JP2017012384A - Wrinkle state analysis device and wrinkle state analysis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique for scoring a wrinkle state so that a view point of visual evaluation of wrinkles by an expert evaluator, is reflected based on a skin image.SOLUTION: A wrinkle state analysis method comprises: a step for acquiring a skin brightness image of an analysis object portion; a step for extracting one or more linear texture images in one or more prescribed angles corresponding to the analysis object portion; a step for extracting linear component intensities of one or more prescribed angles from the extracted one or more linear texture images; and a step for calculating an estimated visual evaluation score of a skin on the analysis object portion using a relational expression between the linear component intensities of one or more prescribed angles and a visual inspection score of wrinkles on a face image, obtained by regression analysis in which plural sample face images are objects, and the extracted linear component intensities.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a skin image analysis technique.

  Wrinkles that occur on the face, neck, etc. are factors that greatly influence the apparent age of human impression. However, the evaluation of the wrinkle state is generally performed visually, and a method for quantification is being studied. The following Patent Document 1 proposes a method for evaluating the wrinkle state of the skin. This proposed method extracts wrinkle information including the position and size of each wrinkle from the skin image, and evaluates the wrinkle state by comparing the extracted wrinkle information with wrinkle information obtained in the past. .

  The following Patent Document 2 proposes a wrinkle state analysis method for quantifying the skin wrinkle state based on a skin image. This proposed method extracts a linear texture image for each of a plurality of predetermined angles from a skin luminance image of a region to be analyzed, and linear (linear) components of each predetermined angle from each extracted linear texture image Extract each intensity. Then, using the extracted linear (linear) component intensities at a predetermined angle, wrinkle state information of the skin of the analysis target site is generated.

JP 2010-119431 A Japanese Patent Laying-Open No. 2015-62569

  In the proposed method of Patent Document 1 described above, wrinkle information is extracted by, for example, obtaining a three-dimensional shape of a wrinkle portion specified from a certain area of a captured image by a replica. Therefore, the above-mentioned Patent Document 1 does not disclose a specific method for specifying each wrinkle region directly from a captured image.

  In this regard, the above-mentioned Patent Document 2 extracts various linear components appearing on the skin of the analysis target part by extracting linear texture images for each of a plurality of predetermined angles by image processing on the skin luminance image of the analysis target part. It presents that the feature amount of (linear component) can be acquired. Further, Patent Document 2 presents a technical idea that the strength of the acquired linear component indicates the strength of the wrinkle feeling given by the groove group represented as the linear component.

  This invention adds the further idea obtained from the new idea with respect to the method shown by the above-mentioned patent document 2. FIG. That is, the present invention provides a technique for scoring a wrinkle state based on a skin image so that a viewpoint of visual evaluation of wrinkles by a professional evaluator is reflected.

  Each aspect of the present invention employs the following configurations in order to solve the above-described problems.

  The first aspect relates to a wrinkle state analysis method. The wrinkle state analysis method according to the first aspect includes an image acquisition step of acquiring a skin luminance image of an analysis target portion, and one or more predetermined corresponding to the analysis target portion from the skin luminance image acquired in the image acquisition step A line detection process for extracting one or more linear texture images for an angle, and an intensity extraction for extracting the linear component intensity of the one or more predetermined angles from the one or more linear texture images extracted in the line detection process And a relational expression between the linear component intensity of the one or more predetermined angles and the visual score of wrinkles on the face image obtained by regression analysis on a plurality of sample face images, and the intensity extraction step And a calculation step of calculating an estimated visual score of skin wrinkles of the analysis target site using the extracted linear component intensity.

  The second aspect relates to a wrinkle state analyzer. The wrinkle state analysis apparatus according to the second aspect includes an image acquisition unit that acquires a skin luminance image of an analysis target part, and one or more predetermined corresponding to the analysis target part from the skin luminance image acquired by the image acquisition unit Line detection means for extracting one or more linear texture images for the angle, and intensity extraction for extracting the linear component intensity of the one or more predetermined angles from the one or more linear texture images extracted by the line detection means And a relational expression between the linear component intensity of the one or more predetermined angles and the visual score of wrinkles on the face image obtained by regression analysis on a plurality of sample face images, and the intensity extracting means Calculating means for calculating an estimated visual score of wrinkles of the skin of the analysis target part using the extracted linear component intensity.

