JP2016193175A - Extraction method of determination part for apparent face impression, extraction method of determining factor for apparent face impression, and differentiation method for apparent face impression - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for extracting a portion having a high influence upon the apparent impression of a face and a factor having a high correlation with the apparent face impression, and a method for differentiating the apparent face impression by using the former method.SOLUTION: A apparent face impression determination factor extracting method comprises: a face image analyzing step of analyzing a plurality of face images having photographed images from a plurality of angles different from each other thereby to measure the physical feature amounts of the face varying angle-dependently; and a factor extracting step of extracting a factor high in correlation to the apparent face impression from the physical feature amounts.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a method of extracting a facial appearance impression determination part, a method of extracting a facial appearance impression determinant, a method of identifying a facial appearance impression, a facial impression analysis device, a facial impression analysis program, and a facial appearance impression analysis. It is about the method.

  A person gives various impressions to others due to various factors such as appearance and behavior. Among them, visual information, especially the appearance of the face, is one of the most important factors that determine the person's impression.

  For this reason, research has been conducted on techniques for objectively estimating the appearance of a person's face. For example, Patent Document 1 discloses a technique for measuring the luminance of a measurement target region on the face of a subject, generating a histogram, and evaluating the skin condition of the subject from the shape of the histogram.

  Patent Document 2 discloses a technique for estimating the age of a subject by acquiring a face image of the subject and calculating the intensity of the spatial frequency therefrom.

JP 2009-131336 A No. 2011-162050

  As described above, parameters for objectively estimating the appearance of the subject's face have been studied, but these are only obtained when the subject's face is observed from a certain angle.

  However, human impressions are formed by various factors, and more accurate estimation means have been desired.

  Therefore, the present invention provides a method of extracting a part having a large influence on the appearance impression of the face, a factor having a high correlation with the appearance impression of the face, and a method of distinguishing the appearance impression of the face based on the factor. Let it be an issue.

In order to solve the above problems, the method for extracting the determined portion of the appearance impression of the face according to the present invention is
A visual evaluation step of presenting to the evaluator a plurality of face images obtained by photographing the evaluation target face from a plurality of different angles, and allowing the evaluator to evaluate the appearance impression of the face from each of the face images;
A line-of-sight analysis step of analyzing line-of-sight data of the evaluator who visually observes the face image, and analyzing an area of the face image watched by the evaluator.

  In this way, by analyzing the area that the evaluator gazes at when observing the face from each angle, it is possible to extract a part having a large influence on the judgment of the appearance impression of the face for each observation angle.

  In a preferred embodiment of the present invention, the appearance impression of the face is an age impression.

The method for extracting a determinant of the appearance impression of a face according to the present invention is as follows:
Analyzing a plurality of face images obtained by photographing a face to be evaluated from a plurality of different angles, and measuring physical features of the face that change in an angle-dependent manner, for example, optical features and morphological features, Characteristic feature measurement process,
And a factor extracting step of extracting a factor having a high correlation with the appearance impression of the face from the physical feature amount.

  Thereby, it is possible to specify a physical feature amount used in discrimination of an appearance impression of the face described later.

In a preferred embodiment of the present invention, the physical feature amount measuring step calculates a histogram of luminance values from the face image,
A histogram analysis step of calculating the physical feature quantity based on at least one of skewness, kurtosis, and mode value from the histogram.

In a preferred embodiment of the present invention, the physical feature amount measuring step includes a filter processing step of applying a Gabor filter to the face image;
And a Gabor filter result analyzing step of calculating the physical feature quantity from the wavelength and / or wave directionality included in the application result of the Gabor filter.

The method for extracting a determinant of the appearance impression of a face according to the present invention is as follows:
A physical feature measurement step for measuring the physical feature of the face to be evaluated;
Analyzing the correlation between the appearance impression of each face when viewed from a plurality of angles and the physical feature amount, and from the physical feature amount, the appearance impression of the face viewed from a predetermined angle And a factor extracting step for extracting a factor having a high correlation.

In a preferred embodiment of the present invention, the physical feature amount measuring step includes a three-dimensional model generation step of generating a three-dimensional model of the face,
A homology modeling step of performing homology modeling of the three-dimensional model;
A principal component analysis step of performing a principal component analysis of the homology model and calculating the physical feature amount.

  In a preferred aspect of the present invention, the factor extraction in the factor extraction step is performed for each age group to be evaluated.

In a preferred embodiment of the present invention, the method for extracting a determinant of the appearance of the face has a visual evaluation value acquisition step of obtaining a visual evaluation value of the appearance of the face by visually evaluating the face,
The factor extracting step includes analyzing the correlation between the physical feature amount and the visual evaluation value to extract the factor.
By using the visual evaluation value, it is possible to specify a factor having a high correlation with an impression received when a person visually observes the face.

In a preferred embodiment of the present invention, the method for extracting a determinant of facial appearance impression includes a physiological quantity measurement step of measuring a physiological quantity related to the skin,
The factor extraction step includes extracting the factor by analyzing a correlation between the physical feature amount and the physiological amount.
By using the physiological amount, it is possible to specify the determinant of the appearance impression of the face that is actually highly correlated with the physiological phenomenon in the biological tissue such as the skin.

  In a preferred embodiment of the present invention, the physiological amount is a physical property on the skin surface and / or an internal physical property of the face.

  In a preferred embodiment of the present invention, the appearance impression of the face is an age impression.

The method for distinguishing the appearance impression of the face according to the present invention is as follows.
Based on the correlation between the physical feature amount of the face that changes in an angle-dependent manner and the appearance impression of the face, the appearance impression of the subject's face is distinguished from the physical feature amount measured from the face image of the subject. And
By using the correlation between the physical feature quantity that changes in an angle-dependent manner and the appearance impression of the face, the appearance impression of the face given to others can be more accurately distinguished.

  In a preferred aspect of the present invention, the physical feature amount includes an amount related to at least one of luminance, saturation, hue, and frequency characteristics in the face image.

  In a preferred aspect of the present invention, the physical feature amount of the subject is measured from a plurality of face images taken from different angles.

The method for distinguishing the appearance impression of the face according to the present invention is as follows.
Based on the correlation between the physical feature amount of the face and the appearance impression of the face viewed from a predetermined angle, the appearance impression of the subject's face is determined from the physical feature amount measured from the face and / or face image of the subject. It is characterized by discrimination.
Thereby, the appearance impression of the face seen from the predetermined angle can be distinguished more accurately.

  In a preferred aspect of the present invention, the physical feature amount includes a value related to the three-dimensional shape of the face.

  In a preferred embodiment of the present invention, the identification of the appearance impression of the face is based on a correlation between the physical feature amount and a visual evaluation value of the appearance impression of the face, based on the measurement value of the physical feature amount for the subject. And calculating a visual evaluation value of the appearance impression of the face.

  In a preferred embodiment of the present invention, the discrimination of the appearance impression of the face is based on a correlation between the physical feature quantity and a physiological quantity related to the skin, based on a measurement value of the physical feature quantity of the subject, and the physiology of the subject. Including calculating a target amount.

  In a preferred embodiment of the present invention, the appearance impression of the face is an age impression.

  In a preferred embodiment of the present invention, the correlation is different for each age group of the subject.

