JP3614783B2 - Face classification - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a classification method for classifying a human face type according to features, a face classification method for extracting feature points from the viewpoint of the face, and classifying the face based on the feature points, and the face The present invention relates to a face map formed by coordinate axes based on a classification method. This face map can be used for image production by makeup, and is particularly suitable for beauty counseling to customers at cosmetics departments and beauty salons, or for education of makeup consultants and hairdressers.
[0002]
[Prior art]
Face makeup is usually done based on your own experience and senses, but when you make up other people's face apart from making up your own face, the target face is In addition, applying makeup according to each request requires considerable experience and special feeling, and it takes a considerable number of years to reach a certain level.
[0003]
Attempts to theoretically explain how to create an image with makeup have been made, but image production with conventional makeup is an application of modeling theory. It can be applied to make-up by grasping the impression. For example, as shown in FIG. 14, a straight line produces a sharp or static image, and a curve produces a soft or dynamic image. Therefore, a sharp image or a soft image can be produced by making up the contours of the eyebrows and mouth or the hair state linearly or curvedly, respectively. Similarly, it is possible to produce a calm image using a descending line, or to produce a robust image using a thick line.
[0004]
[Problems to be solved by the invention]
However, each individual face has its own characteristics. In modeling theory, only one part of the face can be applied, and by applying modeling theory alone, the makeup is disproportionate to each face. It may end up. Also, the makeup techniques used for individual faces are inherently different, but there is no way to show this systematically. There are many needs among many women who want to produce their desired image as a makeup, but there is no appropriate makeup method that can respond to this.
[0005]
The present invention has been made in view of the above problems, and comprehensively grasps the face, scientifically clarifies the factors that cause the face image, and based on the basis, controls the face form or the appearance of the form The purpose of this is to make it possible to perform image production using makeup accurately and easily. Analyzing individual faces and classifying or identifying them accurately It provides a possible face classification method .
[0006]
[Means for Solving the Problems]
The face classification method according to the present invention is a first method for indicating whether a face length and an arrangement state of facial morphological elements including eyes, eyebrows, mouths, and noses are close to the vertical center, the horizontal center, or are separated from each other. Based on the index, the step of positioning the face to be compared with an average face formed from the features of a plurality of samples, and the contour shape of the face and the contour shape of at least one of the face shape elements are curved And comparing the target face with the average face based on a second index indicating whether the target face is straight or straight.
Further, according to another aspect of the present invention, there is provided a first information indicating whether a face length and an arrangement state of facial form elements including eyes, eyebrows, mouths, and noses are close to a vertical center, a horizontal center, or are apart from each other. Based on the index, the step of positioning the face to be compared with an average face formed from the features of a plurality of samples, and the contour shape of the face and the contour shape of at least one of the face shape elements are curved A face classification method for making up a makeup image, comprising: positioning a target face in comparison with the average face based on a second index indicating whether the target is straight or straight .
Furthermore, another aspect of the present invention provides a first index indicating the length of the face and the arrangement of facial morphological elements including eyes, eyebrows, mouth, nose, one of which has a short face length or Indicates the degree to which the facial morphological element is away from the center of the left or right, or the facial morphological element is close to the vertical center, and the other is that the face is long or the facial morphological element is close to the horizontal center Or a first axis indicating the degree to which the facial morphological element is separated from the vertical center, and a second index indicating the facial contour shape and the contour shape of at least one of the facial morphological elements , One of which indicates the degree of the contour of the face and the shape of the facial morphological element being curvilinear, and the other is the second axis indicating the degree of the linearity of the face contour and the shape of the facial morphological element as a coordinate axis Forming a coordinate space, and an average face formed from the features of a plurality of samples. 1 metrics and the placing in the coordinate center based on the second index, or positioned in any of the coordinate position of the coordinate space based on the targeted face immediately to the first index and the second index And a step of classifying a face form as a target or a feature or an impression based on the coordinate position, and a face classification method for making a makeup image .
[0007]
According to this classification method, it is basically possible to classify systematically by analyzing faces using two types of indices.
