JP2010122994A - Face authentication system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a face authentication system hardly affected by a quantization error. <P>SOLUTION: The system includes: an imaging means 1; a face detecting means 2; a cutting-out means 4 for cutting out a plurality of normalized cut-out images A, B, by changing respectively cut-out conditions for cutting out the normalized cut-out images A, B in conformity with a preset size and angle of an image used for authentication, from an image of a face area; a cut-out condition changing means 5 for changing the cut-out condition in the cutting-out means 4; a density gradient image generating means 6 for generating density gradient images from the plurality of normalized cut-out images A, B, respectively; an authentication image generating means 8 for generating one sheet of authentication image from the plurality of density gradient images; and an authentication means 10 for authenticating the face, using the authentication image. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a face authentication system that authenticates a face of a person to be authenticated in an image captured using a density gradient image.

Conventionally, an image including the face of a person to be authenticated is captured, and the authentication target person's face image extracted from the captured image is compared with a pre-registered reference face image indicating the registrant's face for authentication. A face authentication system for determining whether or not a subject is a registrant is provided. For example, there is one disclosed in Patent Document 1. The conventional example described in Patent Document 1 is a face image collation device that realizes personal identification, and performs an image input unit that performs an input process of a target person's face image, and a shading process such as luminance correction of the input face image. A density conversion means, a position normalization processing means for normalizing the position and size of the input face image, and a feature extraction for performing a process of extracting a feature pattern necessary for matching from the normalized face image Means, collation processing means for collating the extracted feature pattern with a standard pattern of face images of each person registered in advance, and determination processing means for judging personal identification from the result of collation.
JP-A-5-20442

  However, in the above conventional example, a quantization error occurs because the position of the input image is shifted or the size is changed when the input image is normalized. And if a quantization error occurs in the boundary part of each organ of the face such as eyes and nose, the characteristics of each individual appearing in the boundary part will be lost, and it will be difficult to distinguish the person to be authenticated from others There was a problem. A technique for suppressing the generation of quantization error as much as possible by a method such as sub-pixel matching used in a stereo camera or the like is known. However, in the face authentication system, the same facial expression and the same face angle are used due to time series changes. Since the image is hardly captured, the effect cannot be exhibited.

  The present invention has been made in view of the above points, and an object of the present invention is to provide a face authentication system that can be hardly affected by a quantization error.

  In order to achieve the above object, the first aspect of the present invention provides an image pickup means for picking up an image including the face of the person to be authenticated, and a face for detecting a face area corresponding to the face from the picked-up image obtained by the image pickup means. A detection unit, a cutout unit that cuts out a plurality of normalized cutout images by changing the cutout conditions for cutting out the normalized cutout images in accordance with the size and angle of the image used for authentication from the face area image, and the cutout conditions in the cutout unit A plurality of densities obtained by replacing the pixel value of each pixel with a numerical value corresponding to the direction of the density gradient at the position of the target pixel for each of a plurality of normalized cut-out images obtained by the cut-out means A density gradient image generating unit that generates a gradient image, and one authentication image in which a quantization error is reduced from all the density gradient images obtained by the density gradient image generating unit And authentication image generating means for forming, characterized by comprising an authentication means for performing face authentication using the authentication image obtained by the authentication image generating means. Note that the authentication image is used for authentication of the authentication target person, but if the reference face image of the authentication target person is not registered, the authentication image is stored as the reference face image of the authentication target person.

  According to a second aspect of the present invention, in the first aspect of the invention, the cutout condition changing means changes the cutout condition by displacing the face region in at least one of the vertical and horizontal directions.

  According to a third aspect of the present invention, in the second aspect of the present invention, the cutout condition changing means changes the amount of displacement according to the size of the face area.

  According to a fourth aspect of the present invention, in the first aspect of the invention, the cutout condition changing means changes the cutout condition by changing a magnification of the size of the face area.

  According to a fifth aspect of the present invention, in the first aspect of the invention, the cutout condition changing means changes the cutout condition by changing an angle at which the face area is rotated.

  According to a sixth aspect of the present invention, in the invention according to any one of the second to fifth aspects, the authentication image generating means indicates the direction of the density gradient between the target pixel and the surrounding pixels for all the pixels of each density gradient image. The difference between the angles is calculated, and for all the pixels, the average value of the angles indicating the direction of the density gradient of all the density gradient images in the target pixel is used as the pixel value of the authentication image.

