Information processing system, information processing method, and recording medium

Abstract

This information processing system comprises: an extracting means that extracts, on the basis of an eye image including an iris region of a subject, an iris feature indicating a feature of an iris region and a partial iris feature indicating a feature of a non-masked portion of the iris region that is not masked by a masking object; and an authentication means that authenticates the subject on the basis of: weights estimated on the basis of the eye image; the iris feature; and the partial iris feature.

Description

Information processing system, information processing method, and recording medium

[0001] The present disclosure relates to the technical fields of an information processing system, an information processing method, and a recording medium.

[0002] As an example of this type of system, a system has been proposed that performs iris authentication using an image including the eyes of a subject wearing color contact lenses (see Patent Document 1).

[0003] International Publication No. 2023 / 047572

[0004] An object of this disclosure is to provide an information processing system, an information processing method, and a recording medium that aim to improve upon the techniques described in prior art documents.

[0005] One aspect of the information processing system includes an extraction means that extracts, based on an eye image including an iris region of a target, iris features that indicate characteristics of the iris region and partial iris features that indicate characteristics of an unobstructed portion of the iris region that is not obscured by an obstructing object, and an authentication means that authenticates the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0006] One aspect of the information processing method extracts, based on an eye image including an iris region of a target, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an obstructing object, and authenticates the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0007] One aspect of the recording medium has recorded thereon a computer program that causes a computer to execute an information processing method that extracts, based on an eye image including an iris region of a target, iris features that indicate characteristics of the iris region and partial iris features that indicate characteristics of an unobstructed portion of the iris region that is not obscured by an obstructing object, and authenticates the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0008] FIG. 1 is a diagram illustrating an example of the configuration of an information processing system. FIG. 2 is a diagram illustrating the concept of an iris image. FIG. 3 is a flowchart illustrating an example of the operation of the information processing system. FIG. 4 is a diagram illustrating another example of the configuration of the information processing system. FIG. 5 is a flowchart illustrating another example of the operation of the information processing system. FIG. 6 is a diagram illustrating the concept of the learning operation of a weighted score estimation unit. FIG. 7 is a diagram illustrating another example of the configuration of an information processing system. FIG. 8 is a diagram illustrating an example of the range of a partial iris image.

[0009] First Embodiment A first embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to Figures 1 to 3. Hereinafter, the first embodiment of the information processing system, the information processing method, and the recording medium will be described using an information processing system 1.

[0010] 1, the information processing system 1 includes an extraction unit 11 and an authentication unit 12. The extraction unit 11 extracts iris features and partial iris features based on an eye image including a target iris region. The iris features are comparable features that indicate the features of the iris region. The partial iris features are comparable features that indicate the features of an unoccluded portion of the iris region. The unoccluded portion is a portion of the iris region that is not occluded by an occluding object.

[0011] Here, the iris region will be described with reference to FIGS. 2( a) and 2(b). FIG. 2(a) is an example of an eye image including the eye of a target. An example of a method for estimating the iris region from an eye image will be described below. As shown in FIG. 2(a), the iris 51 contains the pupil 52, and therefore the shape of the iris 51 is annular. The outer edge of the iris 51 can be considered to be circular. Similarly, the shape of the pupil 52 can also be considered to be circular. Furthermore, the inner edge of the iris 51 is the same as the edge of the pupil 52. Therefore, the position of the iris 51 can be expressed by the center coordinates and outer radius of the pupil 52 and the center coordinates and outer radius of the outer edge of the iris 51 (i.e., the iris circle).

[0012] The outer radius of the iris 51 may be estimated as follows: The outer edge of the iris 51 may be detected based on a change in brightness between the iris 51 and the sclera (the so-called white of the eye) in the eye image. The outer radius of the iris 51 may be estimated by approximating the outer edge of the detected iris 51 with a circle. The inner radius of the iris 51 may be estimated as follows: The pupil 52 may be detected by extracting pixels with a brightness lower than a predetermined value in the eye image. The inner radius of the iris 51 (in other words, the radius of the pupil 52) may be estimated by approximating the edge of the detected pupil 52 with a circle. The central coordinates of the iris 51 may be estimated based on the outer edge of the detected iris 51 or based on the detected pupil 52.

[0013] A portion corresponding to the position of the iris 51, represented by the center coordinates, outer radius, and inner radius of the iris 51, may be extracted from the eye image as the iris region. An image including the extracted iris region may be referred to as an iris image. FIG. 2B shows an iris image 510 as an example of an iris image. In FIG. 2B, the region indicated by reference numeral 51a is an example of the iris region. Note that the method for extracting the iris region from the eye image is not limited to the above-described method, and other methods may also be used. For example, parameters related to the iris region (e.g., the center coordinates, outer radius, and inner radius of the iris 51) may be directly estimated from the eye image using a neural network.

[0014] The eye image may be generated, for example, by irradiating the subject's eye with infrared light and receiving the infrared light reflected from the subject's eye with a camera sensitive to infrared light.The eye image may also be generated by receiving visible light reflected from the subject's eye with a camera sensitive to visible light.

[0015] Next, the non-occluded region will be described with reference to FIGS. 2(c) and 2(d). FIG. 2(c) is an example of an eye image including the eye of a subject wearing colored contact lenses. Note that colored contact lenses may also be referred to as tinted contact lenses. As shown in FIG. 2(c), a portion of the iris 51 is occluded by a colored region 60 of the colored contact lens. The colored contact lens has a colored region and a transparent region, as shown in FIG. 2(d). Generally, the colored region is located near the outer edge of the iris.

[0016] 2(d) shows iris image 510a as another example of an iris image. As described above, when the outer edge of iris 51 is detected based on the change in brightness between iris 51 and the sclera, in a subject wearing color contact lenses, the outer edge of colored region 60 of the color contact lenses is often detected as the outer edge of iris 51. For this reason, iris image 510a includes an area corresponding to colored region 60 of the color contact lenses.

[0017] 2(d), the region indicated by reference symbol 51b is an example of a non-occluded region. For example, non-occluded region 51b may be estimated by estimating a portion of iris image 510a corresponding to colored region 60 based on a luminance value. For example, the range occupied by colored region 60 in a color contact lens may be determined in advance based on commercially available color contact lenses. Based on the determined range, a predetermined range of iris image 510a may be designated as non-occluded region 51b.

[0018] As described above, the iris region may be, for example, a portion of the eye image that corresponds to the position of the iris 51, which is expressed by the central coordinates, outer radius, and inner radius of the iris 51. For this reason, the iris region may include something other than the iris 51 (for example, the colored region 60 of a color contact lens, the eyelid, etc.). For this reason, the iris feature amount, which is a feature amount indicating the characteristics of the iris region, may include, in addition to the feature amount of the iris 51, feature amounts of something other than the iris 51. For example, in the case of the iris image 510a, the iris feature amount may also include feature amounts related to the portion that corresponds to the colored region 60 included in the iris image 510a.

[0019] For example, the extraction unit 11 may extract iris feature amounts and partial iris feature amounts based on an iris image (e.g., at least one of the iris images 510 and 510a). For example, a circular iris region included in the iris image (e.g., at least one of the iris images 510 and 510a) may be expanded in polar coordinates. Specifically, as shown in FIG. 2E, the circular iris region 51a included in the iris image 510 may be converted into a rectangular image 520 by polar coordinate expansion. For example, the extraction unit 11 may extract iris feature amounts based on the iris region expanded in polar coordinates. Similarly, the non-occluded portion (e.g., the non-occluded portion 51b) may also be expanded in polar coordinates. For example, the extraction unit 11 may extract partial iris feature amounts based on the non-occluded portion expanded in polar coordinates.

[0020] In particular, the extraction unit 11 extracts both iris feature amounts and partial iris feature amounts even when the subject is not wearing color contact lenses. In this case, the unoccluded portion (i.e., the portion of the iris region that is not occluded by an occluding object) may be the same as the iris region. In other words, the iris feature amount and the partial iris feature amount may be the same. For example, a portion of the iris region, and a predetermined range of the iris region, may be considered to be the unoccluded portion. In this case, even if the subject is not wearing color contact lenses, the iris region and the unoccluded portion will be different. Therefore, the iris feature amount and the partial iris feature amount will also be different.

[0021] The authentication unit 12 performs authentication of the target based on the weights estimated based on the eye image and the iris feature amounts and partial iris feature amounts extracted by the extraction unit 11. The weights include weights related to the iris feature amounts and weights related to the partial iris feature amounts.

