JP2004185386A - Image collation device - Google Patents

Abstract

An object of the present invention is to provide an image matching device capable of performing high-precision matching without using a complicated shape acquiring unit.
An image matching unit (1) for comparing and matching a first two-dimensional image with a second two-dimensional image, wherein the image converting unit (3) converts the first two-dimensional image into a three-dimensional image and outputs the three-dimensional image. And a direction detection unit 7 for detecting and outputting the shooting angle of the subject in the second two-dimensional image, and a shooting angle output from the direction detection unit 7 based on the three-dimensional image output from the image conversion unit 3. An image creating unit 4 for creating and outputting a third two-dimensional image viewed from the direction in which the third two-dimensional image is viewed, and an image matching unit 8 for matching the third two-dimensional image with the second two-dimensional recorded image.
[Selection diagram] Fig. 1

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image matching device that compares and matches an input two-dimensional image with a previously recorded two-dimensional image.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, an image matching device that compares and matches an input two-dimensional image with a previously recorded two-dimensional image has been proposed. In particular, a face authentication method, which is one of authentication methods using biometrics, is used. Various image collating devices have been proposed.
[0003]
In such an image matching apparatus, a face image of a person to be authenticated (hereinafter, referred to as a recorded face image) is recorded in a database in advance, and an input image of the face of the subject to be referred is input. (Hereinafter, referred to as an input face image) and a recorded face image recorded in a database are compared and collated to perform the authentication of the person to be authenticated.
[0004]
In such an image matching device, differences in various shooting conditions between an input face image and a recorded face image, for example, when capturing an image of a subject's face using an optical system, the front direction of the subject's face and However, there is a problem in that the accuracy of authentication is reduced due to a difference in an angle (hereinafter, referred to as an imaging angle) between the optical axis direction of the imaging device and the like.
[0005]
In order to solve such a problem, for example, a three-dimensional face image model (a three-dimensional face shape model of a wire frame and two-dimensional face image texture data) is recorded as a recorded face image to be recorded in a database, Creates a two-dimensional reference face image from the three-dimensional face image model that matches the shooting angle of the input face image of the input subject to be referred, and compares the reference face image with the input face image Techniques have been proposed to increase the authentication accuracy by performing collation (for example, see Patent Literature 1, Patent Literature 2, and Patent Literature 3).
[0006]
[Patent Document 1]
Japanese Patent Application Laid-Open No. 2000-306095 [Patent Document 2]
JP 2001-283224 A [Patent Document 3]
JP 2001-273496 A
[Problems to be solved by the invention]
However, in the above-described image collating apparatus, it is necessary to create a three-dimensional shape model, so that a complicated photographing device for acquiring the three-dimensional shape of the subject is required. Was needed.
[0008]
The present invention has been made to solve such a problem, and an object of the present invention is to provide an image matching device capable of performing high-accuracy matching without using a complicated shape acquiring unit.
[0009]
[Means for Solving the Problems]
An image matching device of the present invention is an image matching device that compares and matches a first two-dimensional image with a second two-dimensional image, and converts the first two-dimensional image into a three-dimensional image and outputs the three-dimensional image. A conversion unit, a direction detection unit that detects a shooting direction of the subject in the second two-dimensional image, an image creation unit that creates a third two-dimensional image viewed from the shooting direction based on the three-dimensional image, An image collating unit for collating the second two-dimensional image with the third two-dimensional image is provided.
[0010]
With such a configuration, two-dimensional images are used as the input image and the recorded image, respectively, as the images to be collated, so there is no need to use a shape acquiring unit for acquiring a complicated three-dimensional shape. Further, the first two-dimensional image is once converted into a three-dimensional image, and then a third two-dimensional image viewed from the shooting direction of the subject in the second two-dimensional image is created based on the three-dimensional image, Since the matching is performed between the second two-dimensional image and the third two-dimensional image, high-precision matching is performed as compared with the case where two-dimensional images captured from different imaging directions are compared and compared. Becomes possible.
[0011]
Also, an image acquisition unit for acquiring a first two-dimensional image, an image recording unit for recording a two-dimensional recorded image, a second two-dimensional image cut out from the two-dimensional recorded image, and a direction detection unit and an image comparison unit With the configuration including the image cutout unit to be output, the image recording unit only needs to record a two-dimensional recording image, so there is no need to record a large amount of information such as a three-dimensional shape model, An apparatus configuration with a small recording capacity can be realized.
[0012]
Further, since the first two-dimensional image and the second two-dimensional image are each an image representing a human face, an image collating apparatus capable of comparing and collating human face images to perform personal authentication. Can be provided.
