JP2009082338A - Skin discrimination method using entropy - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for easily and accurately discriminating a skin by finding skin property values usable for the discrimination of the skin and its treatment means and to allow anyone to easily, objectively and quantitatively discriminate the skin by providing a method for evaluating the beauty of the skin especially, and a device and a program for executing it. <P>SOLUTION: This skin discrimination method is characterized in evaluating the beauty of the skin using an entropy of the distribution of the skin property values as an index. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for discriminating skin, and more particularly, to a technique for discriminating skin using processed values of skin property values.

  Recognizing that the skin is beautiful by a third party is one of the great wishes of many people as well as women. For this reason, research and development of cosmetics and beauty methods for showing beautiful skin are being actively conducted. However, the skin condition varies greatly depending on the individual, and also changes depending on the aging and living environment. Therefore, in order to appropriately select the type of cosmetics, makeup method, skin care method, etc. It is necessary to objectively judge how the skin to be seen looks to a third party. For example, in a cosmetic department store at a department store, a pharmacy, or a cosmetic store, a simple method for evaluating the degree of skin beauty of a subject is required.

  Various researches have been conducted on the elements that make up the visual beauty of the skin, and methods have been developed to individually evaluate some of these elements. For example, there is a method for evaluating skin properties using a technique for measuring physical quantities such as skin conductance, skin water dispersion loss, sebum amount, skin flexibility, and rate of turnover of the stratum corneum. In recent years, a method for quantitatively evaluating the skin surface shape and optical properties by processing image information obtained by imaging the skin surface and its replica with an appropriate photoelectric conversion means by a program has been reported. Yes.

  However, when a third party looks at the skin, the state of the skin that is visually recognized, that is, the beauty of the visual skin, is formed by a combination of these various elements. For this reason, it is not easy to make a determination from the determination results of the individual elements as described above. Actually, in order to determine the beauty of the visual skin, an expert in skin evaluation analyzes the measurement results based on individual knowledge based on individual measurement data, and the evaluator makes a pair comparison. However, in this case, it is necessary to have a skin evaluation specialist or a certain number of evaluators, and to analyze the collected evaluation data. It was. As described above, it has been difficult to accurately and easily evaluate the visual beauty of the skin with the conventional method.

  In such a technical background, by measuring and processing specific skin property values obtained from skin photos and replicas, we will find a correlation between this value and the skin beauty, and evaluate the skin beauty. Attempts have been made. For example, a method of discriminating the beauty of the skin by applying visible light to the skin replica from two directions and using the reflectance at each wavelength of the reflected light spectrum of the skin as an index (Patent Document 1), the thickness of the subcutaneous fat near the face line To distinguish the beauty of the face line by using the magnitude of the restoring force against the deformation caused by the external force of the skin and the skin as an index (Patent Document 2), particle analysis of the high-luminance part of the two-dimensional image in which the fine light-dark distribution on the skin surface is emphasized A method of quantifying the optical beauty of the skin surface using the correlation between the value and the sensory evaluation of the appearance of the skin (Patent Document 3), and visible light from two directions on the made-up skin A method for distinguishing skin using a difference in spectral spectrum of reflected light obtained by irradiation (Patent Document 4), and high-frequency composition of specular reflection light components included in a digital image of skin acquired under polarized illumination. And a method of evaluating the dispersion value and utilizing a correlation between sensory evaluation beauty skin beauty skin (Patent Document 5) are disclosed. In such a technique, the relationship between measured values and processed values and the beauty of the skin is partially recognized, but the correlation is not very good. As described above, research and elucidation about the numerical value indicating the skin properties that have a high correlation with the beauty of the skin is not yet sufficient, and it is used even more as an index to evaluate the beauty of the skin. There was a search for valuable figures.

  In such a background, a technique for estimating skin age using a fractal dimension of a skin image has been developed (Patent Document 6). The present inventors have also conducted research on a technique for fractal analysis of skin information, and have found that there is a certain relationship between the fractal dimension of skin information and the visual evaluation value of the beauty of skin (Non-Patent Document 1). ).

  On the other hand, it is not known at all that the entropy (average information amount) of the skin property value is used as an index for evaluating the skin.

JP 2003-161656 A Japanese Patent Laid-Open No. 11-164822 Japanese Unexamined Patent Publication No. 7-231883 Japanese Patent Laid-Open No. 10-2798 JP-A-2005-429 JP 2007-050158 A Keihei Kurita, “Fractal Analysis of Skin Information”, The Institute of Electronics, Information and Communication Engineers, IEICE Technical Report, pp. 39-44, 2006

  An object of the present invention is to find a skin property value that can be used for skin discrimination and a processed value thereof, and to provide a technique for easily and accurately discriminating skin. In particular, it is an object of the present invention to provide a method for evaluating the beauty of skin, a device for executing the method, and a program, and to make it possible to easily and objectively and quantitatively differentiate skin by anyone. To do.