  The invention according to claim 7 is the invention according to any one of claims 2 to 5, wherein the authentication image generation means indicates the direction of the density gradient between the target pixel and the surrounding pixels for all the pixels of each density gradient image. An angle is calculated, and for all pixels, an intermediate value of an angle indicating the direction of the density gradient of all density gradient images in the target pixel is used as the pixel value of the authentication image.

  The invention according to claim 8 is the invention according to any one of claims 2 to 5, wherein the authentication image generation means indicates the direction of the density gradient between the target pixel and the surrounding pixels for all the pixels of each density gradient image. In addition to calculating the angle, the angle distribution indicating the direction of the density gradient in the target pixel is calculated, and for all the pixels, the angle indicating the direction of the density gradient having the largest distribution in the target pixel is set as the pixel value of the authentication image. It is characterized by doing.

  According to a ninth aspect of the present invention, in the invention according to any one of the sixth to eighth aspects, the authentication image generating means calculates an angle difference indicating a direction of the density gradient between the authentication image and each density gradient image. The calculation is performed for all the pixels, and the difference result in each pixel is compared with a preset first threshold to count the number of images exceeding the first threshold, and the number of images counted is preset. When the second threshold value is compared with the second threshold value, mask processing is performed.

  The invention of claim 10 is characterized in that, in the invention of claim 9, the second threshold value is a preset fixed value.

  According to an eleventh aspect of the present invention, in the ninth aspect of the invention, the second threshold value is an image in which the difference between the number of the normalized images and the non-normalized images as a whole and the angle difference indicating the direction of the density gradient exceeds the second threshold value. The ratio to the number of sheets is set to be a predetermined value.

  According to the present invention, a plurality of normalized cutout images are generated by changing cutout conditions from each captured image including the face of a person to be authenticated, and an authentication image generated from the density gradient images of the plurality of normalized cutout images. Is used to perform face authentication, so that it is affected by quantization error compared to the case of generating a density gradient image from only one normalized cutout image cut out with a preset cutout condition and performing face authentication. Can be difficult.

  Hereinafter, embodiments of a face authentication system according to the present invention will be described with reference to the drawings. In the present embodiment, as shown in FIG. 1A, an imaging unit 1 that captures an image including a face of a person to be detected, and a face detection unit that detects a face area from a captured image obtained by the imaging unit 1. 2, a first storage unit 3 that stores a captured image obtained by the imaging unit 1, and an image of the face area detected by the face detection unit 2 by reading the captured image from the first storage unit 3 for authentication. A cutout unit 4 for cutting out a plurality of normalized cutout images by changing the cutout conditions for cutting out the normalized cutout image in accordance with the size and angle of the image to be used; and a cutout condition changing unit 5 for changing the cutout conditions in the cutout unit 4 A density gradient image generating unit 6 that generates a density gradient image from each of the plurality of normalized cutout images obtained by the cutout unit 4 and a plurality of density gradient images obtained by the density gradient image generation unit 6 Second storage means 7 for storing, authentication image generating means 8 for generating one authentication image from a plurality of density gradient images read from the second storage means 7, and authentication image generating means 8 The third storage means 9 for storing the authentication image obtained in the above and the authentication means 10 for performing face authentication using the authentication image.

  The imaging unit 1 includes an imaging device such as a CCD (Charge Coupled Device) camera, for example. The imaging unit 1 captures a grayscale image including the face of the person to be authenticated and inputs the image to the face detection unit 2. Input to the storage means 3 to store.

  As shown in FIG. 2 (a), the face detection means 2 is, for example, a rectangular area including a portion corresponding to the face from the grayscale image (a) in FIG. (A) in the figure is detected, and the detection result is input to the clipping means 4 and the clipping condition changing means 5. Since a method for detecting a face area from a captured gray image is well known, detailed description thereof will be omitted here.