  Another aspect of the present invention is a program that causes at least one computer to execute the wrinkle state analysis method according to the first aspect, and is a computer-readable recording medium that records such a program. This recording medium includes a non-transitory tangible medium.

  According to each said aspect, a wrinkle state can be scored so that the viewpoint of the visual evaluation of a wrinkle by a professional evaluator may be reflected based on a skin image.

It is a figure which shows notionally the hardware structural example of the wrinkle state analyzer in 1st embodiment. It is a flowchart which shows the operation example of the wrinkle state analyzer in 1st embodiment. It is a figure which shows notionally the process structural example of the wrinkle state analyzer in 1st embodiment. It is a flowchart which shows the operation example of the wrinkle state analyzer in 2nd embodiment. It is a figure which shows the 1st output example of an estimated visual score. It is a figure which shows the 2nd output example of an estimated visual score. It is a figure which shows notionally the process structural example of the wrinkle state analyzer in 2nd embodiment. It is a flowchart which shows the operation example of the wrinkle state analyzer in 3rd embodiment. It is a figure which shows the example of the position of an analysis object site | part. It is a figure which shows notionally the process structural example of the wrinkle state analyzer in 3rd embodiment. It is a graph which shows the correlation with the angle of a linear component intensity | strength, the binarization threshold value, and a visual score. It is a graph which shows the relationship between the added angle and the determination coefficient of freedom degree adjustment in the regression analysis which added one angle to angle "135 degree | times". It is a graph which shows the relationship between the added angle and the freedom degree adjusted determination coefficient in the regression analysis which added one angle to angle "135 degree" and "85 degree". It is a graph showing the relationship between the estimated visual score S calculated by the single regression equation (1) and the visual score obtained by sensory evaluation. It is a graph showing the relationship between the estimated visual score S calculated by multiple regression equation (2) and the visual score obtained by sensory evaluation. It is a figure which shows the skin luminance image of the analysis object site | part of sample provider # 1 and # 2. It is a table | surface which compares the linear component intensity | strength obtained from sample provider's # 1 and # 2 skin luminance image, and an estimated visual score. It is a figure which shows the example which arranges and outputs the 135 degree | times and 85 degree linear texture image about sample provider # 1 and # 2. It is a figure which shows the example of an output of the superimposition image of the 135 degree | times and 85 degree linear texture image about sample provider # 1 and # 2.

  Embodiments of the present invention will be described below. In addition, each embodiment given below is an illustration, respectively, and this invention is not limited to the structure of each following embodiment.

[First embodiment]
[Hardware configuration]
FIG. 1 is a diagram conceptually illustrating a hardware configuration example of a wrinkle state analyzer (hereinafter simply referred to as an analyzer) 10 in the first embodiment. The analysis device 10 in the first embodiment is a so-called computer, and includes a CPU (Central Processing Unit) 1, a memory 2, an input / output interface (I / F) 3, a communication unit 4, and the like.

In addition to a general CPU, the CPU 1 includes an application specific integrated circuit (ASIC), a DSP (Digital Signal Processor), a GPU (Graphics Processing Unit), and the like.
The memory 2 is a RAM (Random Access Memory), a ROM (Read Only Memory), or an auxiliary storage device (such as a hard disk).

The input / output I / F 3 can be connected to user interface devices such as the output device 5 and the input device 6. The output device 5 is a display device, a projection device (projector), a printing device (printer), a speaker, or the like. The display device is a device that displays a screen corresponding to drawing data processed by the CPU 1 or the like, such as an LCD (Liquid Crystal Display) or a CRT (Cathode Ray Tube) display. The input device 6 is a device that accepts an input of a user operation such as a keyboard and a mouse, a microphone (microphone), an imaging device (camera, video), and the like. The display device and the input device 6 may be integrated and realized as a touch panel.
The communication unit 4 performs communication with other computers, exchange of signals with other devices, and the like. An imaging device (not shown), a portable recording medium, and the like can be connected to the communication unit 4.

