JP4824636B2 - Face detection apparatus, method and program - Google Patents

Description

  The present invention relates to a face detection apparatus, method, and program.

  Conventionally, a technique for detecting a human face by controlling a predetermined parameter even under various environments has been proposed. When an unidentifiable error occurs, the personal identification device of Patent Document 1 makes the physical feature re-image acquisition result different from the previous one at the time of re-image acquisition of the image acquisition means for acquiring the human physical feature. Specifically, when the face is tilted, the most effective parameter is determined, and the imaging camera is moved and adjusted so that an image of the front face is obtained. When half of the face becomes dark, the most effective parameter is determined, and the illumination light is moved and adjusted so that regular uniform illumination is obtained.

When recognizing a person using a person's face image, the face recognition apparatus of Patent Document 2 obtains an optimal image for recognizing the person without being affected by the surrounding lighting environment or changes in the person's clothes. To do.
Japanese Patent Application Laid-Open No. 11-316836 JP 2005-202731 A

  However, Patent Document 1 does not describe a specific method for obtaining the most effective parameter. For this reason, it is not guaranteed whether the parameter changed by the identification device of Patent Document 1 is optimal. In addition, the face recognition device of Patent Document 2 needs to learn in advance for parameter control, and when performing face matching, the parameter is changed to various values, and the optimum parameter is determined based on the similarity of each face. It must be set, and complicated processing is required.

  For this reason, the techniques of Patent Documents 1 and 2 have a problem that, in a situation where illumination, brightness, and the like change sequentially, the change cannot be dealt with and the face of a person cannot be detected quickly.

  The present invention has been proposed to solve the above-described problems, and provides a face detection apparatus, method, and program capable of detecting a person's face quickly in response to a change in a shooting environment. The purpose is to do.

Face detecting apparatus according to the present invention, the imaging parameters are changed and the image pickup means for generating an image by capturing, a parameter changing means for changing the imaging parameters of said imaging means by said parameter changing means the face of a person The face detection likelihood calculating means for calculating the face detection likelihood indicating the face-likeness based on the image generated by the imaging means and the face detection likelihood calculated by the face detection likelihood calculating means Imaging control means for controlling the imaging means so as to be an imaging parameter when the face looks most like.

  According to the face detection device, each time the imaging parameter is changed, the face detection likelihood indicating the likelihood of a face is calculated based on the image generated by the imaging unit, and the calculated face detection likelihood is the highest. Even if the shooting environment changes, the imaging parameters are changed in accordance with the change by controlling the imaging means so that the imaging parameter is set to indicate the face likeness, so that the face can be detected quickly. it can.

The face detection apparatus according to the present invention includes an imaging unit that captures an image of a person's face and generates an image, an image conversion unit that converts an image generated by the imaging unit, and an image conversion parameter of the image conversion unit is changed. and parameter changing means for, said by parameter changing means every time the image transformation parameters are changed, on the basis of the image conversion unit image generated by the face detection likelihood of calculating a face detection likelihood indicating a likelihood face Calculation means, and image conversion control means for controlling the image conversion means so that the face detection likelihood calculated by the face detection likelihood calculation means becomes an image conversion parameter when the face likelihood is most likely. Face detection device.

  According to the face detection device, each time the image conversion parameter is changed, the face detection likelihood indicating the likelihood of a face is calculated based on the image generated by the image conversion means, and the calculated face detection likelihood is the highest. Even if the shooting environment changes, the image conversion parameters are changed according to the change by controlling the image conversion means so that it becomes the image conversion parameter when the face looks like. can do.

  The face detection apparatus, method, and program according to the present invention can detect a face quickly in response to a change in the shooting environment.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

[First Embodiment]
FIG. 1 is a block diagram showing a configuration of a face detection apparatus according to an embodiment of the present invention. The face detection apparatus includes a camera 10 that captures an image of a person's face, and a computer 20 that performs face detection processing based on an image generated by the camera 10.

