using the reciprocal of the same magnification ratio. In this case, the resolution conversion unit 3172 can generate batch data including input images II of the same resolution. In this case, the resolution conversion unit 3172 may reduce the resolution of batch data of training images TI of a batch size extracted by the batch data extraction unit 3171 at different times using the reciprocal of different magnification ratios to generate input images II. It is sufficient for the input image generation unit 317 to prepare input images II such that input images II of various resolutions are uniformly included in the entire dataset of training images TI acquired by the training image acquisition unit 316.

[0057] The operations of steps S34, S35, and S36 may be similar to the operations of steps S21, S24, and S25 described with reference to Fig. 3. The data used for the operations in the second embodiment was an iris image LI for authentication, but the data used for the operations in the third embodiment is different in that it is an input image II prepared for learning.

[0058] The iris image acquisition unit 211 acquires one input image II from batch data of input images II of the batch size (step S34). The generation unit 213 generates a resolution-converted input image RII by converting the resolution of the input image II using the magnification ratio used when the resolution conversion unit 3172 reduced the training image TI (step S35). Here, the resolution conversion unit 3172 and the generation unit 213 perform reduction and resolution conversion using the same magnification ratio, so the resolution-converted input image RII has the same resolution as the training image TI. The converted feature extraction unit 214 extracts input feature OIC, which is a feature of the resolution-converted input image RII (step S36). Furthermore, the converted feature extraction unit 214 extracts training feature TC, which is a feature of the training image TI, or the training feature TC, which is a feature of the training image TI, may be stored in advance in the storage device 22 as a set with the training image TI.

[0059] The iris image acquisition unit 211 determines whether all input images II of the batch data of the batch-size input image II have been processed (step S39). If all input images II of the batch data of the batch-size input image II have not been processed (step S39: No), the process proceeds to step S34. That is, the iris information estimation unit 300 performs the operations of steps S34 to S38 for all input images II of the batch data of the batch-size input image II. If the calculation is performed by a GPU or multi-threading, the estimation calculation for each batch-size input image may be performed in parallel. It is also possible to perform some processing in parallel, and to mix serial and parallel processing.

[0060] If all input images II of the batch data of the batch size of input images II have been processed (step S39: Yes), the learning unit 318 causes the generation unit 213 to learn a method for generating resolution-converted images RI. Specifically, the learning unit 318 causes the super-resolution model SM used by the generation unit 213 to learn a method for generating resolution-converted images RI, and constructs the super-resolution model SM. More specifically, the learning unit 318 adjusts learning parameters included in the super-resolution model SM. The learning unit 318 causes the generation unit 213 to learn a method for generating resolution-converted images RI based on at least one of a first loss function in which the loss increases as the similarity between the training feature amount TC and the input feature amount OIC increases, and a second loss function in which the loss increases as the similarity between the training image TI and the resolution-converted input image RII decreases. The learning unit 318 may optimize the iris information estimation unit 300 based on the loss function.

[0061] First, the loss function calculation unit 3181 performs calculations using at least one of a first loss function in which the loss increases as the training feature TC and the input feature OIC become less similar, and a second loss function in which the loss increases as the training image TI and the resolution-converted input image RII become less similar (step S40).

[0062] The loss function calculation unit 3181 may input the learned features TC, which are the correct individual labels, and the input features OIC of the resolution-converted input image RII extracted by the converted feature extraction unit 214, and output a first loss value indicating the degree of dissimilarity between them. The loss function calculation unit 3181 may compare a one-hot vector generated from the learned features TC, which are the correct individual labels, with a feature vector as the input features OIC extracted by the converted feature extraction unit 214, using a cross-entropy loss function, to obtain the first loss.

[0063] Furthermore, the loss function calculation unit 3181 may input the training image TI, which is a high-resolution image, and the resolution-converted input image RII generated by the generation unit 213, and output a second loss value indicating the degree of dissimilarity between them. The loss function calculation unit 3181 may compare the training image TI and the resolution-converted input image RII generated by the generation unit 213 using an L1 distance loss function to obtain the second loss.

[0064] The loss function calculation unit 3181 is not limited to the cross-entropy loss function or the L1 distance loss function, and may use other loss functions such as the KL divergence function or the L2 distance function.

[0065] The loss function calculation unit 3181 may weight the calculated loss according to the magnification rate calculated by the calculation unit 212. Generally, super-resolution processing at a large magnification rate is often more difficult than super-resolution processing at a small magnification rate. In other words, authentication processing using a super-resolution image obtained by super-resolution processing at a large magnification rate often results in lower authentication accuracy than authentication processing using a super-resolution image obtained by super-resolution processing at a small magnification rate. Therefore, the loss function calculation unit 3181 may use a loss function that assigns a larger weight to the loss resulting from super-resolution processing at a large magnification rate. In other words, the learning unit 318 may cause the generation unit 213 to perform learning based on a loss function in which the weight of the loss corresponding to an input image II generated by using the first magnification rate as an arbitrary magnification rate is larger than the weight of the loss corresponding to an input image II generated by using a second magnification rate smaller than the first magnification rate as an arbitrary magnification rate. First, the learning unit 318 may cause the generation unit 213 to perform learning based on a loss function in which the larger the enlargement factor, the greater the weight of the loss corresponding to the resolution-converted input image RII generated using that enlargement factor. In this way, the contribution of super-resolution processing to learning increases when the enlargement factor is large. Then, the learning unit 318 can construct a super-resolution model SM whose authentication performance is less dependent on the enlargement factor.

[0066] The loss function calculation unit 3181 may separately weight the first loss and the second loss according to the magnification rate, or may weight the first loss and the second loss according to the magnification rate, take the sum of the weights, and output a single loss.

[0067] For example, if the resolution conversion unit 3172 generates batch data of a batch size that uniformly includes input images II of various resolutions, the loss function calculation unit 3181 may calculate the loss of the batch data of the batch size by weighting the loss of each input image II according to the various resolutions. The loss function calculation unit 3181 may calculate the average value of each weighted loss and output it as the loss of the batch data of the batch size. Specifically, the resolution conversion unit 3172 may calculate the loss of the batch data of the batch size by weighting the loss of each input image II according to the enlargement ratio used in step S33 to generate each input image II. As an example, if the first enlargement ratio used to generate the first input image II is greater than the second enlargement ratio used to generate the second input image II, the resolution conversion unit 3172 may calculate the loss of the batch data of the batch size by weighting the loss of the first input image II according to the enlargement ratio used in step S33 to generate each input image II so that the weight for the loss of the first input image II is greater than the weight for the loss of the second input image II.

[0068] On the other hand, for example, if the resolution conversion unit 3172 generates batch data of a batch size including input images II of the same resolution, the loss function calculation unit 3181 may calculate the loss of the batch data of the batch size by applying the same weight to the loss of each input image II. For example, the loss function calculation unit 3181 may calculate an average loss value, which is the average value of the losses of each input image II. In this case, since the batch data of the batch size generated by the resolution conversion unit 3172 at different times each have different resolutions, the loss function calculation unit 3181 may apply a weight to the average loss value according to the resolution.