  The cutout unit 4 cuts out a face image from the grayscale image read from the first storage unit 3 based on the detection result of the face detection unit 2 as shown in FIG. Then, the face image is appropriately rotated and the size thereof is appropriately enlarged or reduced to obtain a normalized cutout image A having a preset size and matching angle. For example, when the predetermined size set in advance is 20 × 20 pixels and the size of the clipped face image is 30 × 30 pixels, the size of the face image is reduced by 2/3 times in the vertical and horizontal directions. Further, the cutout unit 4 cuts out a face image from the grayscale image according to the cutout condition given from the cutout condition changing unit 5 described later, and obtains a plurality of normalized cutout images B separately from the normalized cutout image A.

  The cutout condition changing unit 5 changes the cutout conditions when the cutout unit 4 cuts out by changing the position and size of the region with respect to the grayscale image, and inputs the cutout conditions to the cutout unit 4. For example, the cutout condition is changed by displacing the position of the region in at least one of the horizontal direction (x direction) and the vertical direction (y direction). In the present embodiment, as shown in FIG. 3A, the region is displaced by 0.5 pixel or 1 pixel in the x direction, and the region is displaced by 0.5 pixel or 1 pixel in the y direction. In combination with the case, the cutout conditions are changed in a total of 24 ways, and 24 normalized cutout images B are cut out by the cutout unit 4 under the cutout conditions. In the present embodiment, the region is displaced by 0.5 pixel or 1 pixel, but the displacement amount is not limited to this, and may be changed according to the size of the region.

  The density gradient image generating means 6 first calculates the density gradient dx in the x direction and the density gradient dy in the y direction at each pixel position for a total of 25 pixels of the normalized cutout images A and B, respectively. To do. For the calculation of the density gradients dx, dy, for example, a known Sobel filter is used. Then, for the target pixel that is the target of the density gradient calculation and each of the eight pixels in the vicinity of the upper left, upper, upper right, left, right, lower left, lower, lower right of the target pixel (pixels in the vicinity of eight) By applying the 3 × 3 x-direction Sobel filter and the y-direction Sobel filter, the density gradient dx that is the differential value in the x-axis direction and the density gradient dy that is the differential value in the y-axis direction at the target pixel are obtained. Can be sought.

  Here, using the angle expression in which the numerical value increases clockwise when the right direction is 0 °, the direction of the density gradient at the position of the target pixel (that is, the direction in which the pixel value decreases in the normalized cut-out images A and B). (Hereinafter referred to as “density gradient direction value”) θ is represented by θ = arctan (dy / dx) (when dx> 0, dy ≧ 0). Except for the case where the concentration gradients dx and dy are both 0, the value of the concentration gradient direction value θ can be determined within a range of 0 ° ≦ θ <360 ° for all the concentration gradients dx and dy. Specifically, when dx = 0, θ = 90 ° if dy> 0, θ = 270 ° if dy <0, and θ = arctan (dy / 0 if dx> 0 and dy <0. dx) + 360 °, and if dx <0, θ = arctan (dy / dx) + 180 °. Then, the density gradient image generation unit 6 determines the density gradient direction value θ within the above range for the target pixel from which the density gradients dx and dy are obtained.

  Next, the density gradient image generation means 6 determines the pixel value in the density gradient image for the target pixel for which the density gradient direction value θ has been determined. In the present embodiment, the possible pixel values in the density gradient image are eight values that correspond one-to-one in the eight directions of up, down, left, right, upper right, lower right, upper left, and lower left. By performing the above processing for all the pixels of the normalized cutout images A and B, a plurality of density gradient images based on the normalized cutout images A and B are generated (see FIG. 2C). The density gradient image generated by the density gradient image generation unit 6 is stored in the second storage unit 7.

  The authentication image generation unit 8 generates one authentication image based on the plurality of density gradient images read from the second storage unit 7. Specifically, the average value of the density gradient direction values θ in the target pixels of all density gradient images is calculated, and the pixel value of the target pixel in the authentication image is determined based on the average value. And the image for authentication is produced | generated by performing the said calculation about all the pixels (refer FIG.3 (b)). Note that the authentication image is input to the authentication means 10 at the subsequent stage and used for authentication of the authentication target person. However, when the reference face image of the authentication target person is not registered, a third memory composed of a nonvolatile memory is used. The authentication image is stored in the storage unit 9 as the reference face image of the person to be authenticated.