[0008]
In addition, the first axis indicating the face length or the arrangement state of facial morphological elements including eyes, eyebrows, mouths and noses, and the facial contour shape and facial morphological elements including eyes, eyebrows, mouths and noses. By positioning the target face in a coordinate space consisting of two axes of the second axis indicating the contour shape of at least one of the morphological elements, the face can be accurately classified by form. This also makes it possible to easily grasp the feature points of the faces and the degree of similarity between the faces.
[0009]
In addition, the face map according to the present invention includes a balance axis indicating a face length or an arrangement state of facial morphological elements including eyes, eyebrows, mouths and noses, and a face contour shape or eyes, eyebrows, mouths and noses. A representative image display and / or a representative face image of the image is arranged in a coordinate space consisting of shape characteristic axes indicating the contour shape of at least one of the morphological elements of the face. is there. By using this face map, it is possible to easily understand what kind of feature points and images the target face has, and the desired image for makeup It is easy to know from the viewpoint of the form of the face what kind of makeup should be applied to produce the performance.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
[Facial taxonomy]
In order to construct an index for classifying faces according to features, firstly, in order to clarify the human face recognition mechanism and its face identification mechanism, the similarity of faces is used to Experimented.
[0011]
There were 12 test subjects, and 40 facial photographs of women in their 20s and 30s were used as samples.
[0012]
In the experiment, three faces A, B, and C are simultaneously displayed and presented on the monitor screen as shown in FIG. The subject chooses two more similar ones and enters the number from the keyboard.
[0013]
Forty photographs were divided into two groups, and all subjects evaluated both groups. ₂₀ C ₃ = 1140 assessments were repeated for all combinations per group, and the cumulative frequency (selection frequency) selected to be similar was recorded. For later analysis, the length, area, and angle of the face, such as the length, size, and eye lift, were measured on the screen. 2 to 4 show examples of measured items, FIG. 2 shows measurement items centered on the area of the face, and FIG. 3 shows the arrangement of facial morphological elements such as eyes, eyebrows, mouth, and nose. The measurement items centering on the relationship are shown, and FIG. 4 shows the measurement items related to the shape of each form element of the face such as the eyes, mouth, and eyebrows.
[0014]
For the above measured values, Z scores were standardized for each face so that the average value was 0 and the variance was 1, and this was taken as the degree of deviation from the average value. FIG. 5 shows the deviance of items in typical forms for samples S1 and S2 having particularly high similarity. In addition to the samples S1 and S2, FIG. 5 shows the maximum value (MAX) and the minimum value (MIN) of 40 samples for reference.
[0015]
As shown in FIG. 5, it can be seen that a face with a high degree of similarity has a high degree of similarity in terms of form, and in particular, the values of the eyes and the contour are approximated. It can also be seen that the value of the portion with a large deviation is also approximated.
[0016]
From the above analysis, it is understood that whether the deviation from the average of the facial shape is common and whether the contour and the eyes are similar are important for the determination of the similarity of the face.
[0017]
Next, the selection frequency was analyzed by multidimensional scaling. Multidimensional scaling is a method of placing an object in space according to the degree of similarity. This analysis reveals which index the subject used to determine the similarity, that is, the index that determined the person's face.
[0018]
A multidimensional analysis yielded a three-dimensional solution. That is, it became clear that the similarity of the face was judged mainly using three indicators.
[0019]
Three types of indices in each of the three dimensions were estimated by multiple regression analysis using the coordinate value of each face as a reference variable and the measurement value of each face form as an explanatory variable. As a result, in the first dimension, the face length is “short” to “long” and the arrangement of the eyebrows, nose and mouth is centered on the left and right or on the top and bottom. The second dimension is the shape of the eye that is round and curved or elongated and straight, and the third dimension is the shape of the eyebrow and mouth that is round and curved or elongated and straight. It was estimated to be the dimension shown.
[0020]
FIG. 6 shows an example in which a sample of a face having a high degree of similarity is placed in a different dimensional space. (A) is placed in the coordinate space of “face length, part placement” and “eye shape”. (B) is placed in the coordinate space of “face length, part arrangement” and “brow, mouth” shape, and (c) is “eye shape” and “eyebrow”. , Mouth "in the coordinate space.