  The analysis apparatus 10 may include hardware elements not shown in FIG. 1, and the number of hardware elements is not limited. For example, the analyzer 10 may have a plurality of CPUs 1 or may be connected to a plurality of output devices 5. The hardware configuration of the analyzer 10 is not limited to the content shown in FIG.

[Operation example / Wrinkle state analysis method]
FIG. 2 is a flowchart showing an operation example of the wrinkle state analyzer 10 in the first embodiment. Hereinafter, the wrinkle state analysis method according to the first embodiment will be described with reference to FIG. The wrinkle state analysis method in the first embodiment is executed by at least one computer such as the analysis apparatus 10. For example, the wrinkle state analysis method is realized by the CPU 1 executing a program stored in the memory 2.

  As shown in FIG. 2, the wrinkle state analysis method executed by the analyzer 10 includes an image acquisition step (S21), a line detection step (S22), an intensity extraction step (S23), and a calculation step (S24). . Hereinafter, these steps will be described in detail.

  In the image acquisition step (S21), the analysis device 10 acquires a skin luminance image of the analysis target part. The analysis target site is not limited as long as it is a site where wrinkles can occur in a part of human skin. “Wrinkles” may be classified into a plurality of types depending on the cause of occurrence, such as aging, drying, and deterioration of dermal function. For example, wrinkles called “small wrinkles” are wrinkles generated by the dry state of the epidermis, and are known to be repairable by improving the dry state by skin care. Further, it is known that wrinkles caused by a decrease in dermal function have deeper grooves than “small wrinkles” and are difficult to repair only by improving the dry state of the epidermis. The analysis target part may be determined depending on which type of wrinkle state is analyzed among these types. For example, when analyzing the state of fine wrinkles, a predetermined range where fine wrinkles are likely to occur, such as cheeks, around the eyes, around the mouth, and neck, is selected as the analysis target part. Moreover, when analyzing the state of wrinkles caused by aging, for example, a predetermined range of the corner of the eye, the forehead, and the neck is selected as the analysis target part.

  The skin luminance image is an image in which a skin analysis target part is captured, and is an image having light / dark information for each pixel. The skin luminance image may be a gray scale image obtained by converting a color image captured by a general camera into a gray scale. The skin luminance image may be acquired as an image file in JPEG (Joint Photographic Experts Group) format, BMP (Bitmap image) format, TIFF (Tagged Image File Format) format, GIF (Graphic Interchange Format) format, or the like. Alternatively, it may be acquired as data in which brightness information of each pixel is listed. In addition, since it is preferable that the pixel value has a linear relationship with the luminance, it is preferable that γ correction is not performed on the skin luminance image used here. When γ correction is applied to the image captured by the camera, it is preferable to convert the pixel value into a linear relationship with the luminance by inverse γ correction. This embodiment does not limit the data format of the skin luminance image. The analysis device 10 can also acquire the skin luminance image from the portable recording medium, another device, or the like via the communication unit 4.

  In the line detection step (S22), the analysis apparatus 10 extracts one or more linear texture images for one or more predetermined angles corresponding to the analysis target site from the skin luminance image acquired in the image acquisition step (S21). To do. A linear texture image for one predetermined angle is an image representing only a linear component extending in the direction of the predetermined angle in the skin luminance image. Here, strictly speaking, the linear component represented by the linear texture image may include not only a component existing in the predetermined angle direction but also a component existing within a range of the allowable angle from the predetermined angle. The range of the allowable angle depends on the linear texture image extraction method.

  The linear component represented by the linear texture image is a linear pattern that appears in the skin luminance image, and almost corresponds to a linear groove (dent) formed on the skin surface of the analysis target site. Therefore, according to the line detection step (S22), various linear components appearing on the skin of the analysis target site, including grooves that are visually recognized as wrinkles and grooves that are difficult to be visually recognized (such as shallow skin grooves). It can be extracted as a linear texture image.