[0022] The weights associated with the iris feature amounts and the weights associated with the partial iris feature amounts may be estimated as follows. For example, the likelihood of wearing color contact lenses may be calculated based on the luminance value of the iris region included in the eye image. Alternatively, the likelihood of wearing color contact lenses may be calculated from the eye image using a neural network, for example. The weights associated with the iris feature amounts and the weights associated with the partial iris feature amounts may be estimated based on the calculated likelihoods. Specifically, the smaller the likelihood, the larger the weight associated with the iris feature amount may be relative to the weight associated with the partial iris feature amount. The larger the likelihood, the smaller the weight associated with the iris feature amount may be relative to the weight associated with the partial iris feature amount. Note that the method of estimating the weights associated with the iris feature amounts and the weights associated with the partial iris feature amounts is not limited to the above-described method, and other methods may also be used.

[0023] The authentication unit 12 may authenticate the target using the method described below. For example, the authentication unit 12 may apply a weight related to the iris feature to an iris feature, and may also apply a weight related to the partial iris feature to a partial iris feature. The authentication unit 12 may calculate a fusion feature by combining the weighted iris feature and the weighted partial iris feature. The authentication unit 12 may authenticate the target by comparing the fusion feature with a feature registered in advance.

[0024] For example, a feature corresponding to an iris feature and a feature corresponding to a partial iris feature may be registered in advance. The authentication unit 12 may calculate a matching score for the iris feature by comparing the iris feature extracted by the extraction unit 11 with the registered feature (i.e., the feature corresponding to the iris feature). The authentication unit 12 may calculate a matching score for the partial iris feature by comparing the partial iris feature extracted by the extraction unit 11 with the registered feature (i.e., the feature corresponding to the partial iris feature). The authentication unit 12 may calculate a weighted average of the matching score for the iris feature and the matching score for the partial iris feature using the weights for the iris feature and the weights for the partial iris feature. The authentication unit 12 may authenticate the target by comparing the calculated weighted average with a predetermined threshold.

[0025] The operation of the information processing system 1 will be described with reference to the flowchart in Fig. 3. In Fig. 3, the extraction unit 11 extracts iris features based on an eye image of the target (step S101). The extraction unit 11 extracts partial iris features based on the eye image of the target (step S102). Note that the processes of steps S101 and S102 may be performed in parallel. Thereafter, the authentication unit 12 authenticates the target based on the weights estimated based on the eye image and the iris features and partial iris features extracted by the extraction unit 11 (step S103).

[0026] In this way, the information processing system 1 extracts, based on an eye image including the iris region of a target, iris features that indicate the characteristics of the iris region and partial iris features that indicate the characteristics of the non-occluded part of the iris region that is not occluded by an obstructing object, and performs an information processing method to authenticate the target based on the weights estimated based on the eye image, the iris features, and the partial iris features.

[0027] Information processing system 1 may be realized by a computer reading a computer program recorded on a recording medium. In this case, the recording medium stores the computer program that causes the computer to execute an information processing method that extracts, based on an eye image including the iris region of a target, iris features that indicate characteristics of the iris region and partial iris features that indicate characteristics of an unobstructed portion of the iris region that is not obscured by an obstructing object, and authenticates the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0028] (Technical Effect) Unlike the face, which changes over time, the pattern of the iris is extremely unlikely to change. It is extremely difficult to replicate the pattern of the iris. For example, even if most of the face is covered by a covering such as a mask, it is possible to capture an image of the iris as long as the eyes are exposed. For these reasons, iris authentication can be said to be an excellent authentication technology in terms of both security and user convenience.

[0029] The number of people wearing color contact lenses is increasing, especially among young women. When color contact lenses are worn, a portion of the iris is obscured by the colored region of the color contact lenses (e.g., colored region 60 in FIG. 2(c)). As a result, fewer iris-related features are extracted from an eye image of a person wearing color contact lenses than from an eye image of a person not wearing color contact lenses. In many cases, pre-registered features are those obtained when the person is not wearing color contact lenses. When the iris-related features of a person wearing color contact lenses are compared with the registered features, the degree of match between the two is low. As a result, there is a risk that the wearer of color contact lenses will not be properly authenticated.

[0030] In response to this, for example, it is conceivable to switch the authentication method depending on whether the subject is not wearing color contact lenses or is wearing color contact lenses.

[0031] The colored region of a color contact lens is patterned with at least one of dots and lines. For example, the size and color of the dots and the thickness and color of the lines are arbitrary. The density of the dots and the density of the lines are also arbitrary. In other words, the color shade (in other words, the brightness value) of the colored region varies depending on the color contact lens. For this reason, it may not be possible to accurately determine from an eye image whether a subject is wearing color contact lenses. Furthermore, if the density of at least one of the dots and lines is low, the iris pattern may be detected in the portion of the iris region where the colored region is superimposed.

[0032] In the information processing system 1, the authentication unit 12 authenticates the target based on weights (for example, weights related to iris feature amounts and weights related to partial iris feature amounts) estimated based on the eye image and the iris feature amounts and partial iris feature amounts extracted by the extraction unit 11. The concept of authentication in the information processing system 1 is as follows.

[0033] For example, when the weight associated with the iris feature is greater than the weight associated with the partial iris feature, authentication of the subject is performed by placing more emphasis on the degree of match between the extracted feature and the registered feature in the entire iris region than on the degree of match between the extracted feature and the registered feature in the non-region region of the subject, whereas when the weight associated with the partial iris feature is greater than the weight associated with the iris feature, authentication of the subject is performed by placing more emphasis on the degree of match between the extracted feature and the registered feature in the non-region region than on the degree of match between the extracted feature and the registered feature in the entire iris.

[0034] If the subject is wearing color contact lenses, the weight for the partial iris feature is greater than the weight for the iris feature. In this case, subject authentication is performed with emphasis on the degree of match of the feature in the unoccluded part. Therefore, even if the degree of match of the feature in the entire iris region is low when the subject is registered, the subject can be properly authenticated as long as the degree of match of the feature in the unoccluded part is high.

[0035] If the subject is not wearing colored contact lenses, the weighting for the iris feature is greater than the weighting for the partial iris feature. In this case, the subject is authenticated with emphasis on the degree of match of the feature across the entire iris region. If the subject is enrolled, the degree of match of the feature across the entire iris region and the degree of match of the feature across the unoccluded portion are both high. By placing emphasis on the degree of match of the feature across the entire iris region, the reliability of authentication can be improved.

[0036] Additionally, there is no need to determine whether or not the subject is wearing color contact lenses in the information processing system 1. Therefore, compared to a configuration in which the authentication method is switched between when the subject is not wearing color contact lenses and when the subject is wearing color contact lenses, the configuration of the information processing system 1 can be made less complex.

[0037] Therefore, the information processing system 1 can appropriately authenticate a subject regardless of whether the subject is wearing colored contact lenses. Note that, although colored contact lenses are given as an example of an obstructing object in the above-described embodiment, the technical concept of the information processing system 1 can also be applied to cases where part of the iris region is obstructed by, for example, eyelids, glasses, hair, eyelashes, clothing (e.g., hats, turbans, hoods), etc.

[0038] Second Embodiment A second embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to Figures 4 and 5. Below, the second embodiment of the information processing system, the information processing method, and the recording medium will be described using an information processing system 2. Note that, with regard to the second embodiment, descriptions that overlap with the description of the first embodiment will be omitted as appropriate.

[0039] 4 , the information processing system 2 includes a calculation device 21, a storage device 22, and a communication device 23. The information processing system 2 may further include an input device 24 and an output device 25. Note that the information processing system 2 does not necessarily include at least one of the input device 24 and the output device 25. In the information processing system 2, 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.

[0040] The arithmetic device 21 may include, for example, at least one of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), a TPU (Tensor Processing Unit), and a quantum processor.

[0041] The storage device 22 may include, for example, 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 an optical disk array. In other words, the storage device 22 may include a non-transitory recording medium. The storage device 22 is capable of storing 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 is temporarily used by the arithmetic device 21 when the arithmetic device 21 is executing a computer program.

[0042] The communication device 23 may be capable of communicating with devices external to the information processing system 2. The communication device 23 may perform wired communication or wireless communication.