The face impression analysis apparatus according to the present invention is
A face impression analysis device that analyzes an appearance impression of a subject's face based on a correlation between a physical feature amount of the face that changes in an angle-dependent manner and an appearance impression of the face,
A face image acquisition means for acquiring a face image;
A physical feature amount calculating means for calculating a physical feature amount from the face image;
Correlation storage means for storing the correlation;
Visual impression analysis means for analyzing the visual impression of the subject's face from the physical feature value of the subject based on the correlation.
By setting it as the apparatus of such a structure, the discrimination method of the appearance impression of the face which concerns on this invention can be implemented.

  In a preferred aspect of the present invention, the physical feature amount includes an amount related to at least one of luminance, saturation, hue, and frequency characteristics in the face image.

In a preferred embodiment of the present invention, the face image acquisition means generates means for generating a three-dimensional model of the subject's face from stereo images of the subject's face taken by at least two cameras, rotation of the three-dimensional model, Alternatively, there is provided means for changing an angle at which the face image is acquired by rotation of a viewpoint for observing the three-dimensional model.
Thereby, the test subject's face image acquisition from a plurality of angles can be performed from the three-dimensional model, and the burden on the test subject can be suppressed even when a large number of face images are acquired.

In a preferred aspect of the present invention, the face image acquisition unit includes a pseudo light source setting unit that sets a pseudo light source to be irradiated to the three-dimensional model when the face image is acquired.
Thereby, it becomes easy to make it the same illumination condition about several face images from which an acquisition angle differs, and it can suppress the influence which it has on physical feature-values, such as an optical feature-value, by illumination conditions. In addition, it is possible to arbitrarily set an illumination angle, an irradiation direction, and the like, and perform evaluation assuming faces in various environments.

The face impression analysis apparatus according to the present invention is
A facial impression analysis device that analyzes a visual impression of a subject's face based on a correlation between a physical feature amount of the face and an appearance impression of the face seen from a predetermined angle,
Physical feature amount calculating means for calculating the physical feature amount of the face;
Correlation storage means for storing the correlation;
Visual impression analysis means for analyzing the visual impression of the subject's face from the physical feature value of the subject based on the correlation.

  In a preferred aspect of the present invention, the physical feature amount includes a value related to the three-dimensional shape of the face.

  In a preferred aspect of the present invention, the facial impression is an age impression.

  In a preferred embodiment of the present invention, the correlation is different for each age group of the subject.

In a preferred embodiment of the present invention, the correlation storage means holds a correlation between the physical feature value and a visual evaluation value of an appearance impression of the face obtained by visual evaluation of the face,
The visual impression analysis means calculates a visual evaluation value of the visual impression of the subject from the physical characteristic amount of the subject based on the correlation between the physical characteristic amount and the visual evaluation value of the visual impression of the face. It is characterized by that.

In a preferred embodiment of the present invention, the correlation storage means holds a correlation between the physical feature quantity and a physiological quantity related to the skin,
The visual impression analysis means calculates a physiological quantity relating to the subject's skin from a physical feature quantity relating to the subject based on a correlation between the physical feature quantity and a physiological quantity relating to the skin. .

  In a preferred embodiment of the present invention, the physiological amount is a physical property on the skin surface and / or an internal physical property of the subject's face.

The face impression analysis program according to the present invention is:
A face impression analysis program for analyzing an appearance impression of a subject's face based on a correlation between a physical feature amount of the face that changes in an angle-dependent manner and an appearance impression of the face,
Computer
Means for extracting the physical feature quantity of the subject;
Based on the correlation, means for distinguishing the appearance impression of the subject from the physical feature amount about the subject;
It is made to function as.

The face impression analysis program according to the present invention is:
A face impression analysis program for analyzing the appearance impression of a subject's face based on the correlation between the physical feature amount of the face and the appearance impression of the face seen from a predetermined angle,
Computer
Means for extracting the physical feature quantity of the subject;
Based on the correlation, means for distinguishing the appearance impression of the subject from the physical feature amount about the subject;
It is made to function as.

The method for analyzing the appearance impression of the face according to the present invention is as follows.
A face impression analysis device comprising: a face image acquisition means, a physical feature quantity extraction means, a correlation storage means, and an appearance impression analysis means for analyzing an appearance impression of the subject's face based on the face image of the subject A method for analyzing the appearance impression of a face using
The face image acquiring means acquiring the face image;
The physical feature amount extracting means extracts the physical feature amount of the face from the face image;
The appearance impression analyzing means is based on the correlation between the physical feature quantity of the face that changes in an angle-dependent manner held by the correlation storage means and the appearance impression of the face, based on the physical feature quantity of the subject. Analyzing the appearance impression of the face of the subject.

The method for analyzing the appearance impression of the face according to the present invention is as follows.
The appearance of the face using a face impression analysis device comprising physical feature amount extraction means, correlation storage means, and appearance impression analysis means for analyzing the appearance impression of the subject's face based on the face image of the subject An impression analysis method,
The physical feature extraction means extracting the physical feature of the face;
The appearance impression analyzing means is based on the correlation between the physical feature quantity held by the correlation storage means and the appearance impression of the face seen from a predetermined angle, based on the physical feature quantity of the subject. Analyzing the appearance impression of the face of the subject.

The method for distinguishing cosmetic effects according to the present invention is:
A method for distinguishing makeup effects by distinguishing the appearance of a face using physical features of the face,
A pre-makeup feature quantity measuring step of extracting the physical feature quantity of the face before makeup;
Based on the correlation between the physical feature amount and the visual evaluation value of the appearance of the face viewed from a predetermined angle, a pre-makeup visual evaluation value calculating step for calculating a pre-makeup visual evaluation value;
A post-makeup feature amount measuring step of extracting the physical feature amount of the face after makeup;
Based on the correlation, a post-makeup visual evaluation value calculating step of calculating a post-makeup visual evaluation value;
A cosmetic effect evaluation step of comparing the visual evaluation value before makeup with the visual evaluation value after makeup and evaluating the makeup effect is provided.

  By distinguishing the appearance impression of the face using the correlation between the factor that changes depending on the angle and the appearance impression of the face, the appearance impression of the face can be more accurately distinguished.

It is a flowchart which shows the extraction method of the determinant of the appearance impression of the face which concerns on this invention. It is a functional block diagram of the face impression analysis device according to the present invention. It is a flowchart which shows the discrimination method of the appearance impression of the face by the face impression analyzer which concerns on this invention. It is a flowchart which shows the calculation process of the visual evaluation value / physiological quantity by the face impression analyzer which concerns on this invention. It is a figure which shows the example of the face image used for acquisition of visual evaluation value. (Drawing substitute photo) It is a figure which shows the presentation method of the face image to the evaluator at the time of acquisition of visual evaluation value. (Drawing substitute photo) It is a figure which shows the total result of the visual evaluation value with respect to a face image. It is a figure which shows the area | region division on the face image used for the tabulation of an evaluator's gaze analysis. (Drawing substitute photo) It is a figure which shows the total result of the gaze analysis of the evaluator with respect to a face image. It is a figure which shows the analysis result of the correlation of the principal component analysis result of the evaluation object of 60s, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 20-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 20-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 20-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 20-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 20-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 30-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 30-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 30-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of 30-year-old evaluation object, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of the evaluation object of 40 years old, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of the evaluation object of 40 years old, and the visual evaluation value of an age impression. It is a figure which shows the analysis result of the correlation of the principal component analysis result of the evaluation object of 50s, and the visual evaluation value of an age impression.