[0069] The gradient calculation unit 3182 uses the loss value output by the loss function calculation unit 3181 to calculate the gradient of the learning parameters included in the super-resolution model SM using backpropagation (step S41). The parameter update unit 3183 uses the calculated gradient of the learning parameters to update the values of the learning parameters included in the super-resolution model SM (step S42). Updating the values of the learning parameters in step S42 corresponds to learning the super-resolution model SM. For example, the parameter update unit 3183 may optimize the values of the learning parameters so that the value of the loss function is minimized. Examples of optimization methods used by the parameter update unit 3183 include, but are not limited to, stochastic gradient descent and Adam. Even when using stochastic gradient descent, the parameter update unit 3183 may update the learning parameters using hyperparameters such as weight decay and momentum.

[0070] The input image generation unit 317 determines whether batch data has been extracted from the predetermined training image TI (step S43). If batch data has not been extracted from the predetermined training image TI (step S43: No), the process proceeds to step S32. For example, if the training image acquisition unit 316 has acquired a data set of 320 training images TI and the batch size is 32, the iris information estimation unit 300 may perform the operations of steps S32 to S42 10 times. If batch data has been extracted from the predetermined training image TI (step S43: Yes), the learning unit 318 stores an optimized super-resolution model SM including optimally updated learning parameters in the storage device 22 (step S44). [3-3: Technical Effects of Iris Recognition Device 3]

[0071] The iris authentication device 3 in the third embodiment has the generation unit 213 learn how to generate the resolution-converted image RI based on a loss function in which the loss increases the more dissimilar the training image TI and the resolution-converted input image RII are, thereby improving the accuracy of the super-resolution processing. Furthermore, the iris authentication device 3 in the third embodiment causes the generation unit 213 to learn a method for generating a resolution-converted image RI based on a loss function in which the loss increases as the similarity between the training feature TC and the input feature OIC decreases, thereby generating a resolution-converted image RI suitable for iris authentication. That is, since features extracted from an image that has been subjected to super-resolution processing are used for learning the super-resolution processing, a super-resolution model SM can be constructed that can generate a resolution-converted image RI from which features suitable for iris authentication can be extracted. Since the converted feature OC output by the iris information estimation unit 300 is a feature used for authenticating the person, it is desirable that the resolution-converted image RI be an image from which converted feature OC suitable for authenticating the person can be extracted. In other words, the resolution-converted image RI generated by the super-resolution model SM is an image that has been subjected to super-resolution processing with high accuracy and is also an image suitable for matching.

[0072] Furthermore, since the difficulty of super-resolution processing generally varies depending on the magnification ratio, uniform calculation of the loss function regardless of the magnification ratio may result in insufficient accuracy in the super-resolution processing. In other words, the accuracy of the super-resolution processing may deteriorate when the magnification ratio is large. In contrast, the iris authentication device 3 in the third embodiment uses a loss function that applies a weight according to the magnification ratio, thereby maintaining the accuracy of the super-resolution processing even when the magnification ratio changes. As an example, when the first magnification ratio used to generate the first input image II is greater than the second magnification ratio used to generate the second input image II, the resolution conversion unit 3172 calculates the loss of the batch data of the batch size by applying a weight according to the magnification ratio used in step S33 to generate each input image II so that the weight for the loss of the first input image II is greater than the weight for the loss of the second input image II. As a result, it is possible to construct a super-modified model SM that can generate a resolution-converted image RI that can maintain the accuracy of iris authentication even when a relatively low-resolution iris image LI is input. As a result, the generation unit 213 that uses the super-resolution model SM constructed by the iris authentication device 3 of the third embodiment can realize the generation of a high-resolution resolution-converted image RI that is suitable for highly accurate matching, regardless of the resolution of the iris image LI. In other words, the iris authentication device 3 in the third embodiment has devised a learning method for super-resolution processing, so that it can maintain the accuracy of iris authentication even when an iris image LI with a relatively low resolution that is likely to result in low matching accuracy is input.

[0073] Therefore, the iris authentication device 3 in the third embodiment can construct a super-resolution model SM whose authentication performance is less dependent on the magnification ratio.The iris authentication device 3 in the third embodiment can perform super-resolution processing of iris images that are compatible with various magnification ratios while maintaining authentication accuracy. [4: Fourth embodiment]

[0074] Next, a fourth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the fourth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 3 to which the fourth embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied.

[0075] The iris authentication device 3 in the fourth embodiment may have the same configuration as the iris authentication device 3 in the third embodiment described above. The iris authentication device 3 in the fourth embodiment is different from the iris authentication device 3 in the third embodiment in the generation process of the input image II by the resolution conversion unit 3172 and the calculation process of the loss function by the loss function calculation unit 3181. That is, the iris authentication device 3 in the fourth embodiment is different from the iris authentication device 3 in the third embodiment in the operation of step S33 and the operation of step S40 shown in Fig. 5. The other features of the iris authentication device 3 in the fourth embodiment may be the same as the other features of the iris authentication device 3 in the third embodiment. [4-1: Learning operation performed by iris authentication device 3]

[0076] In the fourth embodiment, a case will be described in which the resolution conversion unit 3172 uses a first enlargement ratio and a second enlargement ratio smaller than the first enlargement ratio as arbitrary enlargement ratios. The resolution conversion unit 3172 generates multiple input images II so that the frequency of generating an input image II according to the inverse of the first enlargement ratio is higher than the frequency of generating an input image II according to the inverse of the second enlargement ratio (step S33). That is, the resolution conversion unit 3172 generates multiple input images II so that the larger the enlargement ratio used, the more input images II are generated. Therefore, the multiple input images II generated by the resolution conversion unit 3172 include more input images II generated using the inverse of a larger enlargement ratio. In other words, the larger the enlargement ratio value, the more frequently the resolution conversion unit 3172 selects an enlargement ratio.

[0077] In general, the larger the magnification ratio of the super-resolution processing, the more difficult it is to perform the super-resolution processing compared to when the magnification ratio of the super-resolution processing is small. Therefore, by increasing the learning of the corresponding super-resolution processing as the magnification ratio of the super-resolution processing, it is expected that a super-resolution model SM capable of super-resolution processing with high accuracy can be constructed even when the magnification ratio of the super-resolution processing is large. In other words, by preparing many images with low resolution that require super-resolution processing using a large magnification ratio as input images II, it is expected that a super-resolution model SM capable of performing super-resolution processing with high accuracy regardless of the magnification ratio can be constructed.