  As a result of repeated studies on the skin beauty, the present inventors have found a causal relationship between the visual evaluation value of the skin beauty and the entropy of the distribution of various skin property values. Using this relationship, the inventors have found that the beauty of the skin can be evaluated using the entropy of the distribution of the skin property value of the subject as an index, and have completed the invention. That is, the present invention is as follows.

(1) A skin discrimination method characterized by evaluating the beauty of skin using the entropy of the distribution of skin property values as an index.
(2) The skin discrimination method according to (1), wherein the skin property value is a color system signal of an image on the skin surface.
(3) The skin discrimination method according to (2), wherein the color system signal is an RGB value, a YUV value, or a Munsell (HVC) value.
(4) A step of acquiring at least one color system signal of the skin surface image, a step of calculating entropy for a distribution of values of at least one component constituting the color system signal, and a preparation in advance Substituting the calculated entropy into a regression equation representing the relationship between the entropy of the distribution of the component values and the visual evaluation value of the beauty of the skin, and calculating the visual evaluation value of the beauty of the skin, (2) or the discrimination method of the skin as described in (3).
(5) The regression equation obtained by multiple regression analysis of the entropy of the distribution of each component value constituting the RGB value, YUV value or Munsell (HVC) value and the visual evaluation value of the beauty of the skin The skin discrimination method according to (4), characterized in that:

(6) The skin discrimination method according to (1), wherein the skin property value is a undulation value of the skin surface.
(7) Represents the relationship between the step of obtaining the undulation value of the skin surface, the step of calculating the entropy of the distribution of the undulation value, and the visual evaluation value of the undulation value distribution prepared in advance and the beauty of the skin And substituting the calculated entropy into a regression equation to calculate a visual evaluation value of the beauty of the skin.
(8) Means for acquiring at least one color system signal of the skin surface image, means for calculating entropy for the distribution of values of at least one component constituting the color system signal, and prepared in advance Means for substituting the calculated entropy into a regression equation representing the relationship between the entropy of the distribution of the component values and the visual evaluation value of the skin beauty, and calculating the visual evaluation value of the skin beauty, A skin discrimination apparatus including means for displaying a visual evaluation value.
(9) Means for obtaining the undulation value of the skin, means for calculating the entropy of the distribution of the undulation value, and regression representing the relationship between the entropy of the distribution of the undulation value prepared in advance and the visual evaluation value of the beauty of the skin A skin discrimination device including means for substituting the calculated entropy into a formula to calculate a visual evaluation value of the beauty of skin, and means for displaying the calculated visual evaluation value.
(10) Means for calculating entropy for the distribution of the values of at least one component constituting the color system signal of the skin surface image, the entropy of the distribution of the component values prepared in advance and the beauty of the skin A skin discrimination program for substituting the calculated entropy into a regression equation representing the relationship between the visual evaluation values of the skin and functioning as a means for calculating a visual evaluation value of the beauty of the skin.
(11) The computer calculates the entropy of the calculated entropy by means of calculating the entropy of the distribution of the undulation value of the skin, the regression equation expressing the relationship between the entropy of the distribution of the undulation value prepared in advance and the visual evaluation value of the beauty of the skin. Is a skin discrimination program for substituting and functioning as a means for calculating a visual evaluation value of the beauty of the skin.

  By using the method and apparatus of the present invention, it is possible to easily, objectively and quantitatively distinguish skin using various skin property values.

In the present invention, “beauty of skin” refers to the degree of preference of the skin state when the properties of the skin that can be objectively recognized by a third party are combined, and the fineness of the texture. It is the degree of skin condition preference when combining properties such as uniformity in texture direction, smoothness, unevenness, smoothness, moistness, softness, wrinkles, elasticity, and glossiness. On the other hand, it does not include subjective beauty that can be obtained by processing and processing objectively recognized information through mental activity.
The “visual evaluation value of the beauty of the skin” is a statistical evaluation value indicating how beautiful a person looks when the skin is visually recognized. This is a statistical evaluation value obtained by repeatedly determining which one looks more beautiful when compared with other skin.

  In the skin discrimination method of the present invention, the skin property value for obtaining an index used for the evaluation of the beauty of the skin preferably includes a color system signal of the skin surface image and a undulation value of the skin surface.