[1] Method for Extracting Face Appearance Impression Determining Section and Method for Extracting Face Appearance Impression Determining Factor According to the Present Invention The present invention analyzes a plurality of face images obtained by photographing faces to be evaluated from different angles. This is a method for extracting a determinant of a facial appearance impression by measuring a physical feature amount of a face that changes in an angle-dependent manner and extracting a factor having a high correlation with the appearance impression of the face. In addition, by analyzing a plurality of face images obtained by photographing the face to be evaluated from different angles, the physical features of the face are measured, and the correlation between the physical impression of the face seen from a predetermined angle and the correlation It includes a method of extracting determinants of facial appearance impressions that extract high factors.

  The “appearance of the face” according to the present invention is an impression that is visually received when another person sees the face to be evaluated. For example, if attention is paid to the age impression, there may be an impression such as a youthfulness that looks younger than the actual age, or an aging impression that looks old. In addition, there are various impressions such as an attractive feeling when viewing the evaluation object. The “physical feature amount of the face that changes depending on the angle” is a physical feature amount that changes depending on the angle at which the face is observed.

  FIG. 1 is a flowchart showing a method for extracting a determinant of an appearance impression of a face according to the present invention. With reference to this, the method for extracting the determinant of the appearance impression of the face according to the present invention will be described in detail.

<Acquisition of face image>
First, in step S11, a face image to be evaluated is acquired. Here, as described above, a plurality of face images obtained by photographing the face to be evaluated from different angles are used.

  The face image is acquired by photographing the evaluation target face image while changing the angle of the evaluation target face or moving the camera. At this time, it is preferable to perform photographing so that the illumination conditions in the plurality of face images are the same.

  Or it is good also as a structure which produces the three-dimensional model of the face of evaluation object, and acquires several face images from there. In this case, first, a stereo image of the face to be evaluated is taken by two or more cameras, and a three-dimensional model is created therefrom. Then, the face image to be evaluated is acquired from a plurality of angles while rotating the three-dimensional model or rotating the viewpoint for observing the three-dimensional model.

  As described above, even when many face images are acquired by using a three-dimensional model, the angle to be evaluated is actually changed because only one time of acquiring a stereo image needs to be taken. However, the burden can be reduced as compared with the method of photographing the face image to be evaluated.

  Also, in terms of setting the shooting angle, it is possible to mechanically rotate the 3D model or the viewpoint for observing the 3D model, so shooting is performed while changing the angle of the face to be evaluated and the camera position. Rather than doing this, it is possible to easily obtain a face image from the intended angle.

  Furthermore, it is also easy to set the same illumination condition for the face image to be evaluated acquired from each angle by irradiating a pseudo light source to the three-dimensional model.

  In addition, in both cases of images obtained by actually photographing the face to be evaluated and images obtained from the three-dimensional model, an average is calculated in order to more accurately analyze the face image from each angle. Correction processing such as brightness adjustment may be performed.

  In addition, in order to extract the appearance impression given by the face to be evaluated more accurately, the correction that deletes the hair, neck, and other parts reflected in the face image, and the part that greatly affects the appearance impression in the face. In order to identify the image, it may be corrected to obtain an image obtained by extracting only an arbitrary part of the face from the face image.

<Facial image analysis>
After acquiring the face image, the analysis is performed in step S12, and the physical feature amount is extracted.

  Examples of the physical feature amount include an optical feature amount such as a brightness histogram of a face image. Specifically, the luminance value of each pixel of the face image is aggregated, and a histogram of luminance values can be obtained by drawing a graph with the horizontal axis representing the luminance value and the vertical axis representing the number of pixels having the luminance value. it can. Furthermore, values such as the skewness and kurtosis of the histogram may be calculated.

  The luminance histogram may be obtained for each of RGB components in a color face image, or may be obtained for the luminance of a monochrome face image or a face image obtained by performing monochrome processing on a color face image. Good.

  Another example of the optical feature amount includes a value related to an angular factor included in the face image. Specifically, by applying a Gabor filter in an arbitrary direction to the face image, it is possible to extract a horizontal component, a vertical component, an oblique component, and the like of a line included in the face image. Furthermore, there is a method of quantifying the angularity factor by calculating an average luminance of the image after applying the Gabor filter, a luminance histogram, and the like.

  In addition, various optical feature amounts such as calculation of frequency characteristics of a face image may be extracted.

  In addition to the optical feature amount, various physical amounts such as a morphological feature amount relating to a three-dimensional or two-dimensional position of each point in the face image may be calculated.

  More specifically, the morphological feature amount may be related to the three-dimensional shape of the face to be evaluated. This can be obtained, for example, by performing homologous modeling of the three-dimensional model of the face to be evaluated as described above and performing principal component analysis. The physical quantity related to the three-dimensional shape of the face obtained in this way is different from the face observation angle, unlike the optical feature quantity obtained by analyzing each face image with a histogram or a Gabor filter.

  Further, for example, in the case where only the morphological feature amount is used as the physical feature amount without using the optical feature amount, instead of acquiring the face image in step S11, the face to be evaluated by another measuring device is used. The three-dimensional shape may be measured, and the measurement result may be adopted as the physical feature amount.

<Acquisition of visual evaluation value>
In step S13, a visual evaluation value for the face image is acquired. This is done by presenting the face image acquired in step S11 to the evaluator and evaluating the impression received by the evaluator from the face image.

  For example, if the purpose is to extract the determinants of the age impression from among the appearance impressions of the face, a plurality of face images with different acquisition angles are presented to the evaluator, There is a method of requesting selection of a face image that looks old. As another example, if the purpose is to extract determinants of the appearance impression of a face related to attraction, after presenting multiple facial images with different acquisition angles to the evaluator as in the case of an age impression And a method for obtaining selection of the most attractive facial image among them. The visual evaluation value thus obtained can be said to be a value that directly indicates the appearance impression of the face to be evaluated.

  When the face image is presented to the evaluator, the gaze information of the evaluator is acquired, and information such as how many times and for what seconds the evaluator has kept the gaze in which area of the face to be evaluated is acquired. .

<Impression determination site extraction>
After the visual evaluation value is acquired in step S13, the gaze information of the evaluator when the face image is presented to the evaluator is analyzed, and the impression determination site is extracted.

  When an evaluator who has evaluated an impression observes a facial image, the evaluator judges the impression regardless of whether it is conscious or unconscious for a region where the gaze is retained for a long time. It can be said that it includes a region to be watched when performing, that is, a determination site of the impression.

  For example, telling the evaluator that an answer for the estimated age of the target displayed on the face image is to be obtained, and when the face image is presented, the area where the evaluator's line of sight has stayed for a long time Includes a decision site.

  Here, as described above, the region where the gaze of the evaluator stayed a lot during the visual evaluation and its periphery are extracted as the determined portions of the appearance impression.

  Note that, for each of the plurality of face images with different acquisition angles presented to the evaluator in step S13, the evaluator's line-of-sight information is analyzed, and the determined part of the appearance impression when observed from each angle is extracted. It is preferable.