[0021]
As can be seen from these examples, it can be seen that in three types of coordinate spaces, groups having high similarity are arranged in a predetermined group (cluster) in any coordinate space. For example, S873, S882, and S777 always form a group, and S755, S766, S770, and S773 form the same group at any coordinates. In this way, it can be seen from the results that “face length, part placement”, “eye shape”, and “eyebrow and mouth shape” are indicators of similarity judgment, supporting the above experimental results. It became a thing.
[0022]
When judging the similarity of the face from the above experiments and analysis, "Whether the length of the face or the arrangement of the parts is close to the top and bottom center, away from each other, or close to the left and right center or away" It was found that “the shape of the parts” played a big role, and by using these as indices, it was found that the face could be systematically classified according to the features of the form.
[0023]
FIG. 7 shows a coordinate space having these indices as coordinate axes. The vertical axis indicates “face length and the arrangement state of parts (brows, eyes, nose, mouth)”, and the horizontal axis indicates “parts (brows, eyes). , Nose, mouth) "as the coordinate axes.
[0024]
This vertical axis shows that the face length is “short” to “long”, the part is “distant from the center of the left / right” to “centered at the center of the left / right”, or “the part is close to the center of the top / bottom” ”To“ away from the vertical center ”. This is consistent with the changes in the face that accompany the growth from children to adults. The downward direction is “childhood”, the upward direction is “adultness”, in other words, It can be seen as a change in the balance of the face when growing from a child to an adult, and here it is named the “balance axis”.
[0025]
In addition, the horizontal axis represents the shape characteristic of whether the shape of the part is curvilinear or linear, and is named “form axis”.
[0026]
When the relationship between the face form and impression when classified by the above index was investigated by multiple regression analysis, the characteristics of each quadrant of the coordinates in FIG. 7 could be clarified. FIG. 8 shows the form and impression located in each quadrant of the coordinate space of FIG. For example, if the face is long and the parts are curved (quad II), the gentle feminine impression is strengthened, or if the face is short and the shape is straight, it gives a fresh and lively impression. And the relationship with the impression became clear.
[0027]
As described above, the face is classified according to the form by classifying the face by using the length of the face and the arrangement state of the facial morphological elements of the eyes, eyebrows, mouth, and nose, and the shape of the facial morphological element as an index. It is possible to classify according to the impression received from the shape of the face. And by classifying in this way, it is possible to produce an image accurately when performing make-up as described later, for example. Further, by assigning a classification according to the present invention to a customer list of a beauty salon or the like, it can be used as a guideline for creating an image suitable for the customer.
[Face Map]
Although the causal relationship between facial form and impression was clarified by the above-mentioned experiment and analysis, the face created based on the knowledge obtained in the process of obtaining the above-mentioned face classification method is shown in FIG. An example of the Tochi map is shown.
[0028]
As with the coordinates shown in FIGS. 7 and 8, the face map 1 has a horizontal axis indicating the balance axis Y indicating “the arrangement of the shape elements of the face length and eyes, eyebrows, mouth, and nose” on the vertical axis. It consists of a two-axis coordinate space 2 with a form axis X indicating the shape of the facial form element on the axis. A female “average face” 3 is arranged in the center (origin position) of the coordinate space 2. Further, images of the face 4 with a long face and the face 5 with a short face are arranged at the top and bottom ends of the balance axis Y on the vertical axis, respectively. At both ends of the balance axis Y, explanations are added such as the length of the face is long, short, the arrangement of parts is centered, and they are separated. In addition, at both ends of the horizontal form axis X, there is an explanation that the shape of the lower half of the face outline and the shape of the eyes, nose, mouth, and eyebrows are curvilinear or linear. .
[0029]
In the first to fourth quadrants of the coordinate space 1 formed by the balance axis Y and the form axis X, explanations of typical images or impressions of the forms located in the respective coordinate spaces 1 are added. For example, in the first quadrant, it is cool, sharp, dynamic, in the second quadrant, friendly and feminine, in the third quadrant, cute and cute, in the fourth quadrant, fresh and active. It is.