  The authentication unit 10 compares the authentication image from the authentication image generation unit 8 with the reference face image of the authentication target person read by the third storage unit 9 to thereby register the authentication target person registered in advance. It is determined whether or not it is the person himself. Since the determination method is well known, detailed description thereof will be omitted here.

  Hereinafter, a process of generating an authentication image in the present embodiment will be described with reference to FIG. First, a grayscale image including the face of the person to be authenticated is captured and input to the face detection means 2, and the captured image is input to and stored in the first storage means 3. The face detection unit 2 detects the face area of the person to be authenticated from the captured grayscale image, and inputs the detection result to the cutout unit 4 and the cutout condition change unit 5. In the cutout unit 4, the region is cut out from the grayscale image to generate a normalized cutout image A, which is input to the density gradient image generation unit 6. Then, the density gradient image generation unit 6 generates a density gradient image of the normalized cutout image A and stores it in the second storage unit 7.

  Next, the cutout condition changing unit 5 changes the cutout condition for displacing the position of the region with respect to the grayscale image and inputs the change to the cutout unit 4. The cutout unit 4 cuts out the grayscale image according to the cutout condition and normalizes the cutout image. B is generated. The generated normalized cutout image B is input to the density gradient image generation unit 6, and the density gradient image generation unit 6 generates a density gradient image of the normalized cutout image B and stores it in the second storage unit 7. Then, the density gradient image of the normalized cutout image B is generated by changing the cutout conditions at any time until the density gradient images of the normalized cutout images A and B of the preset number (25 in this embodiment) are generated. And stored in the second storage means 7. When the process of generating the density gradient image is completed, the authentication image generation means 8 calculates the average value of the density gradient direction values θ in the target pixels of all density gradient images, and the object in the authentication image is based on the average value. The pixel value of the pixel is determined. And the image for authentication is produced | generated by performing the said calculation about all the pixels.

  As described above, in the present embodiment, 25 normalized cutout images A and B are generated by changing the cutout conditions from the grayscale image including the face of the person to be authenticated, and a plurality of normalized cutout images A and B are generated. Face authentication is performed using the authentication image generated from the density gradient image, and the face authentication is performed by generating the density gradient image from only one normalized cutout image A cut out under preset cutout conditions. Compared to the above, it is difficult to be influenced by the quantization error.

  In the present embodiment, the authentication image generation means 8 calculates the average value of the density gradient direction values θ in the target pixels of all density gradient images to determine the pixel value in the target pixel of the authentication image. However, the intermediate value of the density gradient direction value θ may be calculated to determine the pixel value in the target pixel. Alternatively, the distribution of the density gradient direction value θ in the target pixel may be calculated, and the pixel value may be determined from the density gradient direction value θ having the largest distribution in the target pixel.

  In the present embodiment, the clipping condition is changed by displacing the face area detected by the face detecting means 2 in at least one of the vertical, horizontal and horizontal directions in the clipping condition changing means 5. However, the present invention is not limited to this. For example, the extraction condition may be changed by changing the magnification of the size of the region (see FIGS. 4A to 4C). Further, the cutting condition may be changed by rotating the region (see FIGS. 5A to 5C).

  By the way, in the present embodiment, as described above, the quantization error is reduced by generating the authentication image from the density gradient images of the plurality of normalized cutout images A and B, but in order to further reduce the quantization error. Alternatively, the authentication image may be masked. That is, the authentication image generation means 8 calculates the difference in the density gradient direction value θ between the authentication image and each density gradient image for all the pixels, and sets the difference result and the first preset in each pixel. The number of images exceeding the first threshold is counted by comparing with the threshold value, and the mask processing is performed when the number of images counted is compared with a preset second threshold value and exceeds the second threshold value. To apply. Thus, a portion where the concentration gradient direction value θ is likely to be random can be extracted and masked, and the quantization error can be further reduced.

  In the mask process, the second threshold value is set to a fixed value. For this reason, by appropriately setting a fixed value, all pixels in which the difference in the density gradient direction value θ exceeds the first threshold value is higher than a certain value are masked as pixels that are prone to quantization errors. be able to. However, the setting of the second threshold value in the authentication image generating means 8 is not necessarily limited to this. For example, the difference between the total number of normalized cut-out images A and B and the density gradient direction value θ becomes the second threshold value. The threshold value may be set so that the ratio with the number of images exceeding the predetermined ratio. In this case, since the number of pixels used for the mask process is substantially constant regardless of the number of pixels of the captured image, it is possible to prevent variations in the number of pixels used during authentication.