<Measurement of physiological quantity>
In step S14, a physiological quantity related to the skin of the face to be evaluated is measured.
Examples of the physiological amount relating to the skin include physical properties and / or internal physical properties of the skin, composition of the skin tissue, and the like, and those that are known to correlate with the appearance of the face can be used. For example, when the appearance impression of the face is an age impression, a physiological amount relating to the skin whose mechanism for aging has been clarified can be used. Such physiological amounts include physical properties of the skin surface, such as skin surface roughness, physical properties of the horny layer epidermis, such as moisture content of the horny layer, physical properties of the dermis, such as collagen fiber bundle amount and structure, physical properties of the fat layer, such as fat The physical properties of the nipple layer or the nipple layer, such as the position and size of the body, for example, the number per unit area of the nipple structure, the height, and the like. The physiological amount is preferably the physiological amount of the impression determining site.
The physiological amount can be measured by a conventionally known method. In the present invention, the present invention is limited to a case of performing by a non-invasive method (a method not involving surgery).

  Steps S11 to S14 are performed on a plurality of different faces that are sufficient to analyze the next correlation.

<Extraction of visual impression determinants>
In step S15, a determinant of the appearance impression of the face is extracted from the physical feature amount acquired from the face image in step S12.

  Here, the correlation between the physical feature amount acquired in step S12 and the visual evaluation value acquired in step S13 or the physiological amount measured in step S14 is analyzed, and the physical feature amount is visually checked. A factor that determines an evaluation value or a physiological amount is extracted.

For example, two face images acquired from different angles of evaluation targets with different visual evaluation values are compared, and among the physical feature amounts respectively acquired from the face images, there are differences between the two face images Is predicted to have a correlation with the visual evaluation value. Then, by comparing the face images from different angles in the same way, the physical features that have a correlation with the visual evaluation values and are the determinants of the visual impression, and the physical features The correlation with the visual evaluation value can be clarified.
The analysis of the correlation can be performed using a conventionally known method.

  Also, if there is a difference when comparing visual evaluation values for face images from different angles for the same face to be evaluated, it is a factor that correlates with the visual impression when the face is viewed from a predetermined angle. Can be considered. Therefore, the correlation between the visual evaluation value of the appearance impression of the face viewed from a predetermined angle and the physical feature amount that does not change depending on the angle like the three-dimensional shape of the face described above is analyzed. It is also effective.

  Similarly, by analyzing the correlation between the physical feature amount obtained from the face image and the physiological amount related to the skin, the physical feature amount highly correlated with the physiological amount related to the skin can be determined. If it is known that the physiological quantity is involved in the appearance impression of the face, it can be estimated that it is a determinant of the appearance impression of the face.

  As described above, according to the present invention, it is possible to extract determinants of the appearance impression of a face by extracting physical feature amounts from a plurality of face images acquired from different angles and analyzing them. .

  It should be noted that both the acquisition of visual evaluation values and the analysis of the correlation with the physical feature amount and the measurement of the physiological amount and the analysis of the correlation with the physical feature amount may be both performed, or as necessary. Only one of them may be performed. Further, different evaluation values and amounts may be used, and the correlation between the evaluation values and amounts and physical feature amounts extracted from a plurality of face images having different angles may be analyzed.

  Moreover, you may perform the extraction of the determinant of the appearance impression of a face for every age group of evaluation object like 20s, 30s, 40s, for example. Thus, if the analysis for each age group is performed, the correlation between the physical feature amount and the visual impression can be analyzed in more detail.

[2] A method for distinguishing an appearance of a face and an apparatus for analyzing a face impression according to the present invention The present invention acquires a face image of a subject and calculates a physical feature amount by analyzing the face image, thereby the appearance of the face of the subject. A method for distinguishing impressions, and a face impression analyzing apparatus for distinguishing appearance impressions of a face.

  As shown in the functional block diagram of FIG. 2, the face impression analysis apparatus according to the present invention includes a face image acquisition unit 1, a physical feature amount calculation unit 2, a correlation storage unit 3, and a visual evaluation value calculation unit 4. And a physiological quantity calculation means 5.

  The face image acquisition unit 1 further includes a stereo image acquisition unit 11, a 3D model generation unit 12, and a 3D model rotation unit 13.

  The correlation storage means 3 is an expression showing the correlation obtained by analyzing the physical feature quantity and the visual evaluation value in the process of extracting the determinant of the appearance impression of the face, and the physical feature quantity and the physiological Information such as an expression indicating the correlation obtained by analyzing the quantity is held.

  In addition, the face impression analysis apparatus according to the present invention uses a general-purpose computer device including an arithmetic device, a main storage device, an auxiliary storage device, and various input / output devices. This is realized by storing a face impression analysis program for causing the computer device to function as each means described above. And when making this computer apparatus function as a face impression analysis apparatus, a face impression analysis program is expand | deployed on a main memory as needed, and the process by an arithmetic unit is performed.

  In addition, a configuration in which a camera device is used as the face image acquisition unit 1 and a face image obtained thereby is transferred to the physical feature amount calculation unit 2, a configuration in which the correlation storage unit 3 is realized by a general-purpose server device, etc. Any plurality of devices may be configured to be able to communicate via a network such as the Internet or a LAN (Local Area Network) such as USB (Universal Serial Bus) connection or data transfer using a recording medium.

<Acquisition of face image>
FIG. 3 is a flowchart showing a method for distinguishing an appearance impression of a subject's face by the face impression analysis apparatus. First, in step S21, the face image of the subject is acquired using the face image acquisition means 1.

  Here, it is preferable to obtain the face image of the subject from an angle at which a determinant of the appearance impression of the face appears strongly. Similarly to the face image acquisition step in the method for extracting the determinants of the appearance impression of the face, the subject's face is photographed multiple times while changing the angle of the subject's face or moving the camera, and a plurality of face images are obtained. You may get it.

  Alternatively, a three-dimensional model of the subject's face may be created, and one or a plurality of face images may be acquired therefrom. In this case, a stereo image obtained by photographing the subject's face with two or more cameras is acquired using the stereo image acquiring unit 11, and a three-dimensional model is generated using the three-dimensional model generating unit 12. And the angle which observes a test subject's face is set by performing a rotation process to the three-dimensional model using the three-dimensional model rotation means 13, and a face image is acquired.

  As described above, when the configuration using the three-dimensional model is used, the acquisition angle of the face image can be easily set, and further, by irradiating a pseudo light source to the three-dimensional model, various illumination conditions can be obtained. It is also possible to acquire a face image of the assumed subject.

<Calculation of physical features>
In step S22, the physical feature amount calculation unit 2 calculates the physical feature amount from the face image acquired in step S21. Here, in particular, the physical feature amount extracted as the determinant of the appearance impression of the face is calculated by the method for extracting the determinant of the appearance impression of the face described above.

  For example, as calculated in the extraction of determinants of the appearance impression of a face, extraction of optical features such as histogram of facial image brightness, extraction of angular factors using a Gabor filter, morphological feature, etc. Perform extraction.

  When a three-dimensional shape of a three-dimensional model of a face is used as the morphological feature amount, a three-dimensional model group of faces is prepared in advance, homologous modeling with them is performed, and principal component analysis is performed.

  Further, for example, in the case where only the morphological feature amount is used as the physical feature amount without using the optical feature amount, instead of acquiring the face image in step S21, the face to be evaluated by another measuring device. The solid shape may be measured, and the physical feature amount may be calculated from the measurement result in step S22.

<Analysis of appearance of face>
In step S23, the appearance impression of the face is analyzed from the physical feature amount calculated in step S12. Accordingly, at least one of a visual evaluation value of the appearance impression of the subject's face and a physiological amount related to the skin of the subject's face is calculated.