[0030]
Further, representative face images 6 to 9 exaggerating the characteristics of each quadrant are arranged in each quadrant in order to facilitate the determination of the face. Note that the average face 3 at the center and the representative face images 6 to 9 arranged in each quadrant are actually created by morphing, which is a computer graphic technique. For example, the average face 3 in the center is a composite of 40 female face photographs, and has values as shown in FIG. Each representative face image can be produced by exaggerating each feature based on the average face and transforming it with a morphing technique.
[0031]
Next, an example in which the face map 1 shown in FIG. 9 is used as a guideline for makeup will be described with reference to the flowchart of FIG. 11 and FIG.
[0032]
First, for the face to be made up, whether the face length and parts are close to the top / bottom center, away from the balance axis Y of the face map 1, or left / right center, It is determined whether they are separated (step 1; S1). In this case, the balance axis Y first checks whether the face length is long or short. The length of the face is determined by the ratio of “width of face” and “length of chin to forehead”. For the average face, this ratio is about 1: 1.4.
[0033]
If it is difficult to judge by the length of the face, whether the placement of facial morphological elements (parts) such as eyes, nose, mouth, etc. is centered on the left and right, separated, or centered on the top and bottom Identify by. For example, if the distance between the eyes is narrow or the cheeks below the eyes are long, it is determined that the balance is adult. Further, if the eyes are separated from each other, or if the eyes and mouth stick together, it is determined that the balance is childish.
[0034]
Next, determine the shape of the part with the form axis X. In this case, if the lower half of the contour of the face is plump or if the eyes, nose, mouth, eyebrows, etc. are rounded, it is determined that the shape is curved. Further, when the lower half of the contour is square or sharp, or the shape of the eyes, nose, mouth, or eyebrows is linear (sharp), it is determined that the shape is linear.
[0035]
Now, as a result of the analysis of the face, it is determined that the balance is adult-like and the shape gives a curved impression. It is assumed that the face map 1 is positioned at the position shown in the quadrant II as shown in FIG.
[0036]
Therefore, next, a request for makeup is heard and grasped (step 2: S2).
[0037]
Suppose that the make-up request is, for example, that the image is “cool and sharp”.
[0038]
The cool and sharp image is located in the first quadrant of the face map, and the direction from the position of the second quadrant of the current face is confirmed on the face map. The method is found with reference to the face map display (step 3; S3). In this case, as shown in FIG. 12B, it can be understood that the position of the form axis may be moved in a linear direction while keeping the balance axis as it is.
[0039]
Next, makeup is applied based on the image production method found in step 3 (step 4). In this case, make use of the part where the balance is adult-like, and make up the parts in a straight direction along the form axis. Specifically, the shape of the eyebrows and the outline of the lips are expressed linearly. As a result, it is possible to produce a makeup with a cool and sharp image upon request.
[0040]
FIG. 12 (c) shows an example of a case where the make-up request is to give a “fresh and active” impression. In this example, the “fresh and active” image is in the form of being located in the IV quadrant on the face map, and is in a counter-polar relationship from the position of the real face in the II quadrant. Requires considerable technique.
[0041]
The balance axis should be childish. In addition, the position of Mt. Ebisu is slightly lowered, and blusher is put in the center part to make the face length inconspicuous. In addition, the outline of the eyebrows and lips is drawn linearly in order to make the form axis linear. This makes it possible to produce a slightly fresh and active image from an adult-friendly image.
[0042]
As described above, when using the face map composed of the two axes of the balance axis and the form axis shown in the present embodiment, first, to which position the target face belongs in the form and what kind of impression You can easily grasp whether you are giving. Furthermore, by determining the degree of similarity with a representative face image, the coordinates of the target face in the coordinate space can be easily predicted. Then, it is possible to know what kind of makeup should be applied to produce the produced image and the shortcut for the image production.
[0043]
FIG. 13 shows another embodiment of the face map. The face map 10 is located in a coordinate space composed of two axes: a balance axis indicating whether the type is an adult type or a child type on the vertical axis and a form axis indicating whether the shape of the part on the horizontal axis is curved or linear. It shows a lot of impressions of the face. In this way, by displaying a large number of relations between the face and the impression on the above-mentioned two-axis coordinates, it is possible to use for the production of a finer image.