  As will be described in detail in the section of the examples, the present inventors have made a relationship between the result of visual evaluation of the wrinkle state by a professional evaluator and the linear component intensity at each angle for a plurality of sample face images. Was subjected to regression analysis. The result of the visual evaluation of the wrinkle state by this professional evaluator was expressed numerically as a visual score. That is, the visual score is a value obtained by digitizing the wrinkle state. For example, the visual score is determined to be a value indicating the level of the number of wrinkles to be visually recognized (not at all = 1, some at some = 2, some = 3, very at = 4). The state of the wrinkle represented by the visual score is not limited to such an example, and may vary depending on the standard of visual evaluation. Specifically, the wrinkle state may be represented by a visual score by focusing on one type of wrinkles of the plurality of types described above. For example, the state of “wrinkles” (herein referred to as “dermal wrinkles”) caused by deterioration of the dermis state may be represented by a visual score. In this case, the expert evaluator distinguishes “skin wrinkles” from other types of wrinkles such as “small wrinkles” as much as possible, and determines the state (number, length, depth, etc.) of “skin wrinkles”. The level is determined as a visual score.

  The present inventors perform regression analysis on the relationship between the visual score and the linear component intensity for each angle with respect to a plurality of sample face images, thereby correlating the visual score and the correlation with respect to a certain analysis target portion. A high linear component angle was found. Furthermore, the present inventors have found that if the linear component intensity at an angle highly correlated with the visual score is used in a limited manner, the visual score of the wrinkle at the site to be analyzed can be sufficiently explained. Thus, in this embodiment, a limited number of predetermined angles that can sufficiently explain the wrinkle visual score of the analysis target part is determined in advance corresponding to the analysis target part and held in the memory 2. For example, the analysis apparatus 10 extracts one or two linear texture images for one or two predetermined angles. Thereby, compared with the method of extracting a linear texture image about many predetermined angles, processing time can be shortened, maintaining the precision of scoring of a wrinkle state.

  However, the number of predetermined angles to be extracted in the line detection step (S22) may be determined to be three or more. For example, an angle obtained by dividing an angle between 0 degrees and 360 degrees at equiangular intervals (for example, 10 degrees), or a plurality of angles may be used as predetermined angles in descending order of correlation with the visual score. The analysis device 10 can also cause the user to input the one or more predetermined angles by using the input device 6 and obtain the one or more predetermined angles as input data.

  Various well-known methods can be used as a method for extracting a linear texture image. For example, edge detection operators such as Prewitt operator and Sobel operator, line detection operators such as VanderBrug operator, Paton operator and Kasband operator, Wavelet, Steerable, and Gabor filters can be used as extraction methods. The present embodiment does not limit the specific extraction method of the linear texture image. As this extraction method, an embodiment using a Gabor filter will be described later.

  In the intensity extraction step (S23), the analysis apparatus 10 extracts one or more linear component intensities at a predetermined angle from the one or more linear texture images extracted in the line detection step (S22). The linear component intensity at one predetermined angle indicates the characteristic amount (linear component amount) of the linear component extending in the direction of the predetermined angle. Various well-known texture analysis methods can be used for the extraction method of the linear component intensity. The present embodiment does not limit the extraction method of the linear component strength as long as the feature amount of the linear component extending in the direction of each predetermined angle can be obtained.

  For example, the analysis apparatus 10 can calculate a variance value or standard deviation value from the density histogram of the linear texture image, and can use this variance value or standard deviation value as the linear component intensity. A linear texture image showing a large dispersion value is an image with high contrast between light and dark, and an image with high contrast between light and darkness is thought to contain a mixture of strong convex components (bright components) and strong concave components (dark components). It is done. This mixed state can be said to be a state in which the amount of linear components of the image is large. On the other hand, the mixed state in the linear texture image can be considered as a state in which a large number of concavo-convex shapes exist on the skin surface. Therefore, it can be said that the dispersion value obtained from the density histogram of the linear texture image indicates the feature amount of the linear groove formed on the skin surface and extending in the predetermined angle direction, that is, the linear component intensity.