[0078] For this reason, the resolution conversion unit 3172 may be configured so that the smaller the value of the inverse of the magnification ratio, the more likely it is to be selected as a value to be used in the resolution reduction process of the training image TI. The resolution conversion unit 3172 may select the magnification ratio to be used using a probability distribution that makes it easier to generate an input image II with a lower resolution. The resolution conversion unit 3172 may also select the magnification ratio to be used according to a weighted probability distribution. In order to train many super-resolution processes using large magnification ratios, the resolution conversion unit 3172 can select the magnification ratio to be used using a probability distribution that makes it easier to generate low-resolution images with large magnification ratios. In other words, the resolution conversion unit 3172 may be configured so that the lower the resolution of the input image II, the more likely it is to be generated. As a result, larger magnification ratios are more likely to be used in the subsequent super-resolution process by the generation unit 213. The probability distribution used by the resolution conversion unit 3172 to select the magnification ratio may be created using a linear function, a quadratic function, or the like. There are no other limitations on the probability distribution to be used as long as it makes it easier to select low-resolution images with large magnification ratios.

[0079] By devising the generation process of the input image II by the resolution conversion unit 3172, it is possible to construct a super-resolution model SM that can realize super-resolution processing so that the authentication performance does not depend strongly on the magnification ratio.

[0080] The operation of the resolution conversion unit 3172 in the fourth embodiment plays the same role as the weighting calculation by the loss function calculation unit 3181 in the third embodiment. Therefore, in the fourth embodiment, the loss function calculation unit 3181 does not need to weight the loss in the loss calculation. Therefore, in the fourth embodiment, the loss function calculation unit 3181 does not need to perform weighting according to the magnification rate (step S40). [4-2: Technical Effects of Iris Recognition Device 4]

[0081] The iris authentication device 3 of the fourth embodiment constructs a super-resolution model SM by performing machine learning weighted according to the magnification ratio in order to generate a high-resolution, resolution-converted image RI suitable for highly accurate matching, regardless of the resolution of the iris image LI. The generation unit 213 using the super-resolution model SM constructed by the iris authentication device 3 of the fourth embodiment can realize the generation of a high-resolution, resolution-converted image RI suitable for highly accurate matching, regardless of the resolution of the iris image LI.

[0082] Moreover, the iris authentication device 3 of the fourth embodiment also causes the generation unit 213 to learn a method for generating the resolution-converted image RI based on a loss function in which the loss increases as the similarity between the training image TI and the resolution-converted input image RII increases, thereby improving the accuracy of the super-resolution processing. Moreover, the iris authentication device 3 of the fourth embodiment also causes the generation unit 213 to learn a method for generating the resolution-converted image RI based on a loss function in which the loss increases as the similarity between the training feature amount TC and the input feature amount OIC increases, thereby making it possible to generate a resolution-converted image RI suitable for iris authentication. Therefore, the iris authentication device 3 of the fourth embodiment can also construct a super-resolution model SM whose authentication performance is less dependent on the magnification ratio. And the iris authentication device 3 of the fourth embodiment can also perform super-resolution processing of iris images that are compatible with various magnification ratios while maintaining authentication accuracy.

[0083] Both the iris authentication device 3 in the third embodiment and the iris authentication device 3 in the fourth embodiment can achieve highly accurate super-resolution processing regardless of the magnification ratio, and have the effect of enabling highly accurate iris authentication regardless of the magnification ratio of the super-resolution processing, but the iris authentication device 3 in the third embodiment has a simpler construction process for the super-resolution model SM than the iris authentication device 3 in the fourth embodiment. Also, the iris authentication device 3 in the fourth embodiment weights the distribution of the resolution of the input image II and directly manipulates the input image II to be input, so compared to the iris authentication device 3 in the third embodiment, the contribution of the weighting to the construction process is greater, and it is possible to more effectively prevent accuracy from being reduced by the magnification ratio. [5: Fifth embodiment]

[0084] Next, a fifth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the fifth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 5 to which the fifth embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied. [5-1: Configuration of iris authentication device 5]

[0085] 6 is a block diagram showing the configuration of the iris authentication device 5 in the fifth embodiment. In the following explanation, components that have already been explained are given the same reference numerals, and detailed explanations thereof will be omitted. The iris authentication device 5 includes an iris image acquisition unit 211, a calculation unit 212, a generation unit 513, and a converted feature extraction unit 214.

[0086] The generation unit 513 performs super-resolution processing using the super-resolution model SM to generate a resolution-converted image RI by converting the resolution of the iris image LI according to the magnification ratio. The generation unit 513 includes a feature extraction unit 5131, a filter generation unit 5132, and a conversion unit 5133. Details of the operations of the feature extraction unit 5131, the filter generation unit 5132, and the conversion unit 5133 will be described with reference to FIG. 7. [5-2: Super-resolution processing performed by the generation unit 513]

[0087] Next, the super-resolution processing performed by the iris authentication device 5 in the fifth embodiment will be described with reference to Fig. 7. Fig. 7 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device 5 in the fifth embodiment.

[0088] 7, the iris image acquisition unit 211 acquires an iris image LI including the iris of a living body (step S51). The calculation unit 212 calculates the magnification ratio for the iris image LI (step S52).

[0089] The feature extraction unit 5131 extracts pre-conversion feature PC, which is a feature of the iris image LI (step S53). The feature extraction unit 5131 may extract the pre-conversion feature PC from the low-resolution iris image LI using a low-resolution feature extraction model included in the super-resolution model SM. The low-resolution feature extraction model may be a model that, when the low-resolution iris image LI is input, is capable of outputting feature values that are suitable for the filter processing described below. The low-resolution feature extraction model may be constructed, for example, by machine learning, so that, when the iris image LI is input, it is capable of outputting feature values that are suitable for the filter processing described below. The feature extraction unit 5131 may input the iris image LI to the low-resolution feature extraction model and cause it to output the pre-conversion feature PC.

[0090] The filter generation unit 5132 generates one or more conversion filters for converting the pre-conversion feature PC according to the magnification ratio calculated by the calculation unit 212 (step S54). The filter generation unit 5132 may generate one or more conversion filters according to the magnification ratio using a conversion filter generation model included in the super-resolution model SM. The conversion filter generation model may be a model capable of generating a conversion filter that is suitable for filter processing, which will be described later, when the magnification ratio is input. The conversion filter generation model may be configured, for example, by machine learning, so that it can output a conversion filter that is suitable for filter processing, which will be described later, when the magnification ratio is input. The filter generation unit 5132 may input the magnification ratio calculated by the calculation unit 212 into the conversion filter generation model, and cause it to output one or more conversion filters.

[0091] The filter generation unit 5132 may generate a conversion filter for convolution processing. The filter generation unit 5132 may generate a conversion filter with a size of, for example, 3×3. The size of the conversion filter is not limited to 3×3 and may be 5×5. The size of the conversion filter can be determined arbitrarily depending on requirements such as processing speed and processing accuracy. Alternatively, the filter generation unit 5132 may determine the size of the conversion filter. Furthermore, the filter generation unit 5132 may generate, for example, (Cin×Cout) conversion filters. Cin may be a number corresponding to, for example, the number of channels of the pre-conversion feature PC. For example, Cin may be 3 when the iris image LI is a color image, and may be 1 when the iris image LI is a grayscale image. For example, Cout may be 3 when the resolution-converted image RI output by the filter processing is a color image, and may be 1 when the resolution-converted image RI output by the filter processing is a grayscale image.