Examples of the color system signal include RGB values, YUV values, Munsell (HVC) values, L ^* a ^* b ^* values, L ^* C ^* h ^* values, Lab values, and Yxy values. Among these, it is particularly preferable to use RGB values, YUV values, and Munsell (HVC) values. The RGB value represents a color by a combination of the three primary colors of red (R), green (G), and blue (B). For example, in the case of representing each with 256 gradations, about 16.77 million. Color tone expression is possible. The YUV value represents a color by a combination of luminance (Y), color difference (U = blue-Y), and color difference (V = red-Y). The Munsell (HVC) value is a JIS color system that expresses colors with three components of hue, lightness, and saturation. At least one component constituting such a color system signal may be a property value, or a plurality of components may be property values.

  The undulation value is a numerical value indicating how high a point covering the surface of a certain object is from the reference plane.

In order to acquire at least one color system signal of the color system image of the skin surface, first, the subject's skin surface is imaged. The part of the skin from which the image is acquired is not particularly limited as long as it is a part where the skin is to be distinguished, and examples thereof include facial skin such as cheeks and the upper arm inner part. For example, when the discrimination result is used in selection of cosmetics such as foundation and teak, it is preferable to image the cheek. In general, it is preferable to select a region that represents the skin of the subject and avoid a site with many scratches or spots. Moreover, the range for acquiring the image is not particularly limited as long as the information necessary for the analysis can be obtained, but the range of 1 cm * 1 cm to 3 cm * 3 cm is preferable on the skin surface.

The means for acquiring at least one color system signal of the skin surface image is not particularly limited. For example, it can be performed using devices such as a color digital microscope, a color digital camera, a color video camera, and a scanner. Such equipment may be commercially available or manufactured. Preferable examples of commercially available products include i-scope and CCD microscope manufactured by Moritex Co., Ltd., USB video microscope manufactured by Fortesimo Co., Ltd., and digital microscope manufactured by Keyence Corporation.
The imaging magnification may be a magnification suitable for an apparatus used for imaging. For example, when using a digital camera equipped with a macro lens, it is preferable to obtain a close-up photographed image at a distance of about 20 cm from the subject at the same magnification, and when using a video microscope (for example, i-scope manufactured by Moritex), It is preferable to enlarge the skin about 30 to 50 times and take an image. Thus, the amount of information when acquiring an image may be 64 * 64 pixels (dots, pixels) or more when converted into a range of 2 cm * 2 cm, preferably 128 pixels or more, and more preferably Is preferably 300 * 300 pixels or more.

  The colorimetric signal thus obtained is preferably transferred to a computer and then noise is removed using a median filter or smoothed using a smoothing filter. It is preferable to By performing the smoothing process, a large variation in the property value for each pixel can be corrected, and entropy can be calculated with higher accuracy.

  Image capturing and smoothing can be performed using commercially available image analysis software. For example, WinROOF (registered trademark) of Mitani Shoji Co., Ltd., Adobe Photoshop (registered trademark) of Adobe Systems, NanoHunter NS2K-Pro (registered trademark) of Nano Systems Co., Ltd. can be exemplified. Also, smoothing software published on the Internet can be used. The mask size is preferably 3 * 3, 5 * 5, for example.

Further, the specific image signal can be converted into any other image signal according to a conventional method. For example, YUV values, L ^* a ^* b ^* values, L ^* C ^* h ^* values, Lab values, and Yxy values can be converted from RGB values using a conversion formula. A conversion table can be used for conversion to Munsell (HVC) values. For example, when converting from an RGB value to a YUV value, after performing γ correction of the RGB value using commercially available software or the like, it can be converted into a YUV value using, for example, the following equation (1). Also, when converting RGB values to Munsell values, RGB values are converted to XYZ values using commercially available software or software published on the Internet, and then converted to Munsell color system HCV. it can.

  For at least one of the components constituting the color system signal thus obtained, the entropy of the distribution of the component on the image is calculated. The entropy calculation method will be described later.

  On the other hand, as a method for acquiring the undulation value of the skin surface, there is a method using a undulation value obtained by acquiring a replica of the skin of the subject and measuring the surface shape of the replica. The part of the skin from which the replica is acquired is not particularly limited as long as it is a part for which the skin is to be identified, and examples thereof include facial skin such as cheeks and the upper arm inner part. When the evaluation value is used for selecting cosmetics, it is preferable to obtain a cheek replica. For example, a 2 cm * 2 cm measurement region on the cheek may be set, and a replica of the portion including the measurement region may be acquired. The replica agent is not particularly limited, and examples thereof include silicon ASB-01-WW manufactured by Asahi Biomed Co., Ltd. The method of collecting a replica can be performed by a conventional method used for diagnosing skin shape and the like. For example, the replica agent may be applied to the skin placed for about 20 minutes at 20 ° C. and 50% humidity after washing the face, and then collected.