  In addition to the density histogram method (GLHM), a density level difference method (GLDM), a spatial density level dependent method (SGLDM), or the like may be used as the extraction method. Further, the analysis apparatus 10 binarizes the linear texture image using a predetermined binarization threshold, and calculates the area of one value in the obtained binarized image as the linear component intensity. It may be. In this case, in the line detection step (S22), the analysis apparatus 10 may execute up to the binarization process, and the binarized image after the binarization process may be referred to as a linear texture image. The area of one value corresponds to the number of white or black pixels, for example. The analysis apparatus 10 may calculate the ratio of the area of one value to the area of the other value, or the ratio of the area of one value to the entire area as the linear component intensity. The predetermined binarization threshold value is set in advance according to the imaging magnification of the analysis target region shown in the skin luminance image, the imaging environment (brightness, etc.) of the analysis target region, etc., and is stored in the memory 2. In the present embodiment, it is desirable that the predetermined binarization threshold is set to a value from which a linear component intensity having a high correlation with the visual score described above can be extracted. A specific method for determining a desirable binarization threshold will be described in detail in the section of the embodiment.

  In the calculation step (S24), the analysis apparatus 10 calculates the linear component intensity of the one or more predetermined angles and the visual score of wrinkles on the face image obtained by regression analysis on a plurality of sample face images. Using the relational expression and the linear component intensity extracted in the intensity extraction step (S23), an estimated visual score of wrinkles on the skin of the analysis target site is calculated. The predetermined angle corresponding to the intensity of the linear component forming the relational expression is the same as one or more predetermined angles corresponding to the analysis target part, which are targeted in the line detection step (S22). In the present embodiment, the regression analysis described above provides a single regression equation in which the linear component intensity at the first predetermined angle having the highest correlation with the visual score described above is an explanatory variable and the visual score is an objective variable, or the first regression equation thereof. A multiple regression equation having the linear component intensities at two predetermined angles including the predetermined angle as explanatory variables and the visual score as an objective variable is generated and held in advance, and is used as the above-described relational expression. Specific examples of the relational expressions will be described later as examples.

  The analysis apparatus 10 calculates the estimated visual score by substituting the linear component intensity extracted in the intensity extraction step (S23) into the relational expression generated and held in advance in this way. The calculated estimated visual score is a wrinkle state that is estimated to be obtained as an evaluation result when the state of wrinkles of the analysis target portion that is reflected in the skin luminance image acquired in (S21) is visually evaluated by a professional evaluator. Means the score. Thereby, the analyzer 10 can calculate the score which approximates the result of an actual visual evaluation, without making a professional evaluator actually perform visual evaluation regarding the wrinkle state of the analysis target part of the subject.

  The analysis device 10 stores the calculated estimated visual score in the memory 2. For example, the analysis apparatus 10 acquires the identification information of the subject, and stores the calculated estimated visual score in association with the identification information. Further, the analysis device 10 may store the estimated visual score in a portable recording medium via the communication unit 4 or may provide it to another computer.

[Processing configuration]
FIG. 3 is a diagram conceptually illustrating a processing configuration example of the wrinkle state analysis apparatus 10 in the first embodiment. The analysis apparatus 10 in the first embodiment includes an image acquisition unit 11, a line detection unit 12, an intensity extraction unit 13, a calculation unit 14, and the like. Each of these processing modules is realized, for example, by executing a program stored in the memory 2 by the CPU 1. The program may be installed from a portable recording medium such as a CD (Compact Disc) or a memory card or another computer on the network via the input / output I / F 3 and stored in the memory 2. Good.

  The analysis apparatus 10 can execute the above-described wrinkle state analysis method by operating each processing module shown in FIG. That is, the analyzer 10 operates as shown in FIG. In this case, the image acquisition unit 11 executes the above-described image acquisition step (S21), the line detection unit 12 executes the above-described line detection step (S22), and the intensity extraction unit 13 performs the above-described intensity extraction step. (S23) is executed, and the calculation unit 14 executes the above-described calculation step (S24).