[0044]
In each of the above examples, the face map has a balance axis indicating whether it is an adult type or a child type on the vertical axis, and a form axis indicating the shape of the part is on the horizontal axis. The horizontal axis can be changed. Further, the content to be displayed and the face image may be appropriately arranged as necessary.
[0045]
By using the face map according to the present invention, facial features can be easily identified, and this face map can be used as a guideline for makeup. For example, it can be used as a highly useful auxiliary tool in cosmetics counseling for customers in a cosmetics department or a beauty salon, or in the education of a makeup consultant or a hairdresser or in actual business.
[0046]
【The invention's effect】
As described above, according to the face classification method of the present invention, the length of the face or the index relating to the arrangement of the facial morphological elements such as eyes, mouth and nose and the face and the face such as eyes, eyebrows, mouth and nose. By classifying the face according to the index related to the shape of the morphological element, it is possible to classify or identify the features of the face by system. And, as a result of clarifying the causal relationship between facial forms and impressions, it is possible to associate in advance the faces classified by this classification method with the impressions that each classification has, and are classified according to the shape of the face It is possible to easily determine what kind of impression the face has.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating an example of a screen in an experiment of a face identification mechanism.
FIG. 2 is a diagram illustrating measurement items related to the area of a face.
FIG. 3 is a diagram showing measurement items related to the arrangement state of facial morphological elements.
FIG. 4 is a diagram showing measurement items related to the shape of a facial morphological element.
FIG. 5 is a diagram illustrating an example of a morphological deviation degree of a similar face.
FIG. 6 is a diagram illustrating an example of arrangement of coordinates and face images used in the multidimensional scaling method.
FIG. 7 is a diagram showing coordinates composed of two axes, a balance axis and a form axis.
FIG. 8 is a diagram showing an impression of each quadrant of the coordinate space.
FIG. 9 is a diagram showing an example of a face map according to the present invention.
FIG. 10 is a diagram showing the shape of an average face.
FIG. 11 is a flowchart for performing makeup using a face map.
FIG. 12 is a diagram illustrating an example of an image effect when makeup is performed using a face map.
FIG. 13 is a diagram showing another example of a face map.
FIG. 14 is a diagram illustrating an example of a makeup effect based on a conventional modeling theory.
[Explanation of symbols]
1, 10 Face map 2 Coordinate space 3 Average face 4-9 Image representative face

Claims (3)

Based on the first index indicating the length of the face and the arrangement state of the facial morphological elements including the eyes, eyebrows, mouths, and noses are close to the vertical center, the horizontal center, or separated from each other, a plurality of target faces Positioning relative to the average face formed from the features of
Based on a second index indicating whether the contour shape of the face and the contour shape of at least one of the morphological elements of the face are curvilinear or linear, the target face is the average face The stage of positioning compared to
Face classification including Based on the first index indicating the length of the face and the arrangement state of the facial morphological elements including the eyes, eyebrows, mouths, and noses are close to the vertical center, the horizontal center, or separated from each other, a plurality of target faces Positioning relative to the average face formed from the features of
Based on a second index indicating whether the contour shape of the face and the contour shape of at least one of the morphological elements of the face are curvilinear or linear, the target face is the average face The stage of positioning compared to
A face classification method for the production of makeup images. The length and the eyes of the face, eyebrows, mouth, indicates a first index indicating the arrangement of the form elements of the face including the nose, one is shorter in length or facial form elements face away from the left and right center The length of the face is long, the face is long or the face is close to the center, or the face is close to the top and bottom. A first axis indicating the degree of separation from ,
Shows the second index indicating the outline shape of at least one form element of the contour shape and the face of the form elements of the face, one the shape of the contour and the face of the form element of the face indicates the degree is curved The other is to form a coordinate space with the second axis indicating the degree of linearity of the face contour and the shape of the facial morphological element as coordinate axes;
Placing an average face formed from a plurality of sample features in the center of coordinates based on the first and second indices;
Grasping at which coordinate position in the coordinate space the target face is positioned based on the first index and the second index;
Classifying the target face form according to features or impressions based on the coordinate positions;
A face classification method for creating a makeup image.

Images