[0092] The conversion filter generated by the filter generation unit 5132 may be used to increase the resolution of pre-conversion features PC extracted from the low-resolution iris image LI. The pre-conversion features PC extracted from the low-resolution iris image LI may have a size of (Cin×h×w), for example. More specifically, the feature extraction unit 5131 may generate Cin pre-conversion features PC each having a size of (h×w). The feature obtained by increasing the resolution using the conversion filter may have a size of (Cout×H×W), for example. More specifically, Cout high-resolution features each having a size of (H×W) may be generated.

[0093] For example, it is assumed that the calculation unit 212 calculates an expansion ratio constituted by a one-dimensional vector. In this case, the conversion filter generation model may receive the expansion ratio constituted by the one-dimensional vector as input, and output a conversion filter having a size of (Cin×Cout×3×3). More specifically, the conversion filter generation model may receive the expansion ratio constituted by the one-dimensional vector as input, and output Cin×Cout conversion filters having a size of (3×3). Alternatively, the conversion filter generation model may receive the expansion ratio constituted by the one-dimensional vector as input, and output Cin×Cout conversion filters having a size of (h×w).

[0094] Furthermore, the filter generation unit 5132 may generate a transform filter other than a filter for convolution processing. For example, the filter generation unit 5132 may generate a transform filter having the same size as the feature extracted by the feature extraction unit 5131. The size of the feature may be, for example, (Cin×h×w).

[0095] The conversion unit 5133 converts the pre-conversion feature PC through filtering using one or more conversion filters, thereby generating a resolution-converted image RI (step S55). The conversion unit 5133 may perform filtering on the pre-conversion feature PC using the conversion filter generated by the filter generation unit 5132. The conversion unit 5133 may also convert the low-resolution iris image LI using the conversion filter generated by the filter generation unit 5132, thereby generating a resolution-converted image RI that is a highly resolved super-resolution image.

[0096] Before performing the filtering process, the converter 5133 may adjust the size of the pre-conversion feature PC according to the magnification rate. For example, if the magnification rate is 2x, the converter 5133 may insert zeros between pixels of the pre-conversion feature PC to double the size of the pre-conversion feature PC. Furthermore, if the magnification rate is 1.5x, the converter 5133 may insert zeros between every third pixel of the pre-conversion feature PC to double the size of the pre-conversion feature PC. The converter 5133 may also insert a value other than zero between pixels to enlarge the size of the pre-conversion feature PC. For example, the converter 5133 may insert a value obtained by copying the value of an adjacent pixel between pixels to enlarge the size of the pre-conversion feature PC. The converter 5133 may also adjust the size of the pre-conversion feature PC using other methods, without being limited to these. For example, the conversion unit 5133 may increase the size of the pre-conversion feature quantity PC by interpolation using a nearest neighbor method, a linear interpolation method, a bilinear method, a bicubic method, or the like.

[0097] The conversion unit 5133 may perform convolution processing on the interpolated feature using a conversion filter with a stride of 1. Here, the stride refers to the interval at which convolution is applied, and convolution processing with a stride of 1 refers to performing convolution processing by moving the conversion filter at one pixel interval.

[0098] The conversion unit 5133 may perform convolution processing using a filter processing model included in the super-resolution model SM. The filter processing model may be a model capable of outputting a resolution-converted image RI using a conversion filter when a pre-conversion feature PC is input. The filter processing model may be configured, for example, by machine learning, so that when a pre-conversion feature PC is input, the filter processing model can output a resolution-converted image RI using a conversion filter. The conversion unit 5133 may input the pre-conversion feature PC to the filter processing model and cause it to output the resolution-converted image RI. The number of convolution layers realized by the filter processing model is not limited to one layer, and may be multiple layers. In this case, an activation layer such as a ReLU function may be inserted after each convolution layer.

[0099] The conversion unit 5133 may perform a filter process other than the convolution process. For example, the conversion unit 5133 may generate a filter feature having the same size as the pre-conversion feature PC and output the element product of the pre-conversion feature PC and the filter feature. In this case, the number of layers realized by the filter process model is not limited to one layer, and may be multiple layers. Furthermore, multiple layers may be used that combine these layers with activation layers.

[0100] The post-conversion feature amount extraction unit 214 extracts post-conversion feature amounts OC, which are feature amounts of the resolution-converted image RI (step S56). [5-3: Technical Effects of Iris Recognition Device 5]

[0101] The iris authentication device 5 of the fifth embodiment estimates and generates a conversion filter for each magnification ratio of the super-resolution processing. Therefore, a single super-resolution model SM can perform super-resolution processing corresponding to various magnification ratios. The iris authentication device 5 of the fifth embodiment is particularly effective when the resolution of the resolution-converted image RI is fixed. That is, the super-resolution model SM used in the iris authentication device 5 of the fifth embodiment can output a resolution-converted image RI of a desired resolution, regardless of the resolution of the iris image LI. Therefore, by applying the super-resolution model SM, which has been trained and constructed so as to be able to output a resolution-converted image RI corresponding to each existing iris authentication mechanism, to the existing iris authentication mechanism, the existing iris authentication mechanism can perform iris authentication even when an iris image LI of any resolution is input. [6: Sixth embodiment]

[0102] Next, a sixth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the sixth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 6 to which the sixth embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied. [6-1: Configuration of Iris Recognition Device 6]

[0103] 8 is a block diagram showing the configuration of the iris authentication device 6 in the sixth embodiment. In the following explanation, components that have already been explained are given the same reference numerals, and detailed explanations thereof will be omitted. The iris authentication device 6 includes an iris image acquisition unit 211, a calculation unit 212, a generation unit 613, and a converted feature extraction unit 214.

[0104] The generation unit 613 performs super-resolution processing using the super-resolution model SM to generate a resolution-converted image RI by converting the resolution of the iris image LI according to the magnification ratio. The generation unit 613 includes a feature extraction unit 6131, a magnification ratio feature extraction unit 6132, a synthesis unit 6133, and a conversion unit 6134. Note that the generation unit 613 does not necessarily have to include the conversion unit 6134. [6-2: Super-resolution processing performed by the generation unit 613]

[0105] Next, the super-resolution processing performed by the iris authentication device 6 in the sixth embodiment will be described with reference to Fig. 9. Fig. 9 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device 6 in the sixth embodiment.

[0106] 9, the iris image acquisition unit 211 acquires an iris image including the iris of a living body (step S61). The calculation unit 212 calculates the magnification ratio for the iris image (step S62).