[0043] The input device 24 is a device capable of accepting information input to the information processing system 2 from the outside. The input device 24 may include an operation device (e.g., a keyboard, a mouse, a touch panel, etc.) that can be operated by an operator of the information processing system 2. The input device 24 may include a recording medium reading device that can read information recorded on a recording medium that is detachable from the information processing system 2, such as a USB (Universal Serial Bus) memory. Note that when information is input to the information processing system 2 via the communication device 23 (in other words, when the information processing system 2 obtains information via the communication device 23), the communication device 23 may function as an input device.

[0044] The output device 25 is a device capable of outputting information to the outside of the information processing system 2. The output device 25 may output visual information such as text or images, auditory information such as sound, or tactile information such as vibration, as the information. The output device 25 may include, for example, at least one of a display, a speaker, a printer, and a vibration motor. The output device 25 may be capable of outputting information to a recording medium that is detachable from the information processing system 2, such as a USB memory. Note that when the information processing system 2 outputs information via the communication device 23, the communication device 23 may function as the output device.

[0045] The arithmetic device 21 has, as logically realized functional blocks or as physically realized processing circuits, an image acquisition unit 211, an eye position estimation unit 212, an iris image normalization unit 213, a partial iris image normalization unit 214, an iris feature amount extraction unit 215, a partial iris feature amount extraction unit 216, a weighted score estimation unit 217, and an authentication unit 218. The "iris feature amount extraction unit 215" and the "partial iris feature amount extraction unit 216" are components corresponding to the "extraction unit 11" in the first embodiment. The "authentication unit 218" is a component corresponding to the "authentication unit 12" in the first embodiment.

[0046] At least one of the image acquisition unit 211, eye position estimation unit 212, iris image normalization unit 213, partial iris image normalization unit 214, iris feature extraction unit 215, partial iris feature extraction unit 216, weight score estimation unit 217, and authentication unit 218 may be realized in a form that mixes logical functional blocks and physical processing circuits (i.e., hardware). If at least a portion of the image acquisition unit 211, eye position estimation unit 212, iris image normalization unit 213, partial iris image normalization unit 214, iris feature extraction unit 215, partial iris feature extraction unit 216, weight score estimation unit 217, and authentication unit 218 are functional blocks, at least a portion of the image acquisition unit 211, eye position estimation unit 212, iris image normalization unit 213, partial iris image normalization unit 214, iris feature extraction unit 215, partial iris feature extraction unit 216, weight score estimation unit 217, and authentication unit 218 may be realized by the calculation device 21 executing a predetermined computer program.

[0047] The arithmetic device 21 may acquire (in other words, read) the predetermined computer program from the storage device 22. The arithmetic device 21 may read the predetermined computer program stored in a computer-readable, non-transitory recording medium using a recording medium reading device (not shown) included in the information processing system 2. The arithmetic device 21 may acquire (in other words, download or read) the predetermined computer program from a device (not shown) external to the information processing system 2 via the communication device 23. Note that the recording medium on which the predetermined computer program executed by the arithmetic device 21 is recorded may be at least one of an optical disk, a magnetic medium, a magneto-optical disk, a semiconductor memory, and any other medium capable of storing a program.

[0048] The image acquisition unit 211 acquires an eye image of the subject. The image acquisition unit 211 may acquire the eye image from a camera that captures the subject's eye. Alternatively, the image acquisition unit 211 may acquire the eye image from a device that stores the eye image (for example, the storage device 22 or a server device different from the information processing system 1).

[0049] The eye position estimation unit 212 estimates the positions of the target's eyes from the eye images acquired by the image acquisition unit 211. For example, the eye position estimation unit 212 may estimate the positions of the iris and pupils from the eye images as the positions of the target's eyes. The eye position estimation unit 212 may generate an iris image based on the estimated eye positions (for example, the positions of the iris and pupils). Note that a component different from the eye position estimation unit 212 may generate the iris image based on the eye positions estimated by the eye position estimation unit 212.

[0050] The iris image normalization unit 213 performs polar coordinate expansion on the iris region included in the iris image. Polar coordinate expansion is a process of converting the annular iris region 51a into a rectangular image 520, as shown in FIG. 2( e), for example. This process may be performed by, for example, converting the iris region 51a from an xy plane coordinate system to an rθ polar coordinate system. The iris image normalization unit 213 may normalize the image 520, for example. Note that the image generated by the iris image normalization unit 213 performing polar coordinate expansion on the iris region will hereinafter be referred to as an "iris-normalized image" as appropriate.

[0051] Here, the bottom edge of the image 520 shown in FIG. 2E corresponds to the inner edge of the iris 51. The top edge of the image 520 corresponds to the outer edge of the iris 51. The inner edge of the iris 51 may also be referred to as the pupil circle. The outer edge of the iris 51 may also be referred to as the iris circle. The inner radius of the iris 51 corresponds to the radius of the pupil 52. The radius of the pupil 52 may change due to, for example, the brightness of the lighting. Therefore, the inner radius of the iris 51 may change due to, for example, the brightness of the lighting. Furthermore, the size of the iris region 51a in the image of the subject's eye (e.g., distance d in FIG. 2E) may change due to the distance between the camera capturing the subject and the subject. In other words, the size of the iris region included in the iris image acquired by the iris acquisition unit 110 (e.g., the iris region 51a included in the iris image 510) is not constant.

[0052] In order to correct the deviation in size of the iris region 51a that occurs when photographing the subject, the iris image normalization unit 213 may normalize the image 520 by, for example, setting the value of the bottom side of the image 520 to "0" and the value of the top side of the image 520 to "1" and changing the value of the vertical axis of the image 520 so that the value of the vertical axis of the image 520 is between 0 and 1.

[0053] In addition, in order to correct a deviation in size of the iris region 51a that occurs when photographing a subject, the inner radius of the iris image 510 (i.e., the inner radius of the iris 51) and the outer radius of the iris image 510 (i.e., the outer radius of the iris 51) may each be changed to a predetermined value, and then the iris region 51a may be expanded in polar coordinates to obtain a normalized expanded image of the iris region.

[0054] In practice, for example, polar coordinate expansion and normalization may be performed simultaneously by including a variable for normalization in the coordinate transformation matrix for expanding the iris region into polar coordinates. An image in which the iris region has been expanded into polar coordinates (e.g., image 520) may be referred to as a rubber sheet image. For this reason, expanding the iris region into polar coordinates may also be referred to as rubber sheet expansion.

[0055] The partial iris image normalization unit 214 performs polar coordinate expansion on the non-occluded portion of the iris region included in the iris image. The partial iris image normalization unit 214 may normalize a rectangular image generated by performing polar coordinate expansion on the non-occluded portion. Note that the image generated by the partial iris image normalization unit 214 performing polar coordinate expansion on the non-occluded portion will hereinafter be referred to as a "normalized partial iris image" as appropriate.

[0056] The iris feature extraction unit 215 extracts iris features based on the iris-normalized image. The iris feature extraction unit 215 may, for example, divide the iris-normalized image into a plurality of blocks. The iris feature extraction unit 215 may, for example, extract iris features based on the luminance value of each of the plurality of blocks. The process of extracting iris features from an iris-normalized image is simpler than, for example, the process of extracting iris features from an iris image. This makes it possible to reduce the processing load on the iris feature extraction unit 215.

[0057] The partial iris feature extraction unit 216 extracts partial iris features based on the partial iris normalized image. The partial iris feature extraction unit 216 may, for example, divide the partial iris normalized image into a plurality of blocks. The partial iris feature extraction unit 216 may, for example, extract partial iris features based on the luminance value of each of the plurality of blocks. The process of extracting partial iris features from the partial iris normalized image can reduce the input image size and the processing load compared to, for example, the process of extracting partial iris features from an iris image. Therefore, the processing load on the partial iris feature extraction unit 216 can be reduced.

[0058] The weight score estimation unit 217 estimates a weight score based on the eye image acquired by the image acquisition unit 211. For example, the weight score estimation unit 217 may estimate a first weight score related to the iris feature amount and a second weight score related to the partial iris feature amount based on the eye image. For example, the weight score estimation unit 217 may estimate, as the weight score, a vector or a matrix having the first weight score and the second weight score as components based on the eye image. Note that the "weight score" corresponds to the "weight" in the first embodiment.