  FIG. 4A is a flowchart showing a visual evaluation value calculation process by the visual evaluation value calculation means 4. First, in step S31, correlation data between the physical feature quantity and the visual evaluation value is acquired from the correlation storage means 3. Here, as the correlation data, data for each age group of the subject may be prepared and used.

  In step S32, a visual evaluation value is calculated based on the physical feature amount acquired in step S22 and the correlation data acquired in step S31. From the calculated visual evaluation value, the appearance impression of the face of the subject can be distinguished.

  FIG. 4B is a flowchart showing a physiological quantity calculation process by the physiological quantity calculation means 5. This is also the same process as the visual evaluation value calculation process described above. First, in step S41, correlation data between the physical feature quantity and the physiological quantity is acquired from the correlation storage means 3.

  In step S42, the physiological quantity is calculated based on the physical feature quantity acquired in step S22 and the correlation data acquired in step S41. From the calculated physiological quantity, the appearance impression of the subject's face can be differentiated.

  As described above, based on the correlation between the physical feature amount and the visual evaluation value or the physiological amount, the visual evaluation value and / or the physiological amount is calculated from the physical feature amount extracted from the face image of the subject, The appearance impression of the subject's face can be differentiated.

  If the stereo image acquisition unit 11, the 3D model generation unit 12, and the 3D model rotation unit 13 are used, the face image of the subject can be easily acquired from one or a plurality of arbitrary angles as described above. However, the stereo image acquisition means 11 is more expensive than a general monocular camera. Therefore, the stereo image acquisition unit 11 may be omitted, and a general monocular camera or the like may be provided as the face image acquisition unit 1 to acquire the face image of the subject. With such a configuration, the face impression analysis device can be realized at low cost.

  In addition, as a method for distinguishing an appearance impression of a face and a method for using a face impression analyzer according to the present invention, a method for distinguishing a makeup effect can be cited. That is, by comparing the analysis result of the appearance impression performed using the face image of the subject before the makeup with the analysis of the appearance impression performed using the image of the face of the subject after the makeup, It distinguishes the effect on the visual impression.

  In the following, the present invention will be described in more detail with reference to examples for obtaining visual evaluation values of apparent age impressions, extracting impression determination sites, and measuring physical feature amounts. Needless to say, the present invention is not limited to the examples.

[1] Acquisition of visual evaluation value <Acquisition of face image>
A three-dimensional model of the face to be evaluated was created from the stereo images of the face taken using a stereo camera for a total of 280 evaluation objects, 56 persons for each age group in the 20s to 60s. Here, portions such as hair, ears, and neck were removed, and a three-dimensional model representing only the face to be evaluated was used.

  Then, the face image was created by rotating the three-dimensional model and setting six angles that frequently appear when looking at the face of others in daily life. FIG. 5 illustrates seven face images (a) to (g).

  FIG. 5A is a face image when the face to be evaluated is viewed from the front. Different rotation processes were performed on the three-dimensional model in this state, and face images with angles shown in FIGS. 5B to 5G were obtained. FIG. 5B is a face image obtained by rotating the three-dimensional model around the rotation axis in the vertical direction, and viewing the face to be evaluated from an angle of 45 degrees to the right. FIG. 5C is a face image obtained by rotating the three-dimensional model around the rotation axis in the horizontal direction and looking down from the upper side of the face to be evaluated. FIG. 5D is a face image obtained by rotating the three-dimensional model around the rotation axis in the horizontal direction and looking up at the target face from the lower front side. FIG. 5E is a face image obtained by adding both the rotations of FIG. 5B and FIG. 5D to the three-dimensional model and looking up the face to be evaluated from the lower right side. FIG. 5 (f) is a face image obtained by adding both the rotations of FIG. 5 (b) and FIG. 5 (c) to the three-dimensional model, and looking down at the face to be evaluated from the upper right side. FIG. 5G shows a face image obtained by rotating the three-dimensional model around the rotation axis in the vertical direction and viewing the face to be evaluated from an angle of 90 degrees to the right.

<Acquisition of visual evaluation value>
As described above, the face image to be evaluated acquired from the three-dimensional model was presented to 112 evaluators using a monitor, and the visual evaluation value of the age impression of the face appearance was acquired. When presenting the face image to the evaluator, the gaze analysis device was used to analyze how many seconds and how long the evaluator's gaze stayed in which area of the face image. FIG. 6 shows a schematic diagram of a method for presenting a face image to an evaluator.

  First, as shown in FIG. 6 (a), two images are displayed on the monitor, and instructions are given to ask for an answer as to which image appears to be older by looking at them. did.

  Next, as shown in FIG. 6 (b), a symbol as a fixed viewpoint is displayed at the center of the screen, and after prompting the evaluator to see it, as shown in FIG. 6 (c), as shown in FIG. A face image when the face to be evaluated as seen in a) is viewed from the front is presented. By displaying the face image after displaying the fixed viewpoint in this way, the initial position of the evaluator's line of sight when presenting each face image can be made the same, and the evaluator's line of sight analysis can be performed more accurately. .

  After a fixed time has elapsed, as shown in FIG. 6 (d), the same fixed viewpoint as in FIG. 6 (b) is displayed again, and as shown in FIG. 6 (e), FIGS. One of the face images when the face to be evaluated is viewed from a position other than the front is displayed. Then, an answer was obtained as to which of the face image from the front displayed in FIG. 6C and the angled face image displayed in FIG.

  FIG. 7 shows the totaled results of visual evaluation values acquired from 112 evaluators for each evaluation target face image. This shows the proportion of the angled face images shown in FIGS. 5B to 5G that are answered to look older than the face image seen from the front. That is, the vertical axis is an index representing the age impression received by the evaluator. If this value is large, the age impression of the face image viewed from that angle is higher than the front face, and if the value is small, It can be said that the age impression of the face image viewed from this angle is lower than that of the front face.

  According to the result of the aggregation, it was a face image when the face to be evaluated was viewed from below as shown in FIG. 5D that the age impression was lower than the impression of the front face.

  From this result, the angled face images as shown in FIGS. 5B, 5D, 5E, 5F, and 5G have different physical feature amounts from the face image from the front. , It has an influence on the age impression felt by the evaluator, that is, there are some physical features that change in an angle-dependent manner that are determinants of the apparent age impression. Is inferred.

[2] Extraction of Impression Determination Part Here, as described above, the front (a), the right 45 degrees (b), the upper (c), the lower (d), the lower right (as shown in FIG. e) of the evaluator's line-of-sight information acquired by the line-of-sight analysis device when each face image when the face to be evaluated is viewed from seven angles of the upper right (f) and right 90 degrees (g) is presented. Analysis was performed.

  FIG. 8 shows the eye area L1, the nose area L2, the mouth area L3, and the area around the eyes set for each of the face images from the seven angles shown in FIG. It is a figure which shows six areas, the area | region L4, the cheek area | region L5, and the forehead area | region L6.

  FIG. 9 shows the results obtained by measuring the staying time of the evaluator's line of sight to each of these areas L1 to L6. As can be seen from the front face image shown in FIG. 5 (a), the evaluator's line of sight has a long stay in the eye area L1, while the face image from other angles. Resulted in a long gazing time in the cheek region L5.

  From this result, it was suggested that the cheek region is an important region for judging the age impression when the face to be evaluated is viewed from an angle other than the front.