  The analysis device 10 may hold the skin luminance image of the site to be analyzed in advance by the analysis device 10 itself, may be generated by the analysis device 10 itself, or from another computer, a portable recording medium, or the like. You may acquire via input-output I / F3.

[Operation and Effect in First Embodiment]
As described above, in the first embodiment, one or more linear texture images for one or more predetermined angles corresponding to the analysis target region are extracted from the skin luminance image of the analysis target region, and one or more linear shapes are extracted. One or more linear component intensities at a predetermined angle are extracted from the texture image. On the other hand, by regression analysis for a plurality of sample face images, a relational expression between a linear component intensity at one or more predetermined angles and a visual score indicating a result of visual evaluation of wrinkles by a professional evaluator is defined in advance. Yes. And the estimated visual score of the wrinkle of the skin of an analysis object site | part is calculated by applying the linear component intensity | strength about the calculated one or more said predetermined angle to the relational expression.

  Thereby, according to the first embodiment, the actual visual evaluation of the skin wrinkle state of the analysis target part is performed based on the skin luminance image of the analysis target part without causing the professional evaluator to actually perform the visual evaluation. A score approximating the result can be obtained. As a result, it is possible to reduce the labor and burden of the professional evaluator and to reduce the cost for analyzing the wrinkle state. Further, according to the present embodiment, it is possible to stably and quantitatively evaluate the appearance impression of the wrinkled state.

  Furthermore, according to the first embodiment, by using the angle of the linear component strength having a high correlation with the visual score, the linear component strength of a small number of angles is used as a result of the actual visual evaluation by the professional evaluator. An approximate estimated visual score can be calculated. Thus, according to the first embodiment, it is possible to improve the calculation speed of the estimated visual score while maintaining the accuracy as compared with the method of calculating the linear texture image and the linear component intensity with respect to a large number of predetermined angles. it can.

  Further, according to the first embodiment, the wrinkle state represented by the calculated estimated visual score can be limited to a certain type of wrinkle state such as “small wrinkles” and “dermal wrinkles”. In this case, the visual evaluation by the expert evaluator is performed mainly on the state of the target type of wrinkles, and the visual score obtained as the evaluation result is used for the regression analysis. This makes it possible to calculate an estimated visual score that centrally represents a desired type of wrinkle state.

[Second Embodiment]
The second embodiment effectively outputs the result of the wrinkle state analysis in the first embodiment. Hereinafter, the analysis apparatus 10 and the wrinkle state analysis method according to the second embodiment will be described focusing on the content different from the first embodiment. Hereinafter, the description of the same content as the first embodiment is omitted as appropriate. The hardware configuration of the analyzer 10 in the second embodiment is the same as that of the first embodiment shown in FIG.

[Operation example (wrinkle state analysis method)]
Hereinafter, the wrinkle state analysis method according to the second embodiment will be described with reference to FIG. FIG. 4 is a flowchart showing an operation example of the analyzer 10 in the second embodiment. In FIG. 4, the same steps as those in FIG. 2 are denoted by the same reference numerals.

  In addition to the method of the first embodiment shown in FIG. 2, the wrinkle state analysis method in the second embodiment executed by the analyzer 10 further includes a step (S41) of acquiring age information and an output step (S42). Including. Hereinafter, each process will be described in detail with a focus on processes having different contents from the first embodiment.

  In the step (S41), the analysis apparatus 10 acquires the age information of the subject shown in the skin luminance image acquired in the step (S21). The acquired age information may indicate a specific age, an age such as 30's or 40's, or an age range such as 30 or more and less than 35, 35 or more and less than 40 May be indicated. The method for acquiring age information by the analyzer 10 is not limited. For example, the analysis device 10 displays a screen for allowing the subject or other user to input the age information on the display device of the output device 5, and acquires the age information by an input operation using the input device 6 on the screen. be able to. Moreover, the analyzer 10 may acquire age information by recognizing a voice obtained from a microphone. Moreover, the analyzer 10 can also estimate the age from a face image obtained by imaging the subject's face, and can acquire the estimated age as age information. As long as a process (S41) is before execution of a process (S42), it may be performed in any order.