[0059] The weight score estimation unit 217 may estimate the weight score using, for example, a computational model that inputs an eye image and outputs a weight score. The computational model may have a neural network structure. An example of a neural network is a convolutional neural network (CNN). The computational model itself may be referred to as the weight score estimation unit 217. By using such a computational model, the weight score can be estimated relatively easily.

[0060] The authentication unit 218 authenticates the target based on the iris features extracted by the iris feature extraction unit 215, the partial iris features extracted by the partial iris feature extraction unit 216, and the weighted score estimated by the weighted score estimation unit 217.

[0061] Here, the authentication unit 218 has a fusion feature calculation unit 2181. The fusion feature calculation unit 2181 calculates a fusion feature based on the iris feature, the partial iris feature, and the weight score. The fusion feature calculation unit 2181 may, for example, multiply the iris feature by the weight score related to the iris feature (e.g., the above-mentioned first weight score) and multiply the partial iris feature by the weight score related to the partial iris feature (e.g., the above-mentioned second weight score). The fusion feature calculation unit 2181 may calculate the fusion feature by combining, in the channel direction, the iris feature multiplied by the weight score and the partial iris feature multiplied by the weight score.

[0062] Specifically, the fusion feature calculation unit 2181 may multiply the iris feature of a point in the iris-normalized image by a weighting score related to the iris feature. The fusion feature calculation unit 2181 may multiply the partial iris feature of a point in the partial iris-normalized image that corresponds to the above-mentioned point in the iris-normalized image by a weighting score related to the partial iris feature. The fusion feature calculation unit 2181 may calculate the fusion feature of the above-mentioned point by combining the iris feature of the above-mentioned point in the iris-normalized image multiplied by the weighting score and the partial iris feature of the above-mentioned corresponding point in the partial iris-normalized image multiplied by the weighting score. With this configuration, it is possible to calculate the fusion feature relatively easily.

[0063] The authentication unit 218 compares the fusion feature calculated by the fusion feature calculation unit 2181 with features pre-registered in the registered information database 221. The authentication unit 218 calculates a matching score indicating the degree of match between the fusion feature and the registered feature. The authentication unit 218 authenticates the target by comparing the matching score with a predetermined threshold. If the matching score is equal to or greater than the predetermined threshold, the authentication unit 218 may determine that the target is a person registered in the registered information database 221. On the other hand, if the matching score is less than the predetermined threshold, the authentication unit 218 may determine that the target is not registered in the registered information database 221.

[0064] The operation of the information processing system 2 will be described with reference to the flowchart in Fig. 5. In Fig. 5, the image acquisition unit 211 acquires an eye image (step S201). Next, the eye position estimation unit 212 estimates the position of the target's eye from the eye image acquired in the processing of step S201 (step S202). As a result of the processing of step S202, an iris image may be generated.

[0065] The iris image normalization unit 213 generates an iris normalized image from the iris region included in the iris image (step S203). The iris feature extraction unit 215 extracts iris features from the iris normalized image generated in the processing of step S203 (step S204).

[0066] In parallel with the processes of steps S203 and S204, the partial iris image normalization unit 214 generates a partial iris normalized image from an unoccluded portion of the iris region included in the iris image (step S205). The partial iris feature extraction unit 216 extracts partial iris features from the partial iris normalized image generated in the process of step S205 (step S206).

[0067] In parallel with the processes of steps S203 and S204 and steps S205 and S206, the weighted score estimation unit 217 estimates a weighted score based on the eye image acquired in the process of step S201 (step S207).

[0068] Thereafter, the authentication unit 218 performs authentication of the target based on the iris feature extracted in the processing of step S204, the partial iris feature extracted in the processing of step S206, and the weighted score estimated in the processing of step S207 (step S208).

[0069] (Technical effect) According to the information processing system 2 of the second embodiment, similar to the information processing system 1 of the first embodiment described above, it is possible to properly authenticate a subject regardless of whether the subject is wearing colored contact lenses or not.

[0070] Third Embodiment A third embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to Figures 4 and 6. Below, the third embodiment of the information processing system, the information processing method, and the recording medium will be described using an information processing system 2. Note that, with regard to the third embodiment, descriptions that overlap with the descriptions of the first and second embodiments will be omitted as appropriate.

[0071] In the third embodiment, the weighted score estimation unit 217 estimates the weighted score using a computational model that outputs a weighted score when an eye image is input. An example of a learning method for the computational model will be described below with reference to FIG. 6.

[0072] 6 are elements for training the computational model. The authentication loss calculation unit 201, the gradient calculation unit 202, and the parameter update unit 203 may be included in the computation device 21. The authentication loss calculation unit 201, the gradient calculation unit 202, and the parameter update unit 203 may be included in a device different from the information processing system 2.

[0073] The training data used to learn the computational model includes a personal ID, iris features, partial iris features, and an input image. The input image is an image of the eyes of a person identified by the personal ID. The iris features are iris features extracted by the iris feature extraction unit 215 based on the input image. The partial iris features are partial iris features extracted by the partial iris feature extraction unit 216 based on the input image. The training data may also be referred to as learning data.

[0074] The weight score estimation unit 217 may input an input image included in one training data set to a computational model and acquire a weight score output from the computational model. The weight score estimation unit 217 transmits the acquired weight score to the fusion feature calculation unit 2181. The fusion feature calculation unit 2181 calculates a fusion feature based on the iris feature and partial iris feature included in the one training data set and the weight score. The fusion feature calculation unit 2181 transmits the calculated fusion feature to the authentication loss calculation unit 201.

[0075] The weight score estimation unit 217 and the fusion feature calculation unit 218 may perform the above-described operations on a predetermined number (e.g., 128) of training data sets that include a plurality of training data sets, and calculate the average loss. Note that the predetermined number of training data sets may also be referred to as mini-batch training data.

[0076] The authentication loss calculation unit 201 may, for example, perform multi-class classification of multiple (i.e., a predetermined number of) fusion features calculated by the fusion feature calculation unit 2181. The authentication loss calculation unit 201 may perform multi-class classification using, for example, a softmax function. The authentication loss calculation unit 201 may use the softmax function to calculate multiple probabilities that one fusion feature corresponds to each of multiple people. The authentication loss calculation unit 201 may calculate a loss from the calculated multiple probabilities and a personal ID. The authentication loss calculation unit 201 may calculate an overall loss by calculating an average or weighted sum of the multiple losses. Note that the loss calculation method is not limited to the above-mentioned method, and other methods may also be used.

[0077] The gradient calculation unit 202 may calculate the gradient based on a loss function. For example, the gradient calculation unit 202 may calculate the gradient based on the overall loss when the weight related to the loss function is increased by a small value and the overall loss when the weight related to the loss function is decreased by a small value.

[0078] The parameter modification unit 203 may update parameters related to the computation model used by the weight score estimation unit 217 based on the gradient calculated by the gradient calculation unit 202. The parameter modification unit 203 may update the parameters related to the computation model based on the gradient, for example, so that the overall loss is reduced.

[0079] In learning the computational model, the above-described operations may be repeated, for example, until the amount of change in the total loss becomes equal to or less than a predetermined value, or until a predetermined number of iterations is reached.

[0080] (Technical Effect) According to the information processing system 2 according to the third embodiment, it is possible to learn a computational model, and therefore it is possible to construct a weighted score estimation unit 217 that can estimate an appropriate weighted score based on an eye image.

[0081] Fourth Embodiment A fourth embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to Fig. 4. The fourth embodiment of the information processing system, the information processing method, and the recording medium will be described below using an information processing system 2. Note that, with regard to the fourth embodiment, descriptions that overlap with the descriptions of the first to third embodiments will be omitted as appropriate.

[0082] In the fourth embodiment, the weight score estimation unit 217 may estimate the likelihood of wearing color contact lenses from the eye image. The weight score estimation unit 217 may estimate a weight score based on the estimated likelihood. The weight score estimation unit 217 may increase the weight score related to the iris feature amount relative to the weight score related to the partial iris feature amount as the estimated likelihood decreases. The weight score estimation unit 217 may increase the weight score related to the partial iris feature amount relative to the weight related to the iris feature amount as the estimated likelihood increases.

[0083] The weighted score estimation unit 217 may estimate the likelihood using, for example, a computational model that, when an eye image is input, outputs the likelihood of wearing color contact lenses. The computational model may have a neural network structure. For example, the computational model may be trained to output the likelihood of wearing color contact lenses based on the likelihood that the subject is wearing color contact lenses and the likelihood that the subject is not wearing color contact lenses.