[0107] The feature extraction unit 6131 extracts pre-conversion feature PC, which is a feature of the iris image LI (step S63). The feature extraction unit 6131 may extract the pre-conversion feature PC from the low-resolution iris image LI using a low-resolution feature extraction model included in the super-resolution model SM. The low-resolution feature extraction model may be a model capable of generating feature values suitable for at least one of a feature synthesis process and a filtering process, which will be described later, when the low-resolution iris image LI is input. The low-resolution feature extraction model may be constructed, for example, by machine learning, so as to be capable of outputting feature values suitable for at least one of a feature synthesis process and a filtering process, which will be described later, when the iris image LI is input. The feature extraction unit 6131 may input the iris image LI to the low-resolution feature extraction model and cause it to output the pre-conversion feature PC.

[0108] The magnification factor feature extraction unit 6132 extracts a magnification factor feature R C that is a feature of the magnification factor (step S64). The magnification factor feature extraction unit 6132 may generate a magnification factor feature map that is a feature of the magnification factor. The magnification factor feature extraction unit 6132 may extract the magnification factor feature R C using a magnification factor feature extraction model included in the super-resolution model SM. The magnification factor feature extraction model may be configured to be able to output a magnification factor feature R C that is compatible with at least one of feature synthesis processing and filtering processing, which will be described later, when a magnification factor is input. The magnification factor feature extraction unit 6132 may input the magnification factor to the magnification factor feature extraction model and output the magnification factor feature R C. The magnification factor feature extraction unit 6132 may extract a magnification factor feature R C that is the same size as the pre-conversion feature PC.

[0109] The synthesis unit 6133 synthesizes the pre-conversion feature PC and the magnification factor feature RC to convert the pre-conversion feature PC (step S65). The synthesis unit 6133 may synthesize the pre-conversion feature PC and the magnification factor feature RC to generate a synthesized feature. The synthesis unit 6133 may convert the pre-conversion feature PC into a feature independent of the magnification factor through synthesis. The synthesis unit 6133 may perform any of combination, element sum, and element multiplication. The synthesis unit 6133 may synthesize the feature map of the iris image LI and the magnification factor feature map. In this case, the magnification factor feature map generated by the magnification factor feature extraction unit 6132 may have a size of (Cf×h×w). Cf may be, for example, the same number as the number of channels of the pre-conversion feature PC. The synthesis unit 6133 may combine the feature map of the iris image LI and the magnification factor feature map via the channels to generate a synthesized feature map.

[0110] The conversion unit 6134 generates a resolution-converted image RI (step S66). The conversion unit 6134 may generate the resolution-converted image RI using a filtering model included in the super-resolution model SM. The filtering model may be a model capable of outputting the resolution-converted image RI using a conversion filter when converted pre-conversion features PC (combined features) are input. The filtering model may be configured, for example, by machine learning, to be capable of outputting the resolution-converted image RI using the conversion filter when converted pre-conversion features PC (combined features) are input. The conversion filter may be a filter independent of the magnification ratio and can be used regardless of the number of pixels of the iris image LI. The conversion unit 6134 may input the converted pre-conversion features PC (combined features) to the filtering model and cause it to output the resolution-converted image RI. The conversion unit 6134 may output the resolution-converted image RI by performing convolution processing on the combined features. The conversion unit 6134 may perform convolution processing using one convolution layer. Alternatively, the convolution layer may be multiple layers, and the conversion unit 6134 may perform convolution processing using multiple layers that combine convolution layers and activation layers.

[0111] It should be noted that the generation unit 613 does not necessarily have to include an independent conversion unit 6134. The synthesis unit 6133 may generate a resolution-converted image RI by synthesizing a pre-conversion feature PC and a magnification factor feature RC to convert the pre-conversion feature PC and performing convolution processing on the converted pre-conversion feature PC. The synthesis unit 6133 may generate the resolution-converted image RI using the above-mentioned filter processing model.

[0112] The post-conversion feature amount extraction unit 214 extracts post-conversion feature amounts OC, which are feature amounts of the resolution-converted image RI (step S67). [6-3: Technical Effects of Iris Recognition Device 6]

[0113] The iris authentication device 6 of the sixth embodiment can perform super-resolution processing corresponding to various magnification ratios using a single super-resolution model SM by combining the pre-conversion feature amount PC and the magnification ratio feature amount RC. According to the iris authentication device 6 of the sixth embodiment, the magnification ratio feature amount extraction unit 6132 can extract the magnification ratio feature amount RC according to the magnification ratio, so that a resolution-converted image RI can be generated using a common conversion filter independent of the magnification ratio. The iris authentication device 6 of the sixth embodiment is particularly effective when the resolution of the resolution-converted image RI is fixed. That is, the super-resolution model SM used in the iris authentication device 6 of the sixth embodiment can output a resolution-converted image RI of a desired resolution, regardless of the resolution of the iris image LI, similar to the iris authentication device 5 of the fifth embodiment. Therefore, by applying the super-resolution model SM, which has been trained and constructed to be able to output a resolution-converted image RI corresponding to each existing iris authentication mechanism, to the existing iris authentication mechanism, the existing iris authentication mechanism can perform iris authentication with high accuracy even when an iris image LI of any resolution is input. [7: Seventh embodiment]

[0114] Next, a seventh embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the seventh embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 7 to which the seventh embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied. [7-1: Configuration of iris authentication device 7]

[0115] 10 is a block diagram showing the configuration of the iris authentication device 7 in the seventh embodiment. In the following explanation, components that have already been explained are given the same reference numerals, and detailed explanations thereof will be omitted. The iris authentication device 7 includes an iris image acquisition unit 211, a calculation unit 212, a generation unit 713, and a converted feature extraction unit 214.

[0116] The generation unit 713 uses the super-resolution model SM to perform super-resolution processing to generate a resolution-converted image RI by converting the resolution of the iris image LI according to the magnification ratio. The generation unit 713 includes a feature extraction unit 7131, a quantization unit 7132, a filter generation unit 7133, a conversion unit 7134, and a reduction unit 7135.

[0117] The iris authentication device 7 in the seventh embodiment differs from the iris authentication device 5 in the fifth embodiment in that it includes a quantization unit 7132 before the filter generation unit 7133 and a reduction unit 7135 after the conversion unit 7134. [7-2: Super-resolution processing performed by the generation unit 713]

[0118] Next, the super-resolution processing performed by the iris authentication device 7 in the seventh embodiment will be described with reference to Fig. 11. Fig. 11 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device 7 in the seventh embodiment.

[0119] 11, the iris image acquisition unit 211 acquires an iris image LI including the iris of a living body (step S71). The calculation unit 212 calculates the magnification ratio for the iris image LI (step S72).

[0120] The feature extraction unit 7131 extracts pre-conversion feature PC, which is a feature of the iris image LI (step S73). The feature extraction unit 7131 may extract the pre-conversion feature PC from the low-resolution iris image LI using a low-resolution feature extraction model included in the super-resolution model SM. The low-resolution feature extraction model may be a model that, when the low-resolution iris image LI is input, is capable of outputting feature values that are suitable for the filter processing described below. The low-resolution feature extraction model may be constructed, for example, by machine learning, so that, when the iris image LI is input, it is capable of outputting feature values that are suitable for the filter processing described below. The feature extraction unit 7131 may input the iris image LI to the low-resolution feature extraction model and have it output the pre-conversion feature PC.