[0013]
Further, according to the configuration in which the image acquisition unit is an image reading device that reads a first two-dimensional image from a photograph or a printed matter, it is possible to acquire the first two-dimensional image from a photograph or a printed matter in which a face of a subject is photographed. Therefore, it is not necessary to bring the subject to a specific place such as a place where the photographing device is arranged in order to photograph the first two-dimensional image as the input face image.
[0014]
Further, according to the configuration in which the two-dimensional recording image recorded in the image recording unit is a moving image captured by the imaging device, a recording face image is cut out from a moving image captured by a monitoring camera or the like, and the recorded face image is recorded. It can be used as a certain second two-dimensional image.
[0015]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0016]
(First Embodiment)
First, an image matching device according to the present invention will be described as a first embodiment of the present invention. FIG. 1 is a block diagram showing the configuration of the image matching device according to the first embodiment of the present invention.
[0017]
In FIG. 1, an image matching device 1 compares an input face image (hereinafter, referred to as an input face image) with a previously recorded face image (hereinafter, referred to as a recorded face image), and determines the result of the comparison. It is a device that outputs.
[0018]
The image matching device 1 includes an image acquisition unit 2 for acquiring a two-dimensional input face image, an image conversion unit 3 for converting an input face image output from the image acquisition unit 2 into a three-dimensional image, a two-dimensional image recorded in advance. An image recording unit 5 for recording a recording image, an image extracting unit 6 for arbitrarily cutting out one recording face image from the recording image recorded in the image recording unit 5, and one recording face image output from the image extracting unit 6 From the three-dimensional image output from the image conversion unit 3 which is connected to the direction detection unit 7, the direction detection unit 7, and the image conversion unit 3 for detecting the shooting angle of the subject when the recorded face image is shot. An image creation unit 4 for creating a two-dimensional face image (hereinafter referred to as a reference face image) that matches the shooting angle of the recorded face image detected by the detection unit 7, and an image cutout unit 6 and image creation Connected to unit 4 and output from image cutout unit 6 Composed of the recording face image and the image matching portion 8 which compares collating the reference face image output from the image forming unit 4. A signal indicating the result of comparison and comparison between the recorded face image and the reference face image is output from the output terminal 9 from the image matching unit 8.
[0019]
Next, the function of each component will be described in detail.
[0020]
The image acquisition unit 2 can use a video camera device or a digital still camera device that captures a two-dimensional image of a person to be authenticated as a subject, and can use an image reading device such as an image scanner device that reads a photograph of the person to be authenticated. It can also be used. When such an image reading device is used, an input face image of the subject can be acquired even when the subject is not present. FIG. 2 shows an example of the input face image acquired by the image acquisition unit 2.
[0021]
The image conversion unit 3 converts the two-dimensional input face image output from the image acquisition unit 2 into a three-dimensional image. This processing is described in, for example, Hideto Kameshima and two others, “Estimation of Three-Dimensional Shape from Two-Dimensional Face Image by BPLP”, Proceedings of the 8th Image Sensing Symposium, Image Sensing Technology Workshop, July 17, 2002 Days, p. 521-526 can be used. FIGS. 3A and 3B show examples of the wire-frame model 32 and the texture-mapped three-dimensional image 33 converted by the image conversion unit 3, respectively.
[0022]
On the other hand, a two-dimensional image is recorded in the image recording unit 5 (hereinafter, this image is referred to as a recorded image). The two-dimensional recorded image may be an image including one or more human face images. For example, as an example of such a recorded image, a still image or a moving image photographed and recorded by a monitoring camera installed in an ATM or the like can be used.
[0023]
The image clipping unit 6 clips recorded face images 34 to 37 as shown in FIG. 4 from the recorded image recorded in the image recording unit 5. In addition, as a recording medium used for the image recording unit 5, a recording medium arbitrarily selected from known recording media such as a magnetic disk and a semiconductor memory can be used.
[0024]
Next, when a plurality of recorded face images are cut out, the image cutout unit 6 selects one of the recorded face images 34 to 37. The method of this selection is not particularly limited, and the selected image is sent to the image collating unit 8 and the direction detecting unit 7. Here, as an example, it is assumed that the recorded face image 37 is selected and sent to the direction detecting unit 7.
[0025]
The direction detection unit 7 analyzes the recorded face image 37 output from the image clipping unit 6 and calculates the direction (shooting angle) of the face of the subject in the recorded face image 37. The image collating apparatus of the present invention does not particularly limit this method, and various known methods can be used. For example, a method of estimating the direction of the face by calculating the skin color occupation area in the face region (for example, refer to Patent Document 1), and the feature points of the face image facing forward and the feature points of the recorded face image are used. A method of comparing and estimating the direction of the face (for example, see Patent Document 3) can be used. The information on the shooting angle of the recorded face image 37 output from the direction detecting unit 7 is sent to the image creating unit 4.