Next, the undulation value of the replica thus produced is measured. The method for measuring the undulation value is not particularly limited, and a normal method can be used. For example, “Wrinkle Evaluation Method Guidance”, Journal of the Japan Cosmetic Science Society, Vol. 2 (2004).
Specifically, for example, using a commercially available laser type three-dimensional surface roughness meter, laser scanning is performed in the horizontal direction (x) and the vertical direction (y) with respect to the face portion as shown in FIG. You can go and measure. As such a three-dimensional roughness meter, for example, a high-precision three-dimensional image processing apparatus LIP (for example, LIP-50) of Science Systems Co., Ltd., SURFCOM of Tokyo Seimitsu Co., Ltd., VLH, PRIMOS of Lasertec Co., Ltd. (GFM Co., Ltd.) And Derma-TOP-blue (manufactured by Breuckmann).
The interval between scans when measuring the undulation value using these devices is not particularly limited as long as sufficient data can be obtained for calculating entropy, but it is preferably performed at intervals of 10 μm or less.
For example, when scanning using the LIP-50, 1000 scans can be performed at 10 μm intervals in the X or Y direction with respect to the replica region where X * Y is 1 cm * 1 cm.

  When obtaining the undulation values thus obtained, if the sampling periods in the x direction and the y direction are different, it is preferable to correct the sampling period using a Sinc function (see FIG. 2).

  In addition, by applying oblique illumination to the replica using an optical projection device, etc., extracting the shadow part of the replica convex part, and measuring the depth, area ratio, etc. of the concave part of the skin from its area and width, etc. You can also get a value. Acquisition of the undulation value by irradiation with such oblique illumination is preferable in terms of simplicity. The acquisition of the undulation value by this method may be performed using, for example, a 3D replica analysis system for reflection (Asahi Biomed).

  Furthermore, the semi-transparent replica can be irradiated with light, the thickness of the replica can be obtained from the transmitted light quantity, and the undulation value of the replica can be obtained (semi-transparent replica light transmission method). The acquisition of the undulation value by this method may be performed using, for example, the 3D skin analysis system ASA-03 (Asahi Biomed).

  Further, the relief value may be obtained directly from the skin. As such a method, for example, a method in which lattice-like light is applied to the skin and the refractive index of the light is converted into an undulating value can be used, and a commercially available device or the like can be used. If a device such as PRIMOS (manufactured by GFM) or Derma-TOP-blue (manufactured by Breuckmann) is used, the undulation value can be obtained directly from the skin as well as from the replica.

  The entropy of the distribution on the measurement region of the undulation value of the skin surface acquired in this way is calculated.

“Entropy” is also called average amount of information and is well known as a concept of information theory. Entropy, like the thermodynamic concept, is a measure of randomness, irregularity, uncertainty, etc. In other words, it is a measure of how much information a certain source has.
The entropy of the probability distribution P in the probability space is an average value of the selected information amount of the event i−logP (i), and the entropy of the probability distribution P when the event i can take 0 to max is expressed by the following formula (2 ). The sum of the probabilities P (i) for each event i must be one.

For example, when a color system signal of the skin surface image is used as the skin property value, max is determined according to the gradation of each component of the image signal. For example, when a color system signal having 256 gradations is used, each of the pixels forming the image has any information of 0 to 255 gradations. Is 255.
Further, when the skin relief value is used as the skin property value, max is determined according to the maximum value of the skin relief value to be measured. At this time, the minimum value of the undulation value of the skin to be measured is set to zero.

  The skin beauty is evaluated using the entropy of the distribution of the property values of the subject's skin thus obtained as an index. The evaluation may be a relative evaluation or an absolute evaluation. For example, the beauty of the skin of the subject can be evaluated by comparing the entropy of the distribution of the property value of the skin of the subject with that of other skin. In this case, it can be evaluated that the smaller the entropy is, the more beautiful the skin is. In addition, the evaluation result can be statistically processed to create an evaluation standard, and the skin beauty of the subject can be evaluated according to the evaluation standard.

  For example, a method of calculating the visual evaluation value of the skin beauty using a regression equation that represents the relationship between the entropy of the distribution of the skin property value and the visual evaluation value of the skin beauty is specifically mentioned. According to such a method, statistical evaluation such as ranking can be easily performed, and skin discrimination can be performed with higher accuracy.

  The regression equation representing the relationship between the entropy of the distribution of the skin property value and the visual evaluation value of the beauty of the skin can be created by the following method, but is not limited to this method.