[3] Measurement of physical feature amount As described above, a visual evaluation value of an age impression for a 7-angle face image as shown in FIGS. A three-dimensional model representing the three-dimensional shape of the face was created for a total of 280 evaluation subjects, 56 people for each age group in the 60s to 60s. Then, as a physical feature measurement process, homology modeling and principal component analysis are performed for each age group, and the correlation between the visual evaluation value of the age impression for each observation angle and the principal component score obtained by principal component analysis is calculated. Analyzed.

  FIG. 10 shows the ninth principal component obtained by principal component analysis and the visual evaluation value of the age impression when viewed from each angle of FIGS. Show the relationship. FIG. 10A shows the correlation coefficient between the visual evaluation value of the age impression and the ninth principal component.

  FIG. 10B shows the visual evaluation value and the main evaluation score with the vertical axis as the visual evaluation value of the age impression when the face is viewed from the front (FIG. 5A) and the horizontal axis as the main component score of the ninth main component. It is a figure which shows the relationship with a component score. From this, it can be said that the correlation between the visual evaluation value of the age impression when viewed from the front and the ninth principal component is weak. The correlation coefficient shown in FIG. 10A is also a low value of 0.179. FIG. 10C shows a three-dimensional model of the face when the ninth principal component score is low, and FIG. 10D shows the face 3 when the ninth principal component score is high. Images are shown when the dimensional model is viewed from the front.

  FIG. 10 (e) shows the visual evaluation value with the vertical axis as the 45 ° right angle (visual evaluation value of the age impression when the face is viewed from FIG. 5 (b), and the horizontal axis as the main component score of the ninth principal component. 10 is a diagram showing the relationship between the value and the principal component score, and it can be seen that the visual evaluation value tends to be low when the principal component score of the ninth principal component is high. The correlation coefficient shown has an absolute value of -0.490, which is larger than that seen from the front, and Fig. 10 (f) shows the three-dimensional face when the principal component score of the ninth principal component is low. Fig. 10 (g) shows an image when the three-dimensional model of the face when the principal component score of the ninth principal component is high is viewed from an angle of 45 degrees to the right. Compared with the case seen from the front shown in (d), when the main component score is low, the intensity of cheek undulation is high, and the main component score is high. In this case, the cheek is the impression that stretched, has become a thing to be felt more strongly.

  As a result, for an evaluation target in the 60s, when a three-dimensional model of a face is observed from an angle of 45 degrees to the right, or a face image obtained by photographing a face from an angle of 45 degrees is analyzed, It is considered that age impressions can be identified by extracting.

  FIG. 10 (h) shows the visual evaluation value of the age impression when the face is viewed from the lower side (FIG. 5 (d)), and the visual evaluation value with the horizontal axis as the main component score of the ninth main component. It is a figure which shows the relationship with a main component score. As can be seen from the graph, when the main component score of the ninth principal component is high, the visual evaluation value tends to be low, as in the case of viewing from the right 45 degree angle. The correlation coefficient shown in FIG. 10A also has an absolute value of −0.288, which is larger than that seen from the front. FIG. 10 (i) shows the three-dimensional model of the face when the ninth principal component score is low, and FIG. 10 (j) shows the three-dimensional model of the face when the ninth principal component score is high. The images when viewed from below are shown. Compared with the case seen from the front shown in FIGS. 10 (c) and 10 (d), similar to FIGS. 10 (f) and 10 (g), when the main component score is low, the intensity of cheek undulation is mainly When the component score is high, the cheek impression is felt more strongly.

  From this, it is possible to extract a value related to the undulation of the cheek by analyzing a face image obtained by observing a three-dimensional model of the face from the lower side or an image of the face taken from the lower side for an evaluation target in the 60s It is considered effective in distinguishing age impressions.

  As described above, there is a determinant of age impression that cannot be obtained when the face is observed from the front, but is obtained when the subject is observed from the angle of 45 degrees to the right or from the lower side. Can be considered.

  As in the case of the 60s shown in FIG. 10, for each age group, the principal components that are correlated with the visual evaluation value of the age impression and the appearance of the face associated therewith are shown below.

  FIG. 11A is a graph showing the correspondence between the eighth principal component score to be evaluated in the 20s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the evaluation target in the 20s, the higher the eighth principal component score, the higher the estimated age. FIG. 11B shows a face when deformed in the −3SD direction from the average of the eighth principal component score, and FIG. 11C shows a face deformed in the + 3SD direction. When these are seen, it can be seen that when the main component score is high, the size of the eyebrows and cheekbones is large and the cheeks appear to be overwhelmed.

  FIG. 12A is a graph showing the correspondence between the sixth principal component score to be evaluated in the 20s and the visual evaluation value of the age impression from the angle shown in FIG. 5C. As shown here, for the evaluation target in the 20s, the result is that the estimated age increases as the sixth principal component score increases. FIG. 12B shows a face when deformed in the −3SD direction from the average of the sixth principal component scores, and FIG. 12C shows a face when deformed in the + 3SD direction. From these, it can be seen that when the main component score is high, the face is deep and the nose and mouth appear to protrude forward.

  FIG. 13A is a graph showing the correspondence between the ninth principal component score to be evaluated in the 20s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the evaluation target in the 20s, the lower the ninth principal component score, the higher the estimated age. FIG. 16B shows the face when deformed in the −3SD direction from the average of the ninth principal component score, and FIG. 16C shows the face when deformed in the + 3SD direction. From these, when the principal component score is low, the mouth protrudes forward than the nose, and when the principal component score is high, the nose appears to protrude forward from the mouth.

  FIG. 14A is a graph showing the correspondence between the third principal component score to be evaluated in the 20s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the evaluation target in the 20s, the lower the third principal component score, the higher the estimated age. FIG. 14B shows the face when deformed in the −3SD direction from the average of the third principal component scores, and FIG. 14C shows the face when deformed in the + 3SD direction. From these, it can be seen that when the principal component score is low, the chin position appears behind the mouth, and when the principal component score is high, the chin position appears to be ahead of the mouth. .

  FIG. 15A is a graph showing the correspondence between the 10th principal component score to be evaluated in the 20s and the visual evaluation value of the age impression from the angle shown in FIG. 5B. As shown here, for the 20-year-old evaluation target, a result that the estimated age is higher as the tenth principal component score is higher is obtained. FIG. 15B shows the face when deformed in the −3SD direction from the average of the tenth principal component score, and FIG. 15C shows the face when deformed in the + 3SD direction. From these, it can be seen that when the main component score is low, the nose is low and the chin is large, and when the main component score is high, the nose is high and the chin is small.

  From the above, for the evaluation target in the 20s, the state of cheek dandruff when the face is observed from the upper right, the depth of face flare when the face is observed from above, the mouth when the face is observed from the lower right It was suggested that age impressions could be differentiated by analyzing the positional relationship between the front and back of the nose and the positional relationship between the front and back of the chin and mouth when the face was observed from below.

  FIG. 16A is a graph showing the correspondence between the first principal component score to be evaluated in the 30s and the visual evaluation value of the age impression from the angle shown in FIG. 5B. As shown here, for the 30-year-old evaluation target, the higher the first principal component score, the higher the estimated age. FIG. 16B shows the face when deformed in the −3SD direction from the average of the first principal component scores, and FIG. 16C shows the face when deformed in the + 3SD direction. From these, it can be seen that the face size appears larger when the score of the principal component is higher than when the score is low.