[0026]
The image creation unit 4 creates a reference face image 38 from the three-dimensional image 33 output from the image conversion unit 3 so as to match the shooting angle sent from the direction detection unit 7. The method of creating the reference face image 38 from the three-dimensional image 33 can be performed by using a known method (for example, see Patent Literature 1 or Patent Literature 2). FIG. 5 shows an example of a reference face image that has undergone such processing. The reference face image 38 is sent to the image matching unit 8.
[0027]
The image matching unit 8 compares and matches the reference face image 38 output from the image creating unit 4 with the recorded face image 37 output from the image extracting unit 6. As the comparison and matching method, a known method can be used. For example, a method of performing a Karhunen-Loeve expansion on a face image, projecting the image on an eigenspace, and performing comparison and matching using the distance between the projection points can be used. (For example, see Patent Document 1). The result of the comparison and collation of the reference face image 38 and the recorded face image 37 by the image collating unit 8 is sent to the outside through the output terminal 9.
[0028]
It should be noted that the image conversion unit 3, the image creation unit 4, the image cutout unit 6, the direction detection unit 7, and the image comparison unit 8 may each be configured to realize the above-described functions using software. Needless to say.
[0029]
Next, the operation of the image matching device 1 of the present invention will be described.
[0030]
FIG. 6 is a flowchart showing operation steps of the image matching device 1 according to the first embodiment of the present invention.
[0031]
In FIG. 6, first, the input face image 31 is obtained from the image obtaining unit 2 (step S102).
[0032]
Next, the input face image 31 is converted into a three-dimensional image 33 in the image conversion unit 3 (step S103). The information of the three-dimensional image 33, that is, the information of the wire frame and the texture information are temporarily stored in the buffer unit (not shown) of the image matching device 1 from the viewpoint of suppressing duplicate operations. desirable.
[0033]
Next, the image cutout unit 6 selects one recorded face image (here, the recorded face image 37) from the plurality of recorded face images 34 to 37 recorded in the image recording unit 5 (step). S104).
[0034]
Further, the direction detecting unit 7 detects the photographing angle of the recorded face image 37 selected by the image extracting unit 6 (Step S105).
[0035]
The image creating unit 4 creates a reference face image 38 from the information of the three-dimensional image 33 stored in the buffer unit based on the information on the shooting angle output from the direction detecting unit 7 (Step S106).
[0036]
Next, the image collating unit 8 performs collation between the reference face image 38 and the recorded face image 37 (Step S107). Note that, in the example shown in the present embodiment, the comparison result between the reference face image 38 and the recorded face image 37 is “NG (personal authentication impossible)”.
[0037]
If the collation result in step S107 is "OK (identification is possible)", the collation result is output from the output terminal 9 and the process ends (step S110). If the collation result in step S107 is “NG”, the image cutout unit 6 selects another recording face image from the image recording unit 5 different from the recording face image 37 again. If the image has been selected, the collation result “NG” is output from the output terminal 9 and the process ends (step S109). In step S109, if there is a recorded face image that has not been selected in the image recording unit 5, the process returns to step S104, and the image cutout unit 6 selects another recorded face image.
[0038]
As described above, according to the image matching device 1 of the present invention, since the recorded image recorded in the image recording unit 5 is a two-dimensional image, there is no need to record a plurality of three-dimensional images. Requires less capacity.
[0039]
In addition, since the input face image to be obtained may be a two-dimensional image, a special mechanism or the like for obtaining a three-dimensional image is not required when obtaining the input face image.
[0040]
Furthermore, according to the image matching device of the present invention, since the input face image is converted and matched so as to match the shooting angle of the recorded face image, a high authentication rate can be realized.
[0041]
(Second embodiment)
Next, another example of the image matching apparatus will be described as a second embodiment of the present invention.
[0042]
FIG. 7 is a block diagram illustrating a configuration of the image matching device 21 according to the second embodiment of the present invention. In the image matching apparatus 21 shown in FIG. 7, the same components as those of the image matching apparatus 1 shown in the first embodiment are denoted by the same reference numerals, and description thereof will be omitted.
[0043]
The image matching apparatus 21 according to the second embodiment of the present invention is different from the image matching apparatus 1 described in the first embodiment in that a recording unit 25 for recording a recorded image is different from the image matching apparatus 21. Is provided in the image recording device 26 of FIG.
[0044]
The image recording device 26 is a device that records images, for example, an image recording device using a hard disk drive (HDD), a video recorder device, and the like. Further, the connection method between the image recording device 26 and the input terminal 24 of the image matching device 21 may be via a network such as a wired, wireless, or Internet. Note that the essential component of the image recording device 26 is the recording unit 25, and the presence or absence of other components does not matter. Therefore, the description is omitted in FIG.