[0084] Additionally, the computational model may be trained to determine from an eye image whether a subject is wearing colored contact lenses.

[0085] (Technical Effect) According to the information processing system 2 according to the fourth embodiment, it is possible to estimate the weighted score relatively easily.

[0086] Fifth Embodiment A fifth embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to FIG. Hereinafter, the fifth embodiment of the information processing system, the information processing method, and the recording medium will be described using an information processing system 2a. Note that the information processing system 2a may have the same configuration as the information processing system 2, except that the weighted score estimation unit 217 includes an image quality estimation unit 2171 and a weighted score calculation unit 2172. Therefore, for the fifth embodiment, descriptions that overlap with the descriptions of the first to fourth embodiments will be omitted as appropriate.

[0087] 7 , the weighted score estimation unit 217 includes an image quality estimation unit 2171 and a weighted score calculation unit 2172. The image quality estimation unit 2171 estimates a quality score indicating the image quality of the eye image based on the eye image. The weighted score calculation unit 2172 calculates a weighted score based on the estimated quality score.

[0088] The image quality estimation unit 2171 may estimate the quality score by, for example, estimating at least one of the likelihood of wearing colored contact lenses, the occlusion area caused by the colored contact lenses, and the color of the colored contact lenses.

[0089] The image quality estimation unit 217 may estimate the likelihood using, for example, a computational model that, when an eye image is input, outputs the likelihood of wearing color contact lenses (see the fourth embodiment described above).

[0090] For example, the image quality estimation unit 217 may estimate, from the eye image, an occlusion region of the iris region that is occluded by a colored region of the color contact lens (e.g., colored region 60 in FIG. 2C). The image quality estimation unit 217 may calculate, for example, an occlusion likelihood map having the same size as the eye image. Here, the occlusion likelihood map has a likelihood value for each of multiple pixels that make up the occlusion likelihood map, indicating the likelihood that the iris region is occluded by a colored region of the color contact lens. The image quality estimation unit 217 may estimate, as an occlusion region, a pixel whose likelihood value exceeds a predetermined likelihood value. The image quality estimation unit 217 may estimate the area of ​​the occlusion region based on the number of pixels estimated to be an occlusion region.

[0091] For example, when an eye image is input, the image quality estimation unit 217 may estimate the occluded region using a computational model for estimating the occluded region. The computational model may have a neural network structure. For example, the computational model may be trained to perform binary classification of whether each of a plurality of pixels constituting the eye image corresponds to an occluded region or not.

[0092] The image quality estimation unit 217 may estimate the color of the color contact lens (e.g., the color of the colored region of the color contact lens) based on the luminance values ​​of one or more pixels in the eye image corresponding to the occluded region estimated as described above (i.e., the occluded region of the iris region that is occluded by the colored region of the color contact lens). The image quality estimation unit 217 may estimate the color of the color contact lens based on at least one of the average luminance value, the maximum luminance value, and the minimum luminance value. The image quality estimation unit 217 may calculate contrast from the maximum luminance value and the minimum luminance value.

[0093] For example, when an eye image is input, the image quality estimation unit 2171 may estimate the color of the color contact lens using a calculation model for estimating the color of the color contact lens. The calculation model may have a neural network structure.

[0094] An example of a method for calculating the weight score by the weight score calculation unit 2172 will be described below. The weight score calculation unit 2172 calculates a first weight score related to the iris feature amount and a second weight score related to the partial iris feature amount. Here, the weight score calculation unit 2172 calculates the first weight score and the second weight score based on the first quality score related to the likelihood, the second quality score related to the occluded region, and the third quality score related to the color.

[0095] The weighted score calculation unit 2172 calculates, for example, "first weighted score = first quality score × w １ + Second quality score × w ２ + Third quality score x w ３ ". Similarly, the weighted score calculation unit 2172 may calculate the first weighted score using the formula "second weighted score=first quality score×w ４ + Second quality score × w ５ + Third quality score x w ６ The first weighted score may be calculated using the formula: １ , w ２ , w ３ , w ４ , w ５ and w ６ is the weight associated with each quality score.

[0096] The weighted score calculation unit 2172 may calculate the weighted score using, for example, a computation model that receives a quality score as input and outputs a weighted score. The computation model may be trained using a learning method similar to the learning method for the computation model in the third embodiment described above. However, when training the computation model, the quality score is input to the computation model instead of the input image.

[0097] (Technical Effect) According to the information processing system 2a according to the fifth embodiment, it is possible to estimate the weighted score relatively easily.

[0098] Sixth Embodiment An information processing system, an information processing method, and a recording medium according to a fifth embodiment will be described with reference to Fig. 7. The fifth embodiment relating to the information processing system, the information processing method, and the recording medium will be described below using an information processing system 2a. Note that, with regard to the sixth embodiment, descriptions that overlap with the descriptions of the first to fifth embodiments will be omitted as appropriate.

[0099] The image quality estimation unit 2171 may estimate the quality score by estimating at least one of, for example, the likelihood of wearing colored contact lenses, the occluded area caused by the colored contact lenses, the area ratio between the area of ​​the portion of the iris region that is occluded by the colored area of ​​the colored contact lenses and the area of ​​the portion of the iris region that is not occluded by the colored area, the frequency of the colored area of ​​the colored contact lenses, the color of the iris region, and the color of the colored contact lenses.

[0100] For example, the image quality estimation unit 2171 may estimate, from the eye image, an occluded region of the iris region that is occluded by the colored region of the color contact lens (see the fifth embodiment described above). The image quality estimation unit 2171 may estimate the area of ​​the occluded region (i.e., the portion of the iris region that is occluded by the colored region of the color contact lens) based on the estimated occluded region. The image quality estimation unit 2171 may estimate the area of ​​the portion of the iris region that is not occluded by the colored region of the color contact lens based on the estimated occluded region. The image quality estimation unit 2171 may estimate the area ratio between the area of ​​the portion of the iris region that is occluded by the colored region of the color contact lens and the area of ​​the portion of the iris region that is not occluded by the colored region of the color contact lens based on the area of ​​the occluded region and the area of ​​the portion of the iris region that is not occluded by the colored region of the color contact lens.

[0101] For example, when an eye image is input, the image quality estimation unit 2171 may estimate the area ratio using a computational model that estimates the area ratio between the area of ​​the portion of the iris region that is obscured by the colored region of the color contact lens and the area of ​​the portion of the iris region that is not obscured by the colored region. The computational model may have a neural network structure.

[0102] The image quality image estimation unit 217 may, for example, analyze the frequency obtained by applying a Fourier transform to the estimated occlusion area, and estimate the frequency with the highest intensity and its intensity as the frequency of the colored area of ​​the color contact lens.

[0103] For example, when an eye image is input, the image quality estimation unit 2171 may estimate the frequency of the colored region of the color contact lens using a computational model for estimating the frequency of the colored region of the color contact lens. The computational model may have a neural network structure.

[0104] The image quality estimation unit 2171 may estimate a portion of the iris region that is not occluded by the colored region of the color contact lens based on the estimated occlusion region. The image quality estimation unit 2171 may estimate the color of the iris based on the luminance values ​​of one or more pixels in the eye image that correspond to the estimated portion. The image quality estimation unit 217 may estimate the color of the iris based on, for example, at least one of the average luminance value, the maximum luminance value, and the minimum luminance value. The image quality estimation unit 217 may calculate contrast from the maximum luminance value and the minimum luminance value.

[0105] For example, when an eye image is input, the image quality estimation unit 2171 may estimate the iris color using a computational model for estimating the iris color. The computational model may have a neural network structure.

[0106] (Technical Effect) According to the information processing system 2a according to the sixth embodiment, it is possible to estimate the weighted score relatively easily.

[0107] Seventh Embodiment A seventh embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to FIG. Hereinafter, the seventh embodiment of the information processing system, the information processing method, and the recording medium will be described using an information processing system 2b. Note that the information processing system 2b may have the same configuration as the information processing system 2, except that the authentication unit 218 has a fusion score calculation unit 2182. Therefore, for the seventh embodiment, descriptions that overlap with the descriptions of the first to sixth embodiments will be omitted as appropriate.

[0108] 8, the authentication unit 218 has a fusion score calculation unit 2182. Note that the authentication unit 218 does not necessarily have to have the fusion feature calculation unit 2181 (see FIG. 4). In the seventh embodiment, it is assumed that features corresponding to iris features and features corresponding to partial iris features are registered in advance in the registration information database 221.