  In the step (S42), the analysis apparatus 10 outputs the estimated visual score calculated in the step (S24). In the present embodiment, the analysis apparatus 10 specifies the representative visual score corresponding to the age information acquired in step (S41), and uses the specified representative visual score and the estimated visual score calculated in step (S24). Output side by side. In this case, the analysis apparatus 10 holds a representative visual score for each age, each age, or each age range in advance. The analysis device 10 can specify a representative visual score corresponding to the age, age, or age range indicated by the acquired age information from the plurality of representative visual scores held. The representative visual score indicates a standard (typical) wrinkle state of the skin of a person of that age, age, or age range. For example, the representative visual score is set to an average value of visual scores or estimated visual scores obtained with respect to the skin of a person in age, age, or age range.

  Moreover, the analysis apparatus 10 specifies the representative linear texture image corresponding to the age information acquired in the step (S41), and uses the representative linear texture image and the linear texture image extracted in the step (S22). These can be output side by side at a predetermined angle corresponding to the analysis target part. In this case, the analysis device 10 holds in advance a representative linear texture image for each of one or more predetermined angles for each age, each age, or for each age range. The analysis apparatus 10 can specify a representative linear texture image corresponding to the age, age, or age range indicated by the acquired age information from among the plurality of representative linear texture images held. The representative linear texture image is, for example, a visual score or estimation that is close to the representative visual score among those previously extracted from the skin luminance image of the analysis target region of the person in the corresponding age range by the same method as in the step (S22). A linear texture image of a person who has a visual score can be used.

  FIG. 5 is a diagram illustrating a first output example of the estimated visual score. In the example of FIG. 5, the analysis apparatus 10 includes an image RN including the analysis target region of the subject, a linear texture image YT of a predetermined angle PA extracted from the skin luminance image of the analysis target region of the test subject, and a predetermined age in the same age as the test subject. The representative linear texture images AT of the angle PA are arranged and displayed on the display device. Furthermore, the analysis apparatus 10 displays the calculated estimated visual score “Your Wrinkle Index: 4” and the average visual score of the same age as the subject “Average Wrinkle Index: 3” on the display device. The subject who sees this output can grasp the wrinkle state of his / her skin with a specific numerical value (estimated visual score) and can easily compare it with the average of the same age as the subject. In addition, the subject can recognize the distribution and amount of the linear texture extending in the direction of the predetermined angle PA by looking at the linear texture image of the subject, and also compare with the linear texture of the same age. Can do. As shown in the example of FIG. 5, the analysis apparatus 10 further displays the comparison result between the estimated visual score and the representative visual score as text (text “Your skin is more wrinkled than the average of the same age”, etc.) ).

  In the example of FIG. 5, the linear texture images are output side by side only for the predetermined angle PA. However, the analysis apparatus 10 generates a linear texture image of the same age as the linear texture image of the subject for each of the plurality of predetermined angles. Can be displayed side by side. In this way, the state of the linear texture for each angular direction can be confirmed.

  FIG. 6 is a diagram illustrating a second output example of the estimated visual score. As shown in FIG. 6, the analysis apparatus 10 can also display an image obtained by superimposing a plurality of linear texture images extracted in the step (S22) on the display device. In the example of FIG. 6, an image ST in which a linear texture image with a predetermined angle PA1 and a linear texture image with a predetermined angle PA2 are superimposed is displayed. The displayed image ST represents a superposition of linear textures extending in the directions of the angles PA1 and PA2, which are highly correlated with the visual score of the wrinkled state, and excludes linear textures extending in the other angular directions. It can be said that it is an image that emphasizes the appearance characteristics of the wrinkled state. Therefore, the subject who has viewed the displayed superimposed image can easily grasp the appearance of his / her wrinkle state. In addition, the analysis apparatus 10 can output not only a plurality of linear texture images but also an image in which linear components are represented in different colors for each linear texture image. This makes it easier to grasp the state of the linear texture for each angular direction.