It is a figure which shows embodiment of the face authentication system which concerns on this invention, (a) is the schematic of a system, (b) is a flowchart. (A)-(c) is a figure which shows the face image in each process process same as the above. (A) is a figure which shows each normalization cut-out image when a face area is displaced to the up-down and left-right directions, (b) is a figure which shows the dispersion | variation in the density gradient direction value in each coordinate of all the normalization cut-out images. is there. (A)-(c) is a figure which shows each normalized cut-out image at the time of changing the magnification of the size of a face area. (A)-(c) is a figure which shows each normalized cut-out image at the time of rotating a face area | region.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Imaging means 2 Face detection means 4 Clipping means 5 Cutting condition change means 6 Density gradient image generation means 8 Authentication image generation means 10 Authentication means

Claims (11)

  Image capturing means for capturing an image including the face of the person to be authenticated, face detection means for detecting a face area corresponding to the face from the captured image obtained by the image capturing means, and the size of the image used for authentication from the face area image And a cutout means for cutting out a plurality of normalized cutout images by changing the cutout conditions for cutting out the normalized cutout image according to the angle, the cutout condition changing means for changing the cutout conditions in the cutout means, and the cutout means A density gradient image generating means for generating a plurality of density gradient images by replacing the pixel value of each pixel with a numerical value corresponding to the direction of the density gradient at the position of the target pixel for each of the plurality of normalized cut-out images; Authentication image generation means for generating one authentication image with reduced quantization error from all density gradient images obtained by the generation means, and authentication image generation Face authentication system characterized by comprising an authentication means for performing face authentication using the authentication image obtained in stage.   The face authentication system according to claim 1, wherein the cutout condition changing unit changes the cutout condition by displacing the face region in at least one of the vertical and horizontal directions.   The face authentication system according to claim 2, wherein the clipping condition changing unit changes a displacement amount according to a size of the face area.   The face authentication system according to claim 1, wherein the clipping condition changing unit changes the clipping condition by changing a magnification of the size of the face area.   2. The face authentication system according to claim 1, wherein the clipping condition changing unit changes the clipping condition by changing an angle at which the face area is rotated.   The authentication image generation means calculates an angle difference indicating the direction of the density gradient between the target pixel and the surrounding pixels for all pixels of each density gradient image, and for all pixels, all density gradients in the target pixel. 6. The face authentication system according to claim 2, wherein an average value of angles indicating the direction of the density gradient of the image is used as a pixel value of the authentication image.   The authentication image generation means calculates an angle indicating the direction of the density gradient between the target pixel and the surrounding pixels for all the pixels of each density gradient image, and for all the pixels, all the density gradient images in the target pixel are calculated. The face authentication system according to claim 2, wherein an intermediate value of an angle indicating the direction of the density gradient is used as a pixel value of the authentication image.   The authentication image generation means calculates an angle indicating the direction of the density gradient between the target pixel and the surrounding pixels for all pixels of each density gradient image, and calculates an angle distribution indicating the direction of the density gradient at the target pixel. 6. The face according to claim 2, wherein, for all pixels, an angle indicating a direction of a density gradient having the largest distribution in the target pixel is set as a pixel value of the authentication image. Authentication system.   The authentication image generation means calculates the difference in angle indicating the direction of the density gradient between the authentication image and each density gradient image for all the pixels, and sets a difference result and a first preset in each pixel. The number of images exceeding the first threshold is counted by comparing with the threshold value, and the mask processing is performed when the number of images counted is compared with a preset second threshold value and exceeds the second threshold value. The face authentication system according to any one of claims 6 to 8, wherein:   The face authentication system according to claim 9, wherein the second threshold value is a preset fixed value.   The second threshold value is set so that a ratio between the number of images of the normalized image and the non-normalized image as a whole and the number of images in which the difference in angle indicating the direction of the density gradient exceeds the second threshold value is a predetermined value. The face authentication system according to claim 9.