[0109] Fusion score calculation unit 2182 calculates a first matching score indicating the degree of match between the iris features extracted by iris feature extraction unit 215 and features corresponding to iris features registered in advance in registration information database 221. Fusion score calculation unit 2181 calculates a second matching score indicating the degree of match between the partial iris features extracted by partial iris feature extraction unit 216 and features corresponding to partial iris features registered in advance in registration information database 221. Using the weighted score estimated by weighted score estimating unit 217, fusion score calculation unit 2182 calculates a weighted average of the first matching score and the second matching score as the fusion score.

[0110] The authentication unit 218 authenticates the subject by comparing the fusion score with a predetermined threshold. If the fusion score is equal to or greater than the predetermined threshold, the authentication unit 218 may determine that the subject is a person registered in the enrollment information database 221. On the other hand, if the fusion score is less than the predetermined threshold, the authentication unit 218 may determine that the subject is not registered in the enrollment information database 221.

[0111] (Technical Effect) According to the information processing system 2b of the seventh embodiment, it is possible to calculate a fusion score, and therefore, according to the information processing system 2b, it is possible to appropriately authenticate a subject regardless of whether the subject is wearing color contact lenses.

[0112] Eighth Embodiment An eighth embodiment of an information processing system, an information processing method, and a recording medium will be described. Hereinafter, the eighth embodiment of the information processing system, the information processing method, and the recording medium will be described using an information processing system 2. Note that, with regard to the eighth embodiment, descriptions that overlap with the descriptions of the first to seventh embodiments will be omitted as appropriate.

[0113] As described above, the partial iris image normalization unit 214 performs polar coordinate expansion of the unoccluded portion of the iris region included in the iris image. The partial iris image normalization unit 214 may perform polar coordinate expansion of the unoccluded portion, for example, with the center of the eye (e.g., the center of the pupil) as the reference. When the subject is wearing colored contact lenses, the center of the subject's eye and the center of the colored contact lenses often coincide or nearly coincide.

[0114] However, for some reason, there may be a relatively large deviation between the center of the eye of a subject wearing a color contact lens and the center of the color contact lens. In this case, when partial iris image normalization unit 214 performs polar coordinate expansion of the unoccluded portion with reference to the center of the subject's eye, there is a possibility that a relatively large amount of the colored region of the color contact lens will be included in the partial iris normalized image.

[0115] Therefore, the partial iris image normalization unit 214 may detect the outer edge of the iris or the colored region based on, for example, a change in brightness between the iris or the colored region of the color contact lens and the sclera (the so-called white of the eye) in the eye image. The partial iris image normalization unit 214 may perform polar coordinate expansion of the unoccluded portion using the center of a circle that approximates the detected outer edge as a reference.

[0116] If the subject is not wearing colored contact lenses, the detected outer edge corresponds to the outer edge of the iris. If the subject is wearing colored contact lenses, the detected outer edge corresponds to the outer edge of the colored region. If the subject is wearing colored contact lenses, the center of a circle that approximates the detected outer edge can be considered the center of the colored contact lenses.

[0117] (Technical Effect) According to the information processing system 2 according to the eighth embodiment, when the subject is wearing color contact lenses, the partial iris image normalization unit 214 performs polar coordinate expansion of the unobstructed portion with the center of the color contact lenses as the reference. Therefore, even if there is a relatively large deviation between the center of the eye of the subject wearing the color contact lenses and the center of the color contact lenses, it is possible to appropriately perform polar coordinate expansion of the unobstructed portion.

[0118] Ninth Embodiment A ninth embodiment of an information processing system, an information processing method, and a recording medium will be described. The ninth embodiment of the information processing system, the information processing method, and the recording medium will be described below using an information processing system 2. Note that, with regard to the ninth embodiment, descriptions that overlap with the descriptions of the first to eighth embodiments will be omitted as appropriate.

[0119] The partial iris image normalization unit 214 may identify non-occluded portions (i.e., portions of the iris region that are not occluded by an occluding object) from the iris image. The partial iris image normalization unit 214 may, for example, calculate an occlusion likelihood map of the same size as the iris image. Here, the occlusion likelihood map has a likelihood value for each of multiple pixels that make up the occlusion likelihood map, indicating the likelihood that the iris region is occluded by the colored region of the color contact lens. The partial iris image normalization unit 214 may identify pixels whose likelihood value is less than a predetermined likelihood value as non-occluded portions.

[0120] For example, when an iris image is input, the partial iris image normalization unit 214 may identify the unoccluded portion using a computational model that identifies the unoccluded portion. The computational model may have a neural network structure. For example, the computational model may be trained to perform a binary classification of whether each of the multiple pixels that make up the iris corresponds to the unoccluded portion or not.

[0121] In the information processing system 2 according to the ninth embodiment, eye images or iris images of registered individuals may be registered in the registration information database 221. When the unoccluded portion is identified as described above, for example, the range of the unoccluded portion in the iris region varies depending on the subject. Therefore, the authentication unit 218 may extract feature amounts of a portion of the registered eye image or iris image that corresponds to the unoccluded portion identified as described above. Note that the method for extracting the feature amounts of the corresponding portion may be the same as the method used by the partial iris feature extraction unit 216 to extract partial iris feature amounts from a normalized partial iris image. The authentication unit 218 may use the extracted feature amounts to authenticate the subject.

[0122] (Technical Effect) In the information processing system 2 according to the ninth embodiment, the smaller the area of ​​the colored region of the color contact lens, the larger the size of the unoccluded portion. In other words, the larger the area of ​​the colored region of the color contact lens, the smaller the size of the unoccluded portion. For subjects not wearing color contact lenses, the size of the iris region and the size of the unoccluded portion may be the same. This configuration makes it possible to increase the amount of features related to the iris pattern extracted from the unoccluded portion while preventing at least a portion of the colored region of the color contact lens from being included in the unoccluded portion.

[0123] Tenth Embodiment A tenth embodiment of an information processing system, an information processing method, and a recording medium will be described below. The tenth embodiment of the information processing system, the information processing method, and the recording medium will be described below using an information processing system 2. Note that, with regard to the tenth embodiment, descriptions that overlap with the descriptions of the first to ninth embodiments will be omitted as appropriate.

[0124] The partial iris image normalization unit 214 may consider a predetermined range of the iris region from the inner edge of the iris region (i.e., the boundary between the iris and the pupil) as the unoccluded portion. Note that this predetermined range may be set, for example, based on the range that the colored region of a color contact lens can take. The partial iris image normalization unit 214 may define the unoccluded portion as a region of the iris region where the probability that the iris pattern will be occluded by the occluded region is equal to or less than a predetermined probability, for example, based on the size of the occluded region of a commonly sold color contact lens (e.g., the colored region of a color contact lens).

[0125] (Technical Effect) In the tenth embodiment, the size of the unoccluded portion is a fixed value. Therefore, it is possible to extract in advance feature amounts related to the unoccluded portion from eye images or iris images of persons registered in the registration database 221 (i.e., feature amounts corresponding to partial iris feature amounts) and register them in the registration database 221. Therefore, for example, when authenticating a target, the processing load on the information processing system 2 can be reduced compared to when feature amounts related to the unoccluded portion are extracted from eye images or iris images of persons registered in the registration database 221.

[0126] Eleventh Embodiment An eleventh embodiment of an information processing system, an information processing method, and a recording medium will be described with reference to Fig. 9. The eleventh embodiment of the information processing system, the information processing method, and the recording medium will be described below using an information processing system 2. Note that, with regard to the eleventh embodiment, descriptions that overlap with the descriptions of the first to tenth embodiments will be omitted as appropriate.

[0127] The partial iris image normalization unit 214 may consider a predetermined range from the inner edge of the iris region (i.e., the boundary between the iris and the pupil) as the non-occluded portion. Particularly in this embodiment, the partial iris image normalization unit 214 may estimate multiple non-occluded portions.

[0128] 9 , the partial iris image normalization unit 214 may determine a ring-shaped portion of the iris image 510, in which the distance between the inner edge of the iris region and the outer edge of the non-occluded portion is distance d1, as a first non-occluded portion. The partial iris image normalization unit 214 may determine a ring-shaped portion of the iris region, in which the distance between the inner edge of the iris region and the outer edge of the non-occluded portion is distance d2, as a second non-occluded portion. The partial iris image normalization unit 214 may determine a ring-shaped portion of the iris region, in which the distance between the inner edge of the iris region and the outer edge of the non-occluded portion is distance d3, as a third non-occluded portion. Note that the distances d1, d2, and d3 are predetermined values ​​(i.e., fixed values). The number of non-occluded portions is not limited to three, and may be two, four, or more.