[0121] The quantization unit 7132 quantizes the enlargement rate to a predetermined enlargement rate (step S74). The quantization unit 7132 may quantize the input enlargement rate to a value that is a power of 2, such as 2, 4, or 8. In this case, when an enlargement rate of 1.5 times is input, the quantization unit 7132 can output 2 times. Specifically, the quantization unit 7132 quantizes the enlargement rate R by 2. n-1 <R<2 n Search for n such that 2 n may be output as the quantization expansion rate. Note that the predetermined expansion rate does not have to be a power of 2, and any power value such as a power of 1.5 or a power of 2.5 may be used. Furthermore, the predetermined expansion rate does not have to be a value expressed by a power, and other discrete values such as multiples of 2 may be used.

[0122] The filter generation unit 7133 generates one or more transform filters for transforming the pre-transform feature PC according to the quantized magnification ratio (step S75). The filter generation unit 7133 according to the seventh embodiment differs from the filter generation unit 5132 according to the fifth embodiment, which may receive a continuous magnification ratio as an input, in that the filter generation unit 7133 receives a discrete magnification ratio as an input. The filter generation unit 7133 may generate one or more transform filters according to the magnification ratio using a transform filter generation model included in the super-resolution model SM. The transform filter generation model may be a model capable of generating a transform filter suitable for the filter processing described below when the quantized magnification ratio is input. The transform filter generation model may be configured, for example, by machine learning, to output a transform filter suitable for the filter processing described below when the quantized magnification ratio is input. The filter generation unit 7133 may input the magnification ratio quantized by the quantization unit 7132 to the transform filter generation model and cause it to output one or more transform filters.

[0123] The filter generation unit 7133 does not generate conversion filters corresponding to various magnification ratios, but generates conversion filters corresponding to quantized magnification ratios. That is, the conversion filter generation model is constructed by learning to generate conversion filters specialized for limited magnification ratios. As such, the conversion filter generation model in the seventh embodiment is constructed by learning specialized for limited magnification ratios, and therefore, by using the conversion filters generated by the filter generation unit 7133 using the conversion filter generation model, it is possible to achieve more accurate super-resolution processing.

[0124] The conversion unit 7134 generates a first resolution-converted image by converting the pre-conversion feature PC through filter processing using one or more conversion filters (step S76). The conversion unit 7134 may adjust the size of the pre-conversion feature PC according to the magnification rate before performing the filter processing. The conversion unit 7134 may perform convolution processing on the interpolated feature using a conversion filter with a stride of 1. The conversion unit 7134 may generate the first resolution-converted image using a filter processing model included in the super-resolution model SM. The filter processing model may be a model that is capable of outputting the first resolution-converted image using a conversion filter when the pre-conversion feature PC is input. The filter processing model may be configured, for example, by machine learning, so that it is capable of outputting the first resolution-converted image using a conversion filter when the pre-conversion feature PC is input. The conversion unit 7134 may input the pre-conversion feature PC to the filter processing model and cause it to output the first resolution-converted image. The number of convolution layers realized by the filter processing model is not limited to one, but may be multiple. In this case, an activation layer such as a ReLU function may be inserted after each convolution layer.

[0125] The reduction unit 7135 reduces the first resolution-converted image to generate a second resolution-converted image in which the number of pixels in the iris region is the same as the desired number of pixels (step S77). For example, if the magnification rate is 1.5 and the quantized magnification rate is 2, the reduction unit 7135 downsamples the first resolution-converted image, which has been subjected to 2x super-resolution processing on the iris image LI, to a second resolution-converted image with 1.5 times the number of pixels of the iris image LI. The reduction unit 7135 may perform downsampling using a general thinning process or the like.

[0126] The post-conversion feature amount extraction unit 214 extracts post-conversion feature amounts OC, which are feature amounts of the resolution-converted image RI (step S78). [7-3: Technical Effects of Iris Recognition Device 7]

[0127] The iris authentication device 7 of the seventh embodiment estimates and generates a conversion filter corresponding to a quantized magnification ratio of the super-resolution processing. Therefore, a single super-resolution model SM can perform super-resolution processing corresponding to various magnification ratios. The iris authentication device 7 of the seventh embodiment upsamples to 2x, 4x, 8x, etc. using a conversion filter corresponding to the magnification ratio, and then downsamples from that size, thereby achieving continuous magnification ratios with high accuracy. The iris authentication device 7 of the seventh embodiment is particularly effective when the resolution of the resolution-converted image RI is fixed. That is, the super-resolution model SM used in the iris authentication device 7 of the seventh embodiment can output a resolution-converted image RI of a desired resolution, regardless of the resolution of the iris image LI, similar to the iris authentication devices 5 and 6 of the fifth and sixth embodiments. Therefore, by applying a super-resolution model SM, which has been trained and constructed to be able to output a resolution-converted image RI corresponding to each existing iris authentication mechanism, to the existing iris authentication mechanism, the existing iris authentication mechanism can perform iris authentication with high accuracy even when an iris image LI of any resolution is input. [8: Eighth embodiment]

[0128] Next, an eighth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described. Hereinafter, the eighth embodiment of the iris authentication device, the iris authentication method, and the recording medium will be described using an iris authentication device 8 to which the eighth embodiment of the iris authentication device, the iris authentication method, and the recording medium is applied. [8-1: Configuration of Iris Recognition Device 8]

[0129] 12 is a block diagram showing the configuration of the iris authentication device 8 in the eighth embodiment. In the following explanation, components that have already been explained are given the same reference numerals, and detailed explanations thereof will be omitted. The iris authentication device 8 includes an iris image acquisition unit 211, a calculation unit 212, a generation unit 213, a converted feature extraction unit 214, an authentication unit 215, and an adjustment unit 819. [8-2: Iris authentication operation performed by iris authentication device 8]

[0130] Next, the super-resolution processing performed by the iris authentication device 8 in the eighth embodiment will be described with reference to Fig. 13. Fig. 13 is a flowchart showing the flow of the super-resolution processing performed by the iris authentication device 8 in the eighth embodiment.