  The computer 20 includes a CPU, a RAM, a ROM, an input / output port, and the like. As shown in FIG. 1, the computer 20 functionally includes an image input unit 21 that inputs an image, a face detection unit 22 that performs a human face detection process based on the image input to the image input unit 21, A parameter adjustment unit 23 that adjusts parameters of the camera 10 and a camera control unit 24 that controls the camera 10 based on the adjusted parameters are provided.

  In the face detection apparatus configured as described above, the following face detection processing routine is executed. FIG. 2 is a flowchart showing a face detection processing routine executed by the computer 20. In the present embodiment, a case where a camera gain is used as an imaging parameter will be described.

  In step S11, the parameter adjustment unit 23 sets the reference image parameter pc to the initial value p0, and proceeds to step S12. Here, when the camera gain can be set continuously from 1 to 100, for example, an intermediate value 50 is set as the initial value p0. Further, an optimum value under the use environment of the present apparatus may be set as the initial value p0.

  In step S12, the parameter adjustment unit 23 sets the number of repetitions i to zero, and proceeds to step S13.

  In step S13, the parameter adjusting unit 23 sets the current reference image parameter pc as shown in Expression (1) using the reference image parameter pd and the parameter c (i) for image parameter variation, and the step Proceed to S14.

  Here, c (i) is a parameter depending on the number of repetitions i, and is defined as shown in, for example, Expression (2).

  Here, N is a repeat count threshold value. Δp is a change amount of the image parameter. In this embodiment, Δp = 2, for example. Accordingly, the image parameter pc can be changed based on the reference image parameter pd according to the number of repetitions i.

  For example, when N = 5 times and the current reference image parameter pd = 50, pc = 46 when i = 0, pc = 48 when i = 1, pc = 50 when i = 2, i = 3 When pc = 52 and i = 4, pc = 54. Thus, an image is generated with the image parameter pc being changed little by little.

  In step S14, the camera control unit 24 adjusts the camera gain of the camera 10 so that the image parameter pc currently set in the parameter adjustment unit 23 is obtained, and the process proceeds to step S15. As a result, the camera 10 captures an image of a person with the camera gain set by the camera control unit 24 and generates an image. This image is supplied to the face detection unit 22 via the image input unit 21.

  In step S <b> 15, the face detection unit 22 performs face detection processing based on the image supplied from the image input unit 21, and the face detection likelihood that is the reliability of the detected face (an index indicating the likelihood of a face). φ (i) is calculated. Here, a known image processing method can be used for the face detection process, and a method using a neural network that is a learning type face detection process, a method using a support vector machine, or a boosting method is used. Can do.

  As a technique using a neural network, for example, a technique described in “Rowley et al., Neural Network-Based Face Detection (PAMI, January 1998)” can be used. Further, as a boosting technique, for example, the technique described in “Viola et al., Robust Real-Time Face Detection, International Journal of Computer Vision 57 (2), 137-154, 2004” can be used.

  Then, the face detection unit 22 calculates the face detection likelihood φ (i) as described above, records the face detection likelihood φ (i), and then proceeds to step S16. The face detection likelihood φ (i) used in the present embodiment approaches zero as the image feature detected by the face detection process does not look like a face, and increases as the image feature looks like a face.

  In step S <b> 16, the parameter adjustment unit 23 increments the number of repetitions i by 1, and determines whether or not the number of repetitions i after the increment is smaller than the repetition number threshold N. When the number of repetitions i after the increment is smaller than the repetition number threshold N, the process returns to step S13, and when not smaller, the process proceeds to step S17. As a result, the processing from step S13 to step S15 is repeated until the number of repetitions i reaches the repetition number threshold N.

In step S17, the parameter adjustment unit 23 updates the reference image parameter pd that is optimal for the face detection process. Specifically, if i that gives the maximum likelihood in φ (i) {i = 0,..., N−1} is i _max , the parameter adjustment unit 23 obtains pd + c as shown in Equation (3). Using (i _max ) as a new reference image parameter pd, the process proceeds to step S18.

However, i _max satisfies the following equation (4).