[0129] The partial iris image normalization unit 214 may generate a first partial iris normalized image by polar coordinate expansion of the first unoccluded portion. The partial iris image normalization unit 214 may generate a second partial iris normalized image by polar coordinate expansion of the second unoccluded portion. The partial iris image normalization unit 214 may generate a third partial iris normalized image by polar coordinate expansion of the third unoccluded portion.

[0130] The partial iris feature extraction section 216 may extract a first partial iris feature from the first partial iris normalized image. The partial iris feature extraction section 216 may extract a second partial iris feature from the second partial iris normalized image. The partial iris feature extraction section 216 may extract a third partial iris feature from the third partial iris normalized image.

[0131] The weighted score estimation section 217 may estimate, as the weighted scores, a weighted score related to the first partial iris feature amount, a weighted score related to the second partial iris feature amount, and a weighted score related to the third partial iris feature amount.

[0132] The fusion feature calculation unit 2181 of the authentication unit 218 may calculate the fusion feature based on the iris feature, the first partial iris feature, the second partial iris feature, the third partial iris feature, and the weighted scores (for example, the weighted score related to the iris feature, the weighted score related to the first partial iris feature, the weighted score related to the second partial iris feature, and the weighted score related to the third partial iris feature).

[0133] (Technical Effect) The drawing density of at least one of the dots and lines constituting the pattern drawn in the colored region of the color contact lens is arbitrary. Therefore, iris-related features may be extracted even from portions of the iris region that are covered by the colored region. Therefore, by estimating multiple unoccluded portions, it is possible to increase the number of iris pattern-related features extracted from the unoccluded portions. Furthermore, because the sizes of the multiple unoccluded portions are fixed values, it is possible to pre-register multiple partial iris features corresponding to multiple partial iris features (e.g., first partial iris feature, second partial iris feature, and third partial iris feature) in the registration database 221. This reduces the processing load on the information processing system 2.

[0134] <Twelfth embodiment> A twelfth embodiment of an information processing system, an information processing method, and a recording medium will be described. The twelfth embodiment of the information processing system, the information processing method, and the recording medium will be described below using an information processing system 2. Note that, for the twelfth embodiment, descriptions that overlap with the descriptions of the first to eleventh embodiments will be omitted as appropriate.

[0135] The authentication unit 218 may perform multimodal authentication based on information related to the iris of the subject (e.g., iris feature amounts and partial iris feature amounts) and other biometric information different from the iris of the subject. Examples of other biometric information include the area around the eye (periocular), face, fingerprint, veins, voice, and ear acoustics.

[0136] The weighted score estimation unit 217 may estimate, for example, a weighted score related to the iris feature amount, a weighted score related to the partial iris feature amount, and a weighted score related to other biometric information.

[0137] (Technical Features) According to the information processing system 2 according to the twelfth embodiment, it is possible to improve the reliability of authentication of an object.

[0138] <Supplementary Notes> The following supplementary notes are further disclosed regarding the above-described embodiment.

[0139] (Supplementary Note 1) An information processing system comprising: an extraction means for extracting, based on an eye image including an iris region of a target, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an obstruction; and an authentication means for authenticating the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0140] (Supplementary Note 2) The information processing system according to Supplementary Note 1, wherein the occluding object is a colored contact lens, and the information processing system further comprises a weight estimation means for estimating at least one element from the eye image: the likelihood of wearing the colored contact lens, an occluded region that is a part of the iris region occluded by a colored region of the colored contact lens, and the color of the colored contact lens, and estimating the weight based on the estimated at least one element.

[0141] (Supplementary Note 3) The information processing system according to Supplementary Note 1 or 2, wherein the occluding object is a colored contact lens, and the information processing system further comprises a weight estimation means for estimating from the eye image at least one element of the likelihood of wearing the colored contact lens, an occluded region that is a part of the iris region occluded by the colored region of the colored contact lens, an area ratio between the area of ​​the occluded region and the area of ​​the other part of the iris region that is not occluded by the colored region, a frequency of the colored region, a color of the iris, and a color of the colored contact lens, and estimating the weight based on the estimated at least one element.

[0142] (Supplementary Note 4) The information processing system according to any one of Supplementary Notes 1 to 3, further comprising: a weight estimation unit that estimates the weights using the eye image and a computational model having a neural network structure.

[0143] (Supplementary Note 5) The information processing system according to any one of Supplementary Notes 1 to 4, wherein the occluding object is a color contact lens, and the information processing system further comprises a weight estimation means for estimating a likelihood of wearing the color contact lens from the eye image and estimating the weight based on the estimated likelihood of wearing the color contact lens.

[0144] (Supplementary Note 6) The information processing system according to any one of Supplementary Notes 1 to 5, wherein the extraction means extracts the iris feature amount based on a first exfoliated image in which the iris region is exfoliated in polar coordinates, and extracts the partial iris feature amount based on a second exfoliated image in which the non-occluded portion is exfoliated in polar coordinates.

[0145] (Supplementary Note 7) The information processing system according to Supplementary Note 6, further comprising a generating means for generating the second unfolded image, wherein the generating means sets a portion of the iris region corresponding to a predetermined range from inside the iris region as the unoccluded portion.

[0146] (Supplementary Note 8) The information processing system according to Supplementary Note 7, wherein the predetermined range includes a plurality of ranges, and the generating means sets a plurality of portions corresponding to the plurality of ranges respectively as the unoccluded portion.

[0147] (Supplementary Note 9) The information processing system according to Supplementary Note 6, further comprising: a generating unit configured to generate the second unfolded image, wherein the generating unit identifies the unoccluded portion based on the eye image.

[0148] (Supplementary Note 10) The information processing system according to Supplementary Note 9, wherein the occluding object is a colored contact lens, and the generating means identifies the unoccluded portion based on a colored region of the colored contact lens.

[0149] (Supplementary Note 11) The information processing system according to any one of Supplementary Notes 6 to 10, wherein the occluding object is a color contact lens, and further comprising a generating means for generating the second expanded image, wherein the generating means performs polar coordinate expansion of the unoccluded portion with the center of the color contact lens as a reference.

[0150] (Supplementary Note 12) The information processing system according to any one of Supplementary Notes 1 to 11, wherein the weights include a first weight related to the iris feature amount and a second weight related to the partial iris feature amount, and the authentication means calculates a fusion feature amount from a product of the iris feature amount and the first weight and a product of the partial iris feature amount and the second weight.

[0151] (Supplementary Note 13) The information processing system according to Supplementary Note 12, wherein the authentication means calculates the fused feature by combining, in a channel direction, the iris feature to which the first weight has been applied and the partial iris feature to which the second weight has been applied.

[0152] (Supplementary Note 14) The information processing system described in any one of Supplementary Notes 1 to 11, wherein the authentication means calculates a weighted average using the weights of an iris authentication score calculated from the iris feature amount and registered iris feature amount, and a partial iris authentication score calculated from the partial iris feature amount and registered partial iris feature amount, thereby calculating a fusion score.

[0153] (Supplementary Note 15) The information processing system according to any one of Supplementary Notes 1 to 14, wherein the authentication means performs multimodal authentication based on the iris feature amounts and the partial iris feature amounts, which are information relating to characteristics of the iris of the target, and other biometric information different from the iris of the target, and includes weight estimation means for estimating weights relating to the iris feature amounts, weights relating to the partial iris feature amounts, and weights relating to the other biometric information.

[0154] (Supplementary Note 16) An information processing method comprising: extracting, based on an eye image including an iris region of a target, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an obstructing object; and authenticating the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0155] (Supplementary Note 17) The information processing method according to Supplementary Note 16, wherein the occluding object is a colored contact lens, and at least one element of the following is estimated from the eye image: likelihood of wearing the colored contact lens, an occluded region that is a part of the iris region occluded by a colored region of the colored contact lens, and a color of the colored contact lens, and the weight is estimated based on the estimated at least one element.