[0045]
By using such an image matching device 21, for example, an image recorded in the recording unit 25 is recorded on an image recording device 26 such as an ATM or a monitoring device equipped with a photographing device provided remotely via a network, and the recorded face image is recorded. And image information such as a photograph of a license or a snap photograph acquired by the image acquiring unit 2 of the image collating apparatus 21 or recorded in advance, and can be subjected to collation processing as an input face image.
[0046]
In the image matching device 21 according to the embodiment of the present invention, the example in which the image cutout unit 6 is included in the configuration has been described, but the present invention is not limited to this configuration. For example, it goes without saying that the image clipping unit 6 may be included in the configuration of the image storage device 26. In this case, the recorded face image cut out by the image recording device 26 is sent to the image collation device 21 direction detection unit 7 and the image collation unit 8.
[0047]
In the image matching device 21 according to the embodiment of the present invention, in order to increase the matching rate, the input face image acquired by the image acquiring unit 2 has a photographing angle of 0 ° as much as possible, ie, It is desirable to use a frontal face image in which the optical axis coincides with the frontal direction of the subject's face. It is needless to say that the input face image to be input preferably has the highest possible resolution, but an image having a resolution of about 1/4 VGA size (320 × 240 pixels) can be used. In this case, the width of the face only needs to be about 150 pixels.
[0048]
In addition, if the image collating device according to the embodiment of the present invention is used, when a photograph of a license (length × width = 4 cm × 3 cm) is used as an input face image, the recording face image has a shooting angle with respect to the front direction. If the angle is within about 40 °, the authentication can be performed well.
[0049]
In the embodiment of the present invention, a human face has been described as a subject of an image to be collated by the image collating device, but the image collating device of the present invention is not limited to this. For example, it is needless to say that images obtained by photographing other three-dimensional subjects can be compared and collated.
[0050]
【The invention's effect】
As described above, if the image matching device of the present invention is used, matching is performed using the input image and the recorded image, each of which is a two-dimensional image, so that it is necessary to use a shape obtaining unit that obtains a complicated three-dimensional shape. There is no. Further, the input two-dimensional image is once converted into a three-dimensional image, and thereafter, a reference image that matches the shooting angle of the subject in the recorded image is created, and comparison between the recorded image and the reference image is performed. Therefore, highly accurate collation can be performed.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of an image matching device according to a first embodiment of the present invention. FIG. 2 is a diagram illustrating an example of an input face image according to the first embodiment of the present invention. FIG. 3A shows an example of a wire frame model converted by the image conversion unit according to the first embodiment of the present invention. FIG. 3B shows a cubic model converted by the image conversion unit according to the first embodiment of the present invention. FIG. 4 shows an example of an original image. FIG. 4 shows an example of a recorded face image according to the first embodiment of the present invention. FIG. 5 shows an example of a reference face image according to the first embodiment of the present invention. FIG. 6 is a flowchart showing operation steps of the image matching apparatus according to the first embodiment of the present invention. FIG. 7 is a block diagram showing a configuration of the image matching apparatus according to the second embodiment of the present invention. Description】
1, 21 Image collation device 2 Image acquisition unit 3 Image conversion unit 4 Image creation unit 5 Image recording unit 6 Image extraction unit 7 Direction detection unit 8 Image collation unit 9 Output terminal 24 Input terminal 25 Recording unit 26 Image recording device 31 Input Face image (first two-dimensional image)
32 Wireframe model 33 Three-dimensional image 34, 35, 36, 37 Recorded face image (second two-dimensional image)
38 Reference face image (third two-dimensional image)

Claims (5)

An image matching device for comparing and matching a first two-dimensional image and a second two-dimensional image,
An image conversion unit that converts the first two-dimensional image into a three-dimensional image and outputs the three-dimensional image;
A direction detection unit that detects a shooting direction of a subject in the second two-dimensional image;
An image creation unit that creates a third two-dimensional image viewed from the shooting direction based on the three-dimensional image;
An image matching device, comprising: an image matching unit that matches the second two-dimensional image with the third two-dimensional image. An image acquisition unit that acquires the first two-dimensional image;
An image recording unit that records a two-dimensional recording image;
The image collating apparatus according to claim 1, further comprising: an image clipping unit that clips the second two-dimensional image from the two-dimensional recorded image and outputs the second two-dimensional image to the direction detection unit and the image matching unit. . The image matching device according to claim 2, wherein the first two-dimensional image and the second two-dimensional image are images representing a human face, respectively. The image matching device according to claim 3, wherein the image acquisition unit is an image reading device that reads the first two-dimensional image from a photograph or a printed matter. The image matching device according to claim 3, wherein the two-dimensional recorded image recorded in the image recording unit is a moving image captured by a capturing device.

Images