[0131] 13, in the eighth embodiment, as in the second embodiment, the iris authentication device 8 performs the operations from step S21 to step S27. In the eighth embodiment, before the authentication operation by the authentication unit 215 in step S26, the adjustment unit 819 adjusts the threshold value used for authentication by the authentication unit 215 according to the magnification ratio (step S81). That is, in the eighth embodiment, the difficulty level for the authentication unit 215 to authenticate the person is adjusted according to the magnification ratio. The authentication unit 215 authenticates the person as the person in question if the matching score indicating the similarity between the transformed features extracted by the transformed feature extraction unit 214 and the pre-prepared matching features is equal to or greater than the threshold adjusted by the adjustment unit 819 (step S26). [8-3: Technical Effects of Iris Recognition Device 8]

[0132] In iris authentication using an iris image that has been subjected to super-resolution processing, it may be preferable to change the authentication certainty depending on the magnification rate of the super-resolution processing. Since the iris authentication device 8 in the eighth embodiment can adjust the threshold value used by the authentication unit 215 for authentication depending on the magnification rate, even when it is preferable to change the authentication certainty depending on the magnification rate, it is possible to adjust the difficulty of authenticating the person by adjusting the threshold value used for authentication. [9: Ninth embodiment]

[0133] Next, a ninth embodiment of the iris authentication device, the iris authentication system, the iris authentication method, and the recording medium will be described. Hereinafter, the ninth embodiment of the iris authentication device, the iris authentication system, the iris authentication method, and the recording medium will be described using an authentication system 100 to which the ninth embodiment of the iris authentication device, the iris authentication system, the iris authentication method, and the recording medium is applied. [9-1: Configuration of the iris authentication system 100]

[0134] Fig. 14 is a block diagram showing the configuration of an iris authentication system 100 in the ninth embodiment. In the following description, components that have already been described are given the same reference numerals, and detailed description thereof will be omitted. As shown in Fig. 14, the iris authentication system 100 includes a first device 101 and a second device 102. The first device 101 includes an iris image acquisition unit 11 and a calculation unit 12. The second device 102 includes a generation unit 13 and a converted feature extraction unit 14.

[0135] That is, iris image acquisition unit 11, which is a specific example of "iris image acquisition means," calculation unit 12, which is a specific example of "calculation means," generation unit 13, which is a specific example of "generation means," and converted feature extraction unit 14, which is a specific example of "converted feature extraction means," may be provided in different devices. For example, first device 101 may include only iris image acquisition unit 11, and second device 102 may include calculation unit 12, generation unit 13, and converted feature extraction unit 14. Alternatively, iris image acquisition unit 11, calculation unit 12, generation unit 13, and converted feature extraction unit 14 may be provided in different combinations in first device 101 and second device 102. The first device 101 and the second device 102 are capable of communicating with each other and can transmit and receive their respective processing results. As shown in FIG. 14 , an example will be described in which the first device 101 includes an iris image acquisition unit 11 and a calculation unit 12, and the second device 102 includes a generation unit 13 and a converted feature extraction unit 14. In this case, the first device 101 can transmit the calculation result of the calculation unit 12 to the second device 102, and the second device 102 receives the calculation result. The generation unit 13 can generate a resolution-converted image by converting the resolution of the iris image in accordance with the calculation result. Furthermore, the iris authentication system may include three or more devices, and the iris image acquisition unit 11, the calculation unit 12, the generation unit 13, and the converted feature extraction unit 14 may be provided in any combination in each device.

[0136] In the above embodiment, an iris image has been described as an example, but this super-resolution technology can also be applied to other image processing fields such as face recognition, etc. Furthermore, in the second embodiment and onward, cases where the magnification is equal to or greater than 1 have been described, but the magnification is not limited to equal to or greater than 1 and may be less than 1.

[0137] Furthermore, in the iris authentication device in the above embodiment, the magnification is determined based on the number of pixels in the iris region included in the iris image, but the magnification may be determined independently of the number of pixels in the iris region. For example, the magnification used by the iris authentication device for resolution conversion may be determined based on the distance between the imaging device and the living body when capturing the iris image. The magnification used by the iris authentication device may be any magnification that allows appropriate resolution conversion. [10: Note]

[0138] The following additional notes are provided regarding the above-described embodiment. [Appendix 1] an iris image acquisition means for acquiring an iris image including the iris of a living body; a calculation means for calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; a generation means for generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; a post-conversion feature extraction means for extracting post-conversion feature values that are feature values of the resolution-converted image; An iris authentication device comprising: [Appendix 2] the size of the iris region is smaller than the desired size; The magnification is a magnification ratio, The generating means generates a super-resolution image by increasing the resolution of the iris image in accordance with the magnification ratio. 10. The iris authentication device according to claim 1. [Appendix 3] a training image acquisition means for acquiring a training image including the iris region of the desired size; and an input image generating means for generating an input image in which the resolution of the learning image is converted in accordance with the reciprocal of an arbitrary magnification; the generating means generates a resolution-converted input image having the same resolution as the learning image by converting the resolution of the input image according to the arbitrary magnification; The image processing device further includes a learning unit that causes the generating unit to learn a method for generating the resolution-converted image based on a loss function in which the loss increases as the training image and the resolution-converted input image become less similar. 3. The iris authentication device according to claim 1 or 2. [Appendix 4] a training image acquisition means for acquiring a training image including the iris region of the desired size; and an input image generating means for generating an input image in which the resolution of the learning image is converted in accordance with the reciprocal of an arbitrary magnification; the generating means generates a resolution-converted input image having the same resolution as the learning image by converting the resolution of the input image according to the arbitrary magnification; the post-conversion feature extraction means extracts training features that are features of the training image and input features that are features of the resolution-converted input image; The image processing device further includes a learning unit that causes the generating unit to learn a method for generating the resolution-converted image based on a loss function in which the loss increases as the training feature and the input feature become less similar. An iris authentication device according to any one of appendices 1 to 3. [Appendix 5] The learning means causes the generation means to perform learning based on a loss function in which the weight of the loss corresponding to the input image generated by using a first magnification as the arbitrary magnification is greater than the weight of the loss corresponding to the input image generated by using a second magnification smaller than the first magnification as the arbitrary magnification. 5. The iris authentication device according to claim 3 or 4. [Appendix 6] The input image generating means generates the plurality of input images so that the frequency of generating the input images by using a first magnification as the arbitrary magnification is higher than the frequency of generating the input images by using a second magnification, which is smaller than the first magnification, as the arbitrary magnification. An iris authentication device according to any one of appendices 3 to 5. [Appendix 7] The generating means a pre-transformation feature extraction means for extracting pre-transformation feature values that are feature values of the iris image; a filter generating means for generating one or more transform filters for transforming the pre-transform feature quantity in accordance with the magnification; a conversion means for converting the pre-conversion feature quantity by one or more filter processes using the one or more conversion filters, thereby generating the resolution-converted image; Contains An iris authentication device according to any one of appendices 1 to 6. [Appendix 8] The generating means a pre-transformation feature extraction means for extracting pre-transformation feature values that are feature values of the iris image; a magnification feature extraction means for extracting a magnification feature that is a feature of the magnification; a conversion means for generating the resolution-converted image by combining the pre-conversion feature amount and the magnification feature amount and converting the pre-conversion feature amount; Contains An iris authentication device according to any one of appendices 1 to 7. [Appendix 9] The generating means a pre-transformation feature extraction means for extracting pre-transformation feature values that are feature values of the iris image; quantization means for quantizing the magnification to a predetermined magnification; a filter generating means for generating one or more transform filters for transforming the pre-transform feature quantity in accordance with the quantized scaling factor; a conversion means for converting the pre-conversion feature quantity by one or more filter processes using the one or more conversion filters to generate a first resolution-converted image; a reduction means for reducing the first resolution-converted image to generate a second resolution-converted image in which the size of the iris region is the same as the desired size; Contains An iris authentication device according to any one of appendices 1 to 8. [Appendix 10] a determination means for determining that the person is the person in question when a matching score indicating a similarity between the converted feature extracted by the converted feature extracting means and a pre-prepared matching feature is equal to or greater than a threshold; an adjusting means for adjusting the threshold value in accordance with the magnification; Further provided with An iris authentication device according to any one of appendices 1 to 9. [Appendix 11] an authentication means for performing authentication using a score indicating a similarity between the transformed feature and a pre-prepared feature; an output means for outputting the authentication result of the authentication means, the magnification, and the resolution-converted image to the outside of the iris authentication device, and outputting an alert if the magnification is equal to or greater than a predetermined value; Further provided with An iris authentication device according to any one of appendices 1 to 10. [Appendix 12] an iris image acquisition means for acquiring an iris image including the iris of a living body; a calculation means for calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; a generation means for generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; a post-conversion feature extraction means for extracting post-conversion feature values that are feature values of the resolution-converted image; An iris authentication system comprising: [Appendix 13] acquiring an iris image including a biological iris; calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; Extracting post-conversion feature quantities, which are feature quantities of the resolution-converted image. Iris authentication method. [Appendix 14] On the computer, acquiring an iris image including a biological iris; calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; Extracting post-conversion feature quantities, which are feature quantities of the resolution-converted image. A recording medium on which a computer program for executing an iris authentication method is recorded.