[0156] (Supplementary Note 18) The information processing method according to Supplementary Note 16 or 17, wherein the occluding object is a colored contact lens, and at least one element of the following is estimated from the eye image: likelihood of wearing the colored contact lens; an occluded region that is a part of the iris region occluded by the colored region of the colored contact lens; an area ratio between the area of ​​the occluded region and the area of ​​the other part of the iris region that is not occluded by the colored region; a frequency of the colored region; the color of the iris; and the color of the colored contact lens; and the weight is estimated based on the estimated at least one element.

[0157] (Supplementary Note 19) The information processing method according to any one of Supplementary Notes 16 to 18, wherein the weights are estimated using the eye image and a computational model having a neural network structure.

[0158] (Supplementary Note 20) The information processing method according to any one of Supplementary Notes 16 to 19, wherein the occluding object is a color contact lens, and a likelihood of wearing the color contact lens is estimated from the eye image, and the weight is estimated based on the estimated likelihood of wearing the color contact lens.

[0159] (Supplementary Note 21) The information processing method according to any one of Supplementary Notes 16 to 20, wherein the iris feature amount is extracted based on a first expanded image in which the iris region is expanded in polar coordinates, and the partial iris feature amount is extracted based on a second expanded image in which the non-occluded portion is expanded in polar coordinates.

[0160] (Supplementary Note 22) The information processing method according to Supplementary Note 21, wherein a portion of the iris region corresponding to a predetermined range from inside the iris region is set as the non-occluded portion.

[0161] (Supplementary Note 23) The information processing method according to Supplementary Note 22, wherein the predetermined range includes a plurality of ranges, and a plurality of portions corresponding to the plurality of ranges respectively are set as the unobstructed portion.

[0162] (Supplementary Note 24) The information processing method according to Supplementary Note 21, further comprising identifying the unoccluded portion based on the eye image.

[0163] (Supplementary Note 25) The information processing method according to Supplementary Note 24, wherein the occluding object is a colored contact lens, and the non-occluded portion is identified based on a colored region of the colored contact lens.

[0164] (Supplementary Note 26) The information processing method according to any one of Supplementary Notes 21 to 25, wherein the obstructing object is a colored contact lens, and the unobstructed portion is expanded in polar coordinates with the center of the colored contact lens as a reference.

[0165] (Supplementary Note 27) The information processing method according to any one of Supplementary Notes 16 to 26, wherein the weights include a first weight related to the iris feature amount and a second weight related to the partial iris feature amount, and a fusion feature amount is calculated from a product of the iris feature amount and the first weight and a product of the partial iris feature amount and the second weight.

[0166] (Supplementary Note 28) The information processing method according to Supplementary Note 27, wherein the fused feature is calculated by combining, in a channel direction, the iris feature to which the first weight has been applied and the partial iris feature to which the second weight has been applied.

[0167] (Supplementary Note 29) The information processing method described in any one of Supplementary Notes 16 to 26, wherein a fusion score is calculated by calculating a weighted average using the weights of an iris authentication score calculated from the iris feature amount and registered iris feature amount, and a partial iris authentication score calculated from the partial iris feature amount and registered partial iris feature amount.

[0168] (Supplementary Note 30) The information processing method according to any one of Supplementary Notes 16 to 29, wherein multimodal authentication is performed based on the iris feature amounts and the partial iris feature amounts, which are information relating to characteristics of the iris of the target, and other biometric information different from the iris of the target, and weights relating to the iris feature amounts, weights relating to the partial iris feature amounts, and weights relating to the other biometric information are estimated.

[0169] (Supplementary Note 31) A recording medium having recorded thereon a computer program for causing a computer to execute an information processing method of: extracting, based on an eye image including an iris region of a subject, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an obstruction; and authenticating the subject based on weights estimated based on the eye image, the iris features, and the partial iris features.

[0170] This disclosure is not limited to the above-described embodiments, but may be modified as appropriate within the scope of the claims and the gist or idea of ​​the invention as can be read from the entire specification, and information processing systems, information processing methods, and recording media that involve such modifications are also included in the technical scope of this disclosure.

[0171] 1, 2 Information processing system 11 Extraction unit 12, 218 Authentication unit 211 Image acquisition unit 212 Eye position estimation unit 213 Iris image normalization unit 214 Partial iris image normalization unit 215 Iris feature amount extraction unit 216 Partial iris feature amount extraction unit 217 Weight score estimation unit

Claims

1. An information processing system comprising: an extraction means for extracting, based on an eye image including an iris region of a target, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an occluding object; and an authentication means for authenticating the target based on weights estimated based on the eye image, the iris features, and the partial iris features.

2. The information processing system of claim 1, wherein the occluding object is a colored contact lens, and the information processing system is equipped with a weight estimation means for estimating at least one element from the eye image: the likelihood of wearing the colored contact lens, an occluded area which is a part of the iris area that is occluded by the colored area of ​​the colored contact lens, and the color of the colored contact lens, and estimating the weight based on the estimated at least one element.

3. The information processing system of claim 1, wherein the occluding object is a colored contact lens, and the information processing system is equipped with a weight estimation means for estimating from the eye image at least one element of the likelihood of wearing the colored contact lens, an occluded area which is a part of the iris area occluded by the colored area of ​​the colored contact lens, an area ratio between the area of ​​the occluded area and the area of ​​other parts of the iris area not occluded by the colored area, the frequency of the colored area, the color of the iris, and the color of the colored contact lens, and estimating the weight based on the estimated at least one element.

4. The information processing system according to claim 1, further comprising a weight estimation means for estimating the weights using the eye image and a computational model having a neural network structure.

5. The information processing system of claim 1, wherein the occluding object is a colored contact lens, and the information processing system further comprises a weight estimation means for estimating the likelihood of wearing the colored contact lens from the eye image and estimating the weight based on the estimated likelihood of wearing the lens.

6. The information processing system according to claim 1, wherein the extraction means extracts the iris feature amount based on a first expanded image in which the iris area is expanded in polar coordinates, and extracts the partial iris feature amount based on a second expanded image in which the non-occluded portion is expanded in polar coordinates.

7. An information processing system as described in claim 6, further comprising a generating means for generating the second expanded image, wherein the generating means determines that a portion of the iris region corresponding to a predetermined range from inside the iris region is the unobstructed portion.

8. The information processing system according to claim 7, wherein the predetermined range includes a plurality of ranges, and the generating means sets a plurality of portions corresponding to the plurality of ranges respectively as the unobstructed portion.

9. An information processing system according to claim 6, further comprising a generating means for generating the second unfolded image, the generating means identifying the non-occluded portion based on the eye image.

10. The information processing system according to claim 9, wherein the occluding object is a colored contact lens, and the generating means identifies the non-occluded portion based on a colored region of the colored contact lens.

11. The information processing system according to claim 6, wherein the obstructing object is a colored contact lens, and further comprising a generating means for generating the second expanded image, the generating means expanding the unobstructed portion in polar coordinates with the center of the colored contact lens as a reference.

12. The information processing system of claim 1, wherein the weights include a first weight associated with the iris feature and a second weight associated with the partial iris feature, and the authentication means calculates a fusion feature from a product of the iris feature and the first weight and a product of the partial iris feature and the second weight.

13. The information processing system according to claim 12, wherein the authentication means calculates the fused feature by combining, in the channel direction, the iris feature to which the first weight has been applied and the partial iris feature to which the second weight has been applied.

14. The information processing system of claim 1, wherein the authentication means calculates a fusion score by calculating a weighted average using the weights of an iris authentication score calculated from the iris features and registered iris features and a partial iris authentication score calculated from the partial iris features and registered partial iris features.

15. The information processing system of claim 1, wherein the authentication means performs multimodal authentication based on the iris features and partial iris features, which are information relating to the characteristics of the iris of the subject, and other biometric information different from the iris of the subject, and is provided with a weight estimation means for estimating weights relating to the iris features, weights relating to the partial iris features, and weights relating to the other biometric information.

16. An information processing method comprising: extracting, based on an eye image including an iris region of a subject, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an occluding object; and authenticating the subject based on weights estimated based on the eye image, the iris features, and the partial iris features.

17. A recording medium having recorded thereon a computer program for causing a computer to execute an information processing method which extracts, based on an eye image including an iris region of a subject, iris features indicating characteristics of the iris region and partial iris features indicating characteristics of an unobstructed portion of the iris region that is not obscured by an occluding object, and authenticates the subject based on weights estimated based on the eye image, the iris features, and the partial iris features.

Images