  FIG. 17A is a graph showing the correspondence between the fourth principal component score to be evaluated in the thirties and the visual evaluation value of the age impression from the angle shown in FIG. 5B. As shown here, for the 30-year-old evaluation target, the estimated age was higher as the fourth principal component score was lower. FIG. 17B shows the face when deformed in the −3SD direction from the average of the fourth principal component scores, and FIG. 17C shows the face when deformed in the + 3SD direction. From these, it can be seen that the difference in jaw size appears to be large depending on the size of the main component score.

  FIG. 18A is a graph showing the correspondence between the first principal component score to be evaluated in the 30s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the 30-year-old evaluation target, the higher the first principal component score, the higher the estimated age. FIG. 18B shows a face when deformed in the −3SD direction from the average of the first principal component scores, and FIG. 18C shows a face when deformed in the + 3SD direction. From these, it can be seen that the face size appears larger when the score of the principal component is higher than when the score is low.

  FIG. 19A is a graph showing the correspondence between the ninth principal component score to be evaluated in the 30s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the evaluation target in the 30s, the result is that the estimated age is higher as the ninth principal component score is lower. Further, FIG. 19B shows the face when deformed in the −3SD direction from the average of the ninth principal component score, and FIG. 19C shows the face when deformed in the + 3SD direction. When these are seen, it can be seen that a particularly large difference appears in the shape of the cheek depending on the size of the main component score.

  From the above, for the evaluation target in the 30s, the size of the face when the face is observed from an angle of 45 degrees to the right, the size of the jaw when the face is observed from the angle of 45 degrees to the right, By analyzing the size of the face when observed, the shape of the cheek when the face is observed from the lower side, etc., it was suggested that age impressions could be differentiated.

  FIG. 20A is a graph showing the correspondence between the fifth principal component score to be evaluated in the 40s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the 40-year-old evaluation target, the result that the estimated age is higher as the fifth principal component score is lower was obtained. Further, FIG. 20B shows the face when deformed in the −3SD direction from the average of the fifth principal component scores, and FIG. 20C shows the face when deformed in the + 3SD direction. From these, it can be seen that the face looks long when the main component score is high.

  FIG. 21A is a graph showing the correspondence between the second principal component score to be evaluated in the 40s and the visual evaluation value of the age impression from the angle shown in FIG. 5C. As shown here, for the 40-year-old evaluation target, the estimated age was higher as the second principal component score was lower. FIG. 21B shows a face when deformed in the −3SD direction from the average of the second principal component scores, and FIG. 21C shows a face deformed in the + 3SD direction. When these are seen, it can be seen that the face is deeply visible when the main component score is high.

  As described above, for the evaluation target in the 40's, by analyzing the face length when the face is observed from the lower right, the depth of the face when the face is observed from the upper side, etc., The possibility of discrimination was suggested.

  FIG. 22A is a graph showing the correspondence between the fourth principal component score to be evaluated in the 50s and the visual evaluation value of the age impression from the angle shown in FIG. As shown here, for the evaluation target in the 50s, the result was that the higher the fourth principal component score, the higher the estimated age. FIG. 22B shows the face when deformed in the −3SD direction from the average of the fourth principal component scores, and FIG. 22C shows the face when deformed in the + 3SD direction. From these, it can be seen that when the principal component score is low, the chin position appears to be ahead of the mouth, and when the principal component score is high, the chin position appears to be behind the mouth. .

  From this, it was suggested that age impressions could be differentiated by analyzing the position of the jaw when the face was observed from the lower side of the evaluation target in the 50s.

  In this embodiment, in the horizontal direction, the face is viewed from the right side, such as 45 degrees right (b), lower right (e), upper right (f), and right 90 degrees (g) shown in FIG. Although the face image when observed is used, it is considered that the same result can be obtained when the face is observed from the left side. Therefore, the discrimination of the age impression according to the present invention can be similarly performed using the face image when the face is observed from either side in the left-right direction.

  In this way, even for faces with the same three-dimensional shape, the visual evaluation value of the age impression varies depending on the angle at which it is viewed, and there is a factor that shows a correlation with the age impression when viewed from a specific angle Was suggested. By using such factors, it is considered that detailed age impressions can be differentiated for each age group and for each angle at which the face is observed.

  INDUSTRIAL APPLICABILITY The present invention can be used to easily and accurately discriminate a customer's appearance age impression in scenes such as cosmetic sales, counseling, and aesthetics.

DESCRIPTION OF SYMBOLS 1 Face image acquisition means 11 Stereo image acquisition means 12 3D model generation means 13 3D model rotation means 2 Physical feature-value calculation means 3 Correlation storage means 4 Visual evaluation value calculation means 5 Physiological quantity calculation means

Claims (37)