  1) At least one skin selected from a skin color image signal of a skin surface image and a undulation value of the skin surface from skin with sufficiently distributed skin condition and age (hereinafter collectively referred to as a sample). The property value is acquired, and the entropy of the distribution of the acquired property value is calculated. Property value acquisition and entropy calculation can be performed in the same manner as described above. The number of samples used at this time is 30 or more, preferably 50 or more.

  2) Next, prepare an evaluator suitable for representing a third party, present the sample, and evaluate the visual beauty of the skin. These evaluations may be absolute evaluations such as scoring. However, in order to ensure objectivity, another sample (for example, a reference photograph in which the beauty of the skin is gradually changed, see FIG. 3). Relative evaluation such as ranking in comparison with is preferable. In the ranking, if there is no difference, the ranking may be the same. At this time, in order to further ensure objectivity, it will be described that the evaluation is performed only on the visible skin elements. Here, an evaluator suitable for representing a third party may be any person who can understand at least the meaning of visual skin beauty, regardless of age or gender. The number of evaluators is usually 4 or more, preferably 10 or more.

  3) It is preferable to repeat the operation of 2). What is necessary is just to adjust the frequency | count of work suitably by the number of evaluators. In order to obtain an objective evaluation result, the evaluation is usually repeated 3 times or more, preferably 4 times or more, more preferably 5 times or more.

  4) Next, a visual evaluation value of the beauty of the skin is calculated for each sample. Here, as the visual evaluation value, the obtained score itself may be used, or when the relative evaluation is performed by ranking, the ranking itself may be used, or a numerical value may be given in ascending order with beautiful skin. Good. In addition, these evaluation values may be a total for each sample or an average value. For example, when ranking is performed for n samples, the average value of the total score of each sample can be obtained by setting the i-th score to n−i + 1 and can be used as the evaluation value. Further, the deviation value of each sample can be obtained from the average value and the standard deviation of the samples and used as evaluation values, or these values can be divided into arbitrary stages to obtain evaluation values.

  5) Regression analysis is performed on the average of at least one property value obtained in 1) and the evaluation value obtained in 4) to obtain a regression equation (prediction equation). At this time, when colorimetric signals such as RGB values, YUV values, Munsell (HVC) values, etc. are used as property values, the visual evaluation values of the components constituting the colorimetric signal and the beauty of the skin It is preferable to obtain a multiple regression equation by performing multiple regression analysis on the above because a higher correlation can be obtained. Such regression analysis can be performed by a conventional method, for example, using commercially available statistical processing software.

  The entropy of the distribution of the property value of the subject calculated above is substituted into a regression equation representing the correlation between the same property value and the visual evaluation value of the beauty of the skin, and the visual evaluation value of the skin beauty of the subject is calculated. Can be calculated. The visual evaluation value of the beauty of the skin obtained in this way can be displayed as a numerical value as it is, but by processing it into easy-to-use data such as deviation values and predefined ranks, counseling and advice This is preferable because it is easy to use in the scene. For example, the minimum and maximum visual evaluation values of the sample used in the creation of the regression equation used to calculate the visual evaluation value of the beauty of the skin are equally divided into a plurality of arbitrary ranks. Can be displayed with alphabets and numbers, or with words indicating the degree of beauty of the skin.

  In addition, the entropy of the subject is shown on the regression line drawn by the regression equation used, the position and rank in the sample group are shown in the figure, or as a photograph-like image (see FIG. 3) or stereogram It can also be displayed. The display method of the figure is not particularly limited, and can be displayed on, for example, a display of a device or a print medium.

  Further, the skin discrimination apparatus of the present invention is a means for acquiring at least one color system signal of an image on the skin surface, and the entropy of the distribution on the image for at least one component constituting the color system signal. Substituting the calculated entropy into a regression equation representing the relationship between the entropy of the distribution of the component values prepared in advance and the visual evaluation value of the beauty of the skin, and calculating the visual evaluation value of the beauty of the skin Means for calculating and means for displaying the calculated visual evaluation value;

The skin discrimination device of the present invention can be configured as follows, for example. The following configuration is an example, and the present invention is not limited to the configuration of the embodiment.
FIG. 4 is a hardware block diagram of a skin discrimination device using color system image signals obtained from the skin surface and entropy of undulation values. As shown in FIG. 4, the evaluation apparatus 1 includes an input unit 1, a CPU (Central Processing Unit) 2, a ROM (Read Only Memory) 3, a RAM (Random Access Memory) 4, a magnetic disk device 5, a recording unit 6, and an operation. Part 7 and display part 8. These are connected to each other via a bus. The input unit 1 is a device such as a color digital microscope, a color digital camera, a color video camera, a scanner or the like for inputting at least one color system signal of a skin surface image, or a skin such as a three-dimensional roughness meter. It is a device that measures the surface relief.