[0139] At least some of the constituent elements of each of the above-described embodiments can be appropriately combined with at least some of the other constituent elements of each of the above-described embodiments. Some of the constituent elements of each of the above-described embodiments may not be used. Furthermore, to the extent permitted by law, the disclosures of all documents (e.g., published patent applications) cited in this disclosure are incorporated by reference as part of the description of this disclosure.

[0140] This disclosure may be modified as appropriate within the scope of the claims and the technical idea that can be read from the entire specification. Iris authentication devices, iris authentication systems, iris authentication methods, and recording media that incorporate such modifications are also included in the technical idea of this disclosure. [Explanation of symbols]

[0141] 1,2,3,5,6,7,8 Iris authentication device 11,211 Iris image acquisition unit 12,212 Calculation section 2121 Iris circle detection unit 2122 Magnification ratio calculation unit 13,213,513,613,713 Generation part 14,214 Post-conversion feature extraction section 215 Authentication Department 21 Arithmetic unit 22 Storage device 300 Iris Information Estimation Unit 316 Learning Image Acquisition Unit 317 Input Image Generation Unit 3171 Batch Data Extraction Unit 3172 Resolution conversion unit 318 Learning Department 3181 Loss Function Calculation Unit 3182 Gradient Calculation Unit 3183 Parameter Update Unit 5131,6131,7131 Feature extraction unit 5132,7133 Filter generation section 5133,6134,7134 conversion unit 6132 Magnification feature extraction unit 6133 Synthesis Department 7132 Quantization section 7135 Reduction section 819 Adjustment section 100 Iris Recognition System SM Super Resolution Model GM feature generation model LI,H1 Iris image RI resolution conversion image Pre-PC feature OC transformed features TI training images TC learning feature II Input image RII Resolution conversion input image OIC input features RC magnification feature CC matching feature

Claims

1. an iris image acquisition means for acquiring an iris image including the iris of a living body; a calculation means for calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; a generation means for generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; a post-conversion feature extraction means for extracting post-conversion feature values that are feature values of the resolution-converted image; An iris authentication device comprising:

2. the size of the iris region is smaller than the desired size; The magnification is a magnification, The generating means generates a super-resolution image by increasing the resolution of the iris image in accordance with the enlargement ratio as the resolution-converted image. The iris authentication device according to claim 1 .

3. a training image acquisition means for acquiring a training image including the iris region of the desired size; and an input image generating means for generating an input image in which the resolution of the learning image is converted in accordance with the reciprocal of an arbitrary magnification; the generating means generates a resolution-converted input image having the same resolution as the learning image by converting the resolution of the input image according to the arbitrary magnification; The image processing device further includes a learning unit that causes the generating unit to learn a method for generating the resolution-converted image based on a loss function in which the loss increases as the similarity between the training image and the resolution-converted input image decreases.

3. The iris authentication device according to claim 1.

4. a training image acquisition means for acquiring a training image including the iris region of the desired size; and an input image generating means for generating an input image in which the resolution of the learning image is converted in accordance with the reciprocal of an arbitrary magnification; the generating means generates a resolution-converted input image having the same resolution as the learning image by converting the resolution of the input image according to the arbitrary magnification; the converted feature extraction means extracts training features that are features of the training image and input features that are features of the resolution-converted input image; The image processing device further includes a learning unit that causes the generating unit to learn a method for generating the resolution-converted image based on a loss function in which the loss increases as the training feature and the input feature become less similar.

3. The iris authentication device according to claim 1.

5. The learning means causes the generation means to perform learning based on a loss function in which the weight of the loss corresponding to the input image generated by using a first magnification as the arbitrary magnification is greater than the weight of the loss corresponding to the input image generated by using a second magnification smaller than the first magnification as the arbitrary magnification. The iris authentication device according to claim 3.

6. The input image generating means generates the plurality of input images so that the frequency of generating the input images by using a first magnification as the arbitrary magnification is higher than the frequency of generating the input images by using a second magnification, which is smaller than the first magnification, as the arbitrary magnification. The iris authentication device according to claim 3.

7. The generating means a pre-transformation feature extraction means for extracting pre-transformation feature values that are feature values of the iris image; a filter generating means for generating one or more transform filters for transforming the pre-transform feature quantity in accordance with the magnification; a conversion means for converting the pre-conversion feature quantity by one or more filter processes using the one or more conversion filters, thereby generating the resolution-converted image; Contains 3. The iris authentication device according to claim 1.

8. an iris image acquisition means for acquiring an iris image including the iris of a living body; a calculation means for calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; a generation means for generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; a post-conversion feature extraction means for extracting post-conversion feature values that are feature values of the resolution-converted image; An iris authentication system comprising:

9. acquiring an iris image including a biological iris; calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; Extracting post-conversion feature quantities, which are feature quantities of the resolution-converted image. Iris authentication method.

10. On the computer, acquiring an iris image including a biological iris; calculating a magnification for the iris image based on a size of an iris region included in the iris image and a desired size; generating a resolution-converted image by converting the resolution of the iris image in accordance with the magnification; Extracting post-conversion feature quantities, which are feature quantities of the resolution-converted image. A computer program for performing an iris recognition method.

Images