A visual evaluation step of presenting to the evaluator a plurality of face images obtained by photographing the evaluation target face from a plurality of different angles, and allowing the evaluator to evaluate the appearance impression of the face from each of the face images;
Analyzing each line of sight of the evaluator who views the face image, and analyzing a region in which the evaluator gazes at each face image, for each observation angle, To extract the determined part of the appearance impression of the face.   2. The method for extracting a determined part of an appearance impression of a face according to claim 1, wherein the appearance impression of the face is an age impression. Analyzing a plurality of face images obtained by photographing the face to be evaluated from a plurality of different angles, and measuring a physical feature amount of the face that changes in an angle-dependent manner;
A factor extracting step of extracting a factor having a high correlation with the appearance of the face from the physical feature amount, and a method for extracting a determinant of the appearance impression of the face. The physical feature amount measuring step calculates a histogram of luminance values from the face image;
The face analysis method according to claim 3, further comprising: a histogram analysis step of calculating the physical feature amount based on at least one of skewness, kurtosis, and mode value from the histogram. To extract determinants of appearance impressions. The physical feature measurement step includes a filter processing step of applying a Gabor filter to the face image;
The Gabor filter result analysis step of calculating the physical feature amount from the wavelength and / or the directionality of the wave included in the application result of the Gabor filter, characterized by comprising: Extraction method of determinants of facial appearance impression. A physical feature measurement step for measuring the physical feature of the face to be evaluated;
Analyzing the correlation between the appearance impression of each face when viewed from a plurality of angles and the physical feature amount, and from the physical feature amount, the appearance impression of the face viewed from a predetermined angle And a factor extracting step for extracting a factor having a high correlation. A method for extracting a factor that determines the appearance impression of a face. The physical feature measurement step includes a three-dimensional model generation step of generating a three-dimensional model of the face;
A homology modeling step of performing homology modeling of the three-dimensional model;
The method according to claim 6, further comprising: a principal component analysis step of performing a principal component analysis of the homologous model and calculating the physical feature amount.   The method for extracting a determinant of appearance impression of a face according to any one of claims 3 to 7, wherein the extraction of the factor in the factor extraction step is performed for each age group to be evaluated. Visual evaluation value acquisition step of obtaining a visual evaluation value of the appearance impression of the face by visual evaluation of the face,
9. The factor extraction step according to claim 3, wherein the factor extraction step includes analyzing the correlation between the physical feature amount and the visual evaluation value to extract the factor. To extract determinants of appearance impression of human face. Having a physiological quantity measuring step for measuring a physiological quantity relating to the skin;
The factor extraction step includes extracting the factor by analyzing a correlation between the physical feature amount and the physiological amount, according to any one of claims 3 to 9. To extract determinants of appearance impression of human face.   The method of extracting determinants of facial appearance impression according to claim 10, wherein the physiological amount is a physical property and / or an internal physical property of the face of the face.   12. The method for extracting a determinant of an appearance impression of a face according to claim 3, wherein the appearance impression of the face is an age impression.   Based on the correlation between the physical feature amount of the face that changes in an angle-dependent manner and the appearance impression of the face, the appearance impression of the subject's face is distinguished from the physical feature amount measured from the face image of the subject. The method of distinguishing the appearance impression of the face.   The method according to claim 13, wherein the physical feature amount includes an amount related to at least one of luminance, saturation, hue, and frequency characteristics in the face image. .   The method according to claim 13 or 14, wherein the physical feature amount of the subject is measured from a plurality of face images taken from different angles.   Based on the correlation between the physical feature amount of the face and the appearance impression of the face viewed from a predetermined angle, the appearance impression of the subject's face based on the physical feature amount measured from the subject's face and / or face image A method of distinguishing the appearance impression of a face, characterized by distinguishing between.   The method according to claim 16, wherein the physical feature amount includes a value related to the three-dimensional shape of the face.   The identification of the appearance impression of the face is based on the correlation between the physical feature amount and the visual evaluation value of the face appearance impression, and the visual impression of the subject's face is determined from the measured physical feature amount of the subject. 18. The method for distinguishing an appearance impression of a face according to claim 13, further comprising calculating an evaluation value.   The identification of the appearance impression of the face calculates the physiological quantity of the subject from the measured value of the physical feature quantity of the subject based on the correlation between the physical feature quantity and the physiological quantity related to the skin. The method for distinguishing an appearance impression of a face according to any one of claims 13 to 18, characterized by comprising:   20. The method for distinguishing the appearance impression of a face according to claim 13, wherein the appearance impression of the face is an age impression.   21. The method for distinguishing the appearance impression of a face according to any one of claims 13 to 20, wherein the correlation is different for each age group of the subject. A face impression analysis device that analyzes an appearance impression of a subject's face based on a correlation between a physical feature amount of the face that changes in an angle-dependent manner and an appearance impression of the face,
A face image acquisition means for acquiring a face image;
A physical feature amount calculating means for calculating a physical feature amount from the face image;
Correlation storage means for storing the correlation;
A facial impression analysis device comprising: visual impression analysis means for analyzing a visual impression of the subject's face from the physical feature amount of the subject based on the correlation.   The face impression analysis apparatus according to claim 22, wherein the physical feature amount includes an amount related to at least one of luminance, saturation, hue, and frequency characteristics in the face image.   The face image acquisition means generates means for generating a three-dimensional model of the subject's face from stereo images of the subject's face taken by at least two cameras, and observes the rotation of the three-dimensional model or the three-dimensional model. 24. The face impression analyzing apparatus according to claim 22, further comprising means for changing an angle at which the face image is acquired by rotating a viewpoint to be performed.   25. The face according to claim 24, wherein the face image acquisition unit includes a pseudo light source setting unit that sets a pseudo light source to irradiate the three-dimensional model when the face image is acquired. Impression analysis device. A facial impression analysis device that analyzes a visual impression of a subject's face based on a correlation between a physical feature amount of the face and an appearance impression of the face seen from a predetermined angle,
Physical feature amount calculating means for calculating the physical feature amount of the face;
Correlation storage means for storing the correlation;
A facial impression analysis device comprising: visual impression analysis means for analyzing a visual impression of the subject's face from the physical feature amount of the subject based on the correlation.   27. The face impression analysis apparatus according to claim 26, wherein the physical feature amount includes a value related to the three-dimensional shape of the face.   The facial impression analysis apparatus according to any one of claims 22 to 27, wherein the impression of the face is an age impression.   The facial impression analysis apparatus according to any one of claims 22 to 28, wherein the correlation is different for each age group of the subject. The correlation storage means holds a correlation between the physical feature value and a visual evaluation value of an appearance impression of the face obtained by visual evaluation of the face;
The visual impression analysis means calculates a visual evaluation value of the visual impression of the subject from the physical characteristic amount of the subject based on the correlation between the physical characteristic amount and the visual evaluation value of the visual impression of the face. 30. The face impression analysis apparatus according to any one of claims 22 to 29, wherein: The correlation storage means holds a correlation between the physical feature quantity and a physiological quantity related to the skin;
The visual impression analysis means calculates a physiological quantity relating to the subject's skin from a physical feature quantity relating to the subject based on a correlation between the physical feature quantity and a physiological quantity relating to the skin. The face impression analysis device according to any one of claims 22 to 30.   32. The facial impression analysis apparatus according to claim 31, wherein the physiological amount is a physical property on the skin surface and / or an internal physical property of the face of the subject. A face impression analysis program for analyzing an appearance impression of a subject's face based on a correlation between a physical feature amount of the face that changes in an angle-dependent manner and an appearance impression of the face,
Computer
Means for extracting the physical feature quantity of the subject;
Based on the correlation, means for distinguishing the appearance impression of the subject from the physical feature amount about the subject;
A facial impression analysis program characterized by functioning as A face impression analysis program for analyzing the appearance impression of a subject's face based on the correlation between the physical feature amount of the face and the appearance impression of the face seen from a predetermined angle,
Computer
Means for extracting the physical feature quantity of the subject;
Based on the correlation, means for distinguishing the appearance impression of the subject from the physical feature amount about the subject;
A facial impression analysis program characterized by functioning as A face impression analysis device comprising: a face image acquisition means, a physical feature quantity extraction means, a correlation storage means, and an appearance impression analysis means for analyzing an appearance impression of the subject's face based on the face image of the subject A method for analyzing the appearance impression of a face using
The face image acquiring means acquiring the face image;
The physical feature amount extracting means extracts the physical feature amount of the face from the face image;
The appearance impression analyzing means is based on the correlation between the physical feature quantity of the face that changes in an angle-dependent manner held by the correlation storage means and the appearance impression of the face, based on the physical feature quantity of the subject. Analyzing the appearance impression of the face of the subject, and a method for analyzing the appearance impression of the face. The appearance of the face using a face impression analysis device comprising physical feature amount extraction means, correlation storage means, and appearance impression analysis means for analyzing the appearance impression of the subject's face based on the face image of the subject An impression analysis method,
The physical feature extraction means extracting the physical feature of the face;
The appearance impression analyzing means is based on the correlation between the physical feature quantity held by the correlation storage means and the appearance impression of the face seen from a predetermined angle, based on the physical feature quantity of the subject. Analyzing the appearance impression of the face of the subject, and a method for analyzing the appearance impression of the face. A method for distinguishing makeup effects by distinguishing the appearance of a face using physical features of the face,
A pre-makeup feature quantity measuring step of extracting the physical feature quantity of the face before makeup;
Based on the correlation between the physical feature amount and the visual evaluation value of the appearance of the face viewed from a predetermined angle, a pre-makeup visual evaluation value calculation step for calculating a pre-makeup visual evaluation value;
A post-makeup feature amount measuring step of extracting the physical feature amount of the face after makeup;
Based on the correlation, a post-makeup visual evaluation value calculating step of calculating a post-makeup visual evaluation value;
A cosmetic effect identification method comprising: comparing a visual evaluation value before makeup with a visual evaluation value after makeup and evaluating a makeup effect.