The input unit 1 may include either one of a unit that acquires an image signal and a device that acquires a undulation value, or may include both. The CPU 2 executes processing such as data processing such as the above-described smoothing processing, calculation of entropy by a box counting method, calculation of visual evaluation value using a regression equation, and the like according to a program stored in the ROM 3. The ROM 3 stores programs necessary for the functioning of the evaluation apparatus of the present invention, various regression equations necessary for visual evaluation, and the like. The RAM 4 is an OS (Operating) to be executed by the CPU 2.
A part of the (System) program and application program are temporarily stored. The magnetic disk device 5 is used as an external storage of the RAM 4 and has a recording unit 6. The operation unit 7 is operated when inputting necessary data such as predetermined commands and regression equations. The display unit 8 only needs to be able to display a visual evaluation value. For example, a display device such as a CRT (Cathode Ray Tube), a liquid crystal display, or a plasma display, an audio output device such as a speaker, an output from a printer, or the like. Apparatus.

  Further, the present invention may be a program for causing a computer, another device, a machine, or the like to execute part or all of the processing. Further, the present invention may be a program in which such a program is recorded on a computer-readable recording medium.

  Hereinafter, the present invention will be described in detail with reference to examples, but the scope of the present invention is not limited thereby.

<Calculation of entropy of value distribution of each component of RGB value>
Using the image samples of the cheeks of 248 women in their 10s and 50s, the entropy of the distribution of each component of the RGB values and the visual evaluation value of the beauty of the skin (hereinafter also simply referred to as “visual evaluation value”). ). That is, for the 248 persons, 30 minutes after washing the face, the part shown in FIG. 1 was photographed using a commercially available digital camera (Nicon D100 60mm Macro lens), and the central area of the photographing area was 15 * 15 mm (size 300 *). (300 pixels) was extracted, and this part was smoothed with a 3 * 3 mask (all grid values were 1). For each component of the RGB values of each sample, entropy was calculated using a program that executes the calculation of Equation (2) described above.

<Calculation of visual evaluation value (visual evaluation by panel)>
On the other hand, a visual evaluation of the beauty of the skin was performed by a third party on the image samples of the 248 women. As a third party, 4 panelists were selected and 248 cheek photos were presented and ranked in the order in which they were visually recognized as having beautiful skin. Specifically, using the reference photos of 10 levels with different skin beauty shown in FIG. 3, after ranking the 5 levels of A to E in the order of beautifulness, in each of the 5 levels, further Ranking was done in 10 stages. Visual evaluation values (scores) were given to 1 to 10 in order of more beautifulness among the A ranks recognized as the most beautiful. Similarly, in the B rank, the scores from 11 to 20 are beautiful, in the C rank, the scores from 21 to 30 are beautiful, in the D rank, the scores from 31 to 40 are beautiful, and in the E rank, the scores from 41 to 50 are beautiful. Gave. That is, the score of the most beautiful sample is 1, and the score of the least beautiful sample is 50. For each sample, these operations were performed three times independently for each of four panelists, and a total of 12 scores were averaged, and the average value was used as a visual evaluation value of the beauty of the skin.

As a result of the visual evaluation, the frequency (pixel number) distribution of the RGB values of the image of the skin having the smallest visual evaluation value (the most beautiful skin) and the image of the skin having the largest visual evaluation value (the least beautiful skin) is obtained. These are shown in FIGS. 5 and 6, respectively. It can be seen that the image with the least beautiful skin has a higher probability of more various RGB values than the image with the most beautiful skin, that is, the entropy of the distribution of RGB values is large.
Table 1 shows the results of calculating the entropy of the RGB value distribution of these skin images using a program that executes the calculation of the above equation (2). From this, it can be seen that the entropy of the distribution of the components of the RGB values of the image of the least beautiful skin is larger than that of the most beautiful skin.

<Relationship between entropy of distribution of values of each component of RGB value and visual evaluation value>
FIG. 7 shows the average value of entropy calculated for each of the RGB values for each of the five ranks A to E in the beautiful order. From this, the A group evaluated as the most beautiful by the visual evaluation by the panel has the smallest entropy average, and the E group evaluated as the least beautiful by the visual evaluation has the largest entropy average. That is, it can be seen that the entropy of the distribution of the values of all the components of the RGB value is related to the beauty of the skin actually visually evaluated by a third party.