  In step S18, the parameter adjustment unit 23 determines whether or not to continue the process. If the end of the process is not instructed, the process returns to step S12 again to continue the process. Then, the above-described processing is performed using the updated reference image parameter pd. When the end of the process is instructed, the process ends. Then, after this process is completed, the face detection unit 22 detects a person's face based on the image at the updated new reference image parameter pd (camera gain in this embodiment).

  As described above, the face detection apparatus according to the first embodiment images a person using different camera gains, calculates a face detection likelihood representing the likelihood of a face from the generated image, and the face detection likelihood. The camera gain when the degree becomes the most optimal value is used for the next imaging. For this reason, the face detection apparatus can detect the face stably even when the brightness condition, the lighting condition, etc. gradually change when performing face detection continuously.

[Second Embodiment]
Next, a second embodiment of the present invention will be described. In addition, the same code | symbol is attached | subjected to the site | part same as 1st Embodiment, and a different point is mainly demonstrated. In the first embodiment, the camera gain is used as the imaging parameter, but in the second embodiment, a gamma correction coefficient is used as the image conversion parameter instead of the camera gain. At this time, the same processing as in the first embodiment is performed in each step of FIG. 2, but the following processing is performed in the following steps S11, S13, S14, and S17.

  In step S11, the parameter adjustment unit 23 sets, for example, a standard gamma correction coefficient 1.0 as the initial value p0 of the reference image parameter pd.

  In step S13, the parameter adjusting unit 23 sets the image parameter pc by setting Δp in the equation (2) to 0.1. In this case, for example, if N = 5, the image parameter pc takes a value from 0.98 to 1.02.

  In step S <b> 14, the camera control unit 24 controls the camera 10 so that the gamma correction coefficient currently set by the parameter adjustment unit 23 is obtained. As a result, the camera 10 performs gamma correction on the human image using the gamma correction coefficient set by the camera control unit 24, and then outputs the image.

  In step S17, the parameter adjustment unit 23 updates the reference image parameter pd according to Expression (3).

  As described above, in the face detection apparatus according to the second embodiment, the gamma correction coefficient is generally set to 1.0, but the face detection likelihood φ (i) is an optimal value. By setting the gamma correction coefficient, the quality of the image generated by the camera 10 can be improved.

[Third Embodiment]
Next, a third embodiment of the present invention will be described. In addition, the same code | symbol is attached | subjected to the site | part same as embodiment mentioned above, and a different point is mainly demonstrated. In the third embodiment, a threshold value for density conversion is used as the image conversion parameter. The density conversion is performed in the image input unit 21.

  FIG. 3 is a diagram showing a density conversion table representing output levels with respect to input levels. The density conversion table (a) shows the relationship between the input level and the output level when density conversion is not performed. The density conversion table (A) indicates that the input level V is converted to a luminance value 0 and the input level (V + 150) is converted to a luminance value 255. The density conversion table of (a) enables luminance dispersion when the target luminance distribution is concentrated.

  Therefore, in the present embodiment, the image input unit 21 uses this input level V as an image parameter to set a density conversion table that provides an input level V that is optimal for face detection. At this time, in each step of FIG. 2, the same processing as that of the first embodiment is performed, but in the following steps S11, S13, S14, S15, and S17, the following processing is performed.

  In step S11, the parameter adjustment unit 23 sets the initial level so as to satisfy the condition that (1) the width of the input level to be subjected to density conversion is 150, for example, and (2) the input level 128 corresponds to the output level 128. Set the value p0. In the present embodiment, the input level 53 is set as the initial value p0 of the reference image parameter pd.

  In step S13, the parameter adjusting unit 23 sets the image parameter pc as Δp = 3 in the equation (2). In this case, for example, if N = 5, the value of the input level V in FIG. 3 is 47, 50, 53, 56, 59.

  In step S14, the image input unit 21 creates a density conversion table so that the input level V currently set by the parameter adjustment unit 23 is obtained, and performs density conversion of the input image in accordance with the density conversion table.

  In step S <b> 15, the face detection unit 22 executes face detection processing using the image whose density has been converted by the image input unit 21.

  In step S17, the parameter adjustment unit 23 updates the reference image parameter pd according to Expression (3).