<Relationship of entropy of distribution of values of each component of RGB value>
FIG. 8 shows a plot of the entropy of the distribution of each component value of RGB values in a three-dimensional coordinate space. Thus, the entropy of the distribution of the values of each component of the RGB values is significant and highly positive correlation (correlation coefficients: r _RG = 0.837, r _RB = 0.694, r _GB = 898, all It can be seen that p <0.01).

<Relationship between entropy of RGB value distribution and visual evaluation value>
Next, regression analysis was performed using the visual evaluation value obtained above as an objective variable and the entropy of the distribution of each component of the RGB values as explanatory variables. As a result, the correlation coefficient (r) between the visual evaluation value and each entropy is 0.492 (R), 0.702 (G), and 0.709 (B), respectively, and a significant high positive correlation. (All p <0.01). FIG. 9 shows the entropy relationship between the visual evaluation value and the distribution of the B component values.
The regression equation showing the relationship between the visual evaluation value (y) and the entropy (x _B ) of the distribution of the B component values was as follows.
y = 31.366 * x _B −150.600 (3)
The correlation coefficient is 0.709 (p <0.01)

Next, a multiple regression analysis was performed using the visual evaluation value (y) as an objective variable and the entropy (x _R , x _G , x _B ) of the distribution of each component of the RGB values as explanatory variables.
The multiple regression equation showing the relationship between the visual evaluation value (y) and the entropy (x _R , x _G , x _B ) of the distribution of the values of each component was as follows.
y = -10.690 * x _R + 27.035 * x _G + 14.446 * x _B -148.054 (4)
The multiple correlation coefficient is 0.737 (p <0.01)

  From this, it can be seen that the entropy of the distribution of the values of all components of the RGB value and the visual evaluation value have a very high correlation.

<Evaluation of skin beauty using entropy of distribution of values of each component of RGB value>
Next, RGB value distribution data is acquired from cheek photographs of four female subjects not included in the 248 subjects, and each entropy is calculated for these, and the values are substituted into the multiple regression equation. The evaluation value (y) was calculated.
The calculated visual evaluation values of the cheek photographs of the four female subjects were compared with the actual visual evaluation values obtained by the visual evaluation method described above. The results are shown in Table 2. From this, it can be seen that the visual evaluation value using the entropy of the RGB value distribution and the actual visual evaluation value agree very well.

<Relationship between entropy of distribution of values of each component of YUV value and visual evaluation value>
In Example 1, the entropy was similarly calculated by replacing the RGB values with the YUV values. That is, after the YUV value is obtained by performing matrix transformation on the RGB value according to a regular method (see Equation 1), the entropy is calculated for each component of the YUV value by using the program that executes the above-described calculation (2). Calculated. FIG. 10 shows the relationship between the entropy of the Y component value distribution and the visual evaluation value.
Next, regression analysis was performed using the visual evaluation value (y) obtained above as an objective variable and the entropy (x _Y ) of the distribution of the Y component values as explanatory variables.
The regression equation showing the relationship between the visual evaluation value (y) and the entropy (x _Y ) of the distribution of the Y component values was as follows.
y = 29.185 * x _Y -127.08 (5)
The correlation coefficient is 0.677 (p <0.01)

<Relationship between entropy of relief distribution and visual evaluation value>
Using a commercially available silicon from the cheek portion shown in FIG. 1, a 2 cm * 2 cm skin replica specimen at the center is collected, and then a high-precision three-dimensional image processing apparatus LIP-50 manufactured by Science Systems Co., Ltd. 3 is used. Dimensional relief data were obtained. About 1 cm * 1 cm area | region of the center part of the replica, 1000 scans were performed by the space | interval of 10 micrometers in the y direction (vertical direction). Since the sampling period in the x direction and the y direction is different from 9.4 μm and 10 μm, respectively, the LIP-50 uses a sinc function (see FIG. 2), and complement processing is performed so that both intervals in the xy direction are 10 μm. After that, the entropy was calculated using a program for executing the calculation of (2) described above.
FIG. 11 shows the relationship between the visual evaluation value and the entropy of the undulation value distribution.
Next, regression analysis was performed using the visual evaluation value (y) obtained above as an objective variable and the entropy (x) of the distribution of undulation values as an explanatory variable.
The regression equation showing the relationship between the visual evaluation value (y) and the entropy (x) of the distribution of undulation values was as follows.
y = 18.231 * x-94.111 (6)
The correlation coefficient is 0.556 (p <0.01)

<Evaluation of skin beauty using the entropy of the distribution of relief values>
Next, entropy was calculated from the undulation values obtained in the same manner as described above, using a replica specimen of the cheek of a 33-year-old female subject. The entropy was 6.51. This entropy was substituted into the regression equation to obtain a visual evaluation value of 24.6. Separately, when the visual evaluation of the beauty of the skin was actually performed by the above method, the visual evaluation value was 23.3, which was in good agreement with the calculated visual evaluation value.