  As described above, the invention according to the third embodiment sets the threshold value for the density conversion of the input image (the intercept of the input level in FIG. 3) to an optimal value based on the face detection likelihood. In addition, stable face detection can be realized even with fluctuations in the brightness of the environment and lighting conditions.

  Note that the present invention is not limited to the above-described embodiment, and it is needless to say that the present invention can also be applied to a design modified within the scope of the claims.

  For example, the imaging parameter and the image conversion parameter are not limited to those of the above-described embodiment, and can be applied to a camera electronic shutter speed, an aperture value, or a conversion function for density conversion. Further, the imaging parameters and the image conversion parameters may be combinations of arbitrary image parameters. Note that the imaging parameter and the image conversion parameter may be controlled simultaneously.

It is a block diagram which shows the structure of the face detection apparatus which concerns on embodiment of this invention. It is a flowchart which shows the face detection process routine which a computer performs. It is a figure which shows the density | concentration conversion table showing the output level with respect to an input level.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Camera 20 Computer 21 Image input part 22 Face detection part 23 Parameter adjustment part 24 Camera control part

Claims (8)

Imaging means for imaging a person's face and generating an image;
Parameter changing means for changing imaging parameters of the imaging means;
Each time it is the imaging parameters changed by the parameter changing section, on the basis of the image generated by the imaging unit, a face detecting likelihood calculating means for calculating a face detection likelihood indicating a likelihood face,
Imaging control means for controlling the imaging means so that the face detection likelihood calculated by the face detection likelihood calculating means becomes an imaging parameter when the face likelihood is most likely;
A face detection device. The face detection apparatus according to claim 1, wherein the imaging parameter is at least one of a camera gain, a gamma correction coefficient, an electronic shutter speed, and an aperture value. Imaging means for imaging a person's face and generating an image;
Image conversion means for converting the image generated by the imaging means;
Parameter changing means for changing image conversion parameters of the image converting means;
Each time the image transformation parameter by the parameter changing means is changed, on the basis of the image conversion unit image generated by a face detection likelihood calculating means for calculating a face detection likelihood indicating a likelihood face,
Image conversion control means for controlling the image conversion means so that the face detection likelihood calculated by the face detection likelihood calculation means becomes an image conversion parameter when it is most likely to be a face;
A face detection device. The face detection apparatus according to claim 3, wherein the image conversion parameter is a threshold of an output level with respect to an input level in density conversion, or a conversion function for density conversion. The imaging means images a human face according to the imaging parameters to generate an image,
Each time the imaging parameter is changed, based on the image generated by the imaging means, a face detection likelihood indicating the likelihood of a face is calculated,
A face detection method for controlling the imaging means so that the calculated face detection likelihood becomes an imaging parameter when the face likelihood is most likely. Imaging means generates an image by imaging the face of the person,
The image conversion means converts the image generated by the imaging means according to the image conversion parameter,
Each time the image conversion parameter is changed, based on the image generated by the image conversion means, a face detection likelihood indicating the likelihood of a face is calculated,
A face detection method for controlling the image conversion means so that the calculated face detection likelihood becomes an image conversion parameter when the face likelihood is most likely. Computer
Parameter changing means for changing an imaging parameter of an imaging means for capturing an image of a person and generating an image;
Each time the imaging parameter is changed by the parameter changing unit, based on the image generated by the imaging unit, a face detection likelihood calculating unit that calculates a face detection likelihood indicating the likelihood of a face;
Imaging control means for controlling the imaging means so that the face detection likelihood calculated by the face detection likelihood calculating means becomes an imaging parameter when the face likelihood is most likely;
Face detection program to function as. Computer
An image conversion means for converting an image generated by an imaging means for capturing an image of a person's face and generating an image;
Parameter changing means for changing image conversion parameters of the image converting means;
Each time the image conversion parameter is changed by the parameter changing means, based on the image generated by the image converting means, a face detection likelihood calculating means for calculating a face detection likelihood indicating the likelihood of a face;
Image conversion control means for controlling the image conversion means so that the face detection likelihood calculated by the face detection likelihood calculation means becomes an image conversion parameter when it is most likely to be a face;
Face detection program to function as.

Images