  According to the present invention, it is possible to objectively and easily discriminate the appearance beauty of the skin reflected in the eyes of a third party, so that anyone can evaluate the beauty of the subject's skin. By using the method and apparatus of the present invention, it is possible to provide appropriate information to cosmetic users, such as advice on skin counseling, care methods, and selection of cosmetics.

It is a figure which shows the example of the measurement object area | region of a face. It is a figure which shows what correct | amended the undulation value data obtained from the replica by the Sinc function. It is a figure which shows the example of the cheek part photograph used for evaluation of this invention (drawing substitute photograph). It is a hardware block diagram of an example of the skin discrimination apparatus of this invention. It is a figure which shows the frequency (pixel number) distribution of the image of the most beautiful skin (drawing substitute photograph) in Example 1, and its RGB value. It is a figure which shows the image (drawing substitute photograph) of the most unsatisfactory skin in Example 1, and the frequency (pixel number) distribution of the RGB value. It is a figure in Example 1 which shows the relationship between the visual evaluation value of the beauty for every rank, and its entropy. FIG. 3 is a diagram in which the entropy of the distribution of each component value of RGB values in Example 1 is three-dimensionally plotted. It is a scatter diagram which shows the relationship between the visual evaluation value and the entropy of the distribution of the value of the B component in Example 1. FIG. 10 is a scatter diagram illustrating the relationship between the visual evaluation value and the entropy of the Y component value distribution in Example 3. In Example 4, it is a scatter diagram which shows the relationship between the visual evaluation value and the entropy of the distribution of relief values.

Claims (11)

  A skin discrimination method characterized by evaluating the beauty of a skin using an entropy of a distribution of skin property values as an index.   The skin discrimination method according to claim 1, wherein the skin property value is a color system signal of an image of the skin surface.   The skin discrimination method according to claim 2, wherein the color system signal is an RGB value, a YUV value, or a Munsell (HVC) value. Obtaining at least one color system signal of an image of the skin surface;
Calculating entropy for a distribution of values of at least one component constituting the color system signal;
The step of substituting the calculated entropy into a regression equation representing the relationship between the entropy of the distribution of the component value prepared in advance and the visual evaluation value of the beauty of the skin, and calculating the visual evaluation value of the beauty of the skin; The skin discrimination method according to claim 2 or 3, further comprising:   The regression equation is a regression equation obtained by multiple regression analysis of the entropy of the distribution of each component value constituting the RGB value, YUV value or Munsell (HVC) value and the visual evaluation value of the beauty of the skin. The skin discrimination method according to claim 4, wherein:   The skin identification method according to claim 1, wherein the skin property value is a undulation value of the skin surface. Obtaining a relief value on the skin surface;
Calculating the entropy of the distribution of the relief values;
Including the step of substituting the calculated entropy into a regression equation representing a relationship between the entropy of the distribution of the undulation value prepared in advance and the visual evaluation value of the skin beauty, and calculating the visual evaluation value of the skin beauty. The skin discrimination method according to claim 6. Means for obtaining at least one color system signal of the skin surface image;
Means for calculating entropy for a distribution of values of at least one component constituting the color system signal;
Means for substituting the calculated entropy into a regression equation representing the relationship between the entropy of the distribution of the value of the component prepared in advance and the visual evaluation value of the beauty of the skin, and calculating the visual evaluation value of the beauty of the skin;
A skin discrimination apparatus including means for displaying the calculated visual evaluation value. Means for obtaining skin relief values;
Means for calculating the entropy of the distribution of the relief values;
Means for substituting the calculated entropy into the regression equation representing the relationship between the entropy of the distribution of the relief values prepared in advance and the visual evaluation value of the beauty of the skin, and calculating the visual evaluation value of the beauty of the skin;
A skin discrimination apparatus including means for displaying the calculated visual evaluation value. Computer
Means for calculating entropy for a distribution of values of at least one component constituting a color system signal of an image of the skin surface;
Functions as a means for substituting the calculated entropy into a regression equation representing the relationship between the entropy of the distribution of the component values prepared in advance and the visual evaluation value of the beauty of the skin, and calculating the visual evaluation value of the beauty of the skin A skin discrimination program to help you. Computer
Means for calculating the entropy of the distribution of skin relief,
The calculated entropy is substituted into a regression equation that represents the relationship between the entropy of the distribution of the undulation value prepared in advance and the visual evaluation value of the beauty of the skin, and functions as a means for calculating the visual evaluation value of the beauty of the skin. For skin discrimination program.

Images