  The linear texture image and representative linear texture image output in the step (S42) may be images extracted from the skin luminance image by an arbitrary extraction method in the line detection step (S22), An image obtained by further processing by binarization processing or the like may be used. When outputting a superimposed image of linear texture images of a plurality of angles, the extracted images may be superimposed, or the extracted image and a processed image may be superimposed. In other words, the method for extracting the superimposed linear texture image may be different. For example, a linear texture image of a certain angle obtained by applying the Gabor filter is superimposed on a linear texture image of another angle obtained by executing binarization processing after applying the Gabor filter. May be output.

  The above-described output by the analysis device 10 may be a display on a display device or a print on a printing device. Moreover, the analysis apparatus 10 can also output the sound which reads a test subject's estimated visual score and a representative visual score from a speaker. The output method by the analyzer 10 is not limited.

[Processing configuration]
FIG. 7 is a diagram conceptually illustrating a processing configuration example of the wrinkle state analysis apparatus 10 in the second embodiment. The analysis apparatus 10 in the second embodiment further includes an information acquisition unit 16 and an output processing unit 17 in addition to the processing configuration of the first embodiment. Each of these processing modules is also realized, for example, by executing a program stored in the memory 2 by the CPU 1.

  The analysis apparatus 10 can execute the above-described wrinkle state analysis method by operating each processing module shown in FIG. That is, the analyzer 10 operates as shown in FIG. In this case, the information acquisition part 16 performs the process (S41) which acquires age information, and the output process part 17 performs the above-mentioned output process (S42).

[Operation and Effect in Second Embodiment]
In the second embodiment, the representative visual score corresponding to the age information of the subject and the estimated visual score calculated for the subject are output side by side. As a result, the subject can grasp the wrinkle state of his / her skin with specific numerical values (estimated visual score), and the relative position of the subject's wrinkle state with the standard wrinkle state of the same age It can be easily grasped.

  In the second embodiment, the representative linear texture image corresponding to the subject's age information and the linear texture image extracted from the subject's skin luminance image are output side by side for each predetermined angle corresponding to the analysis target region. Is done. Accordingly, the subject can recognize the wrinkle state (linear component state) of his / her skin for each predetermined angle direction in the image, and can also compare it with a standard wrinkle state of the same age.

  In the second embodiment, an image in which a plurality of linear texture images representing a plurality of linear textures having a predetermined angle are superimposed is output. Since each linear texture image represents a linear component having an angle showing a high correlation with the visual score of the wrinkle state, this superimposed image emphasizes the linear component that has a strong influence on the appearance of the wrinkle state. Therefore, by the output of the superimposed image, the subject can easily grasp the appearance impression of his / her wrinkle state. In addition, in the superimposed image, by expressing the linear component with a different color for each linear texture image, it is possible to make it easier to grasp the state of the linear texture for each angular direction.

  As described above, according to the second embodiment, since the wrinkle state of the subject can be effectively output, the counseling of the wrinkle state and the counseling of the skin care product and the cosmetics for improving the wrinkle state are the first. Two embodiments can be utilized.

[Third embodiment]
Hereinafter, the analysis device 10 and the wrinkle state analysis method according to the third embodiment will be described focusing on the contents different from the above-described embodiments. In the third embodiment, a specific method for acquiring a skin luminance image that can increase the analysis accuracy of the wrinkle state is newly added. Hereinafter, the description of the same content as the first and second embodiments is omitted as appropriate. The hardware configuration of the analyzer 10 in the third embodiment is the same as that of the first embodiment shown in FIG.

[Operation example (wrinkle state analysis method)]
Hereinafter, the wrinkle state analysis method according to the third embodiment will be described with reference to FIG. FIG. 8 is a flowchart showing an operation example of the analyzer 10 in the third embodiment. In FIG. 8, the same steps as those in FIG.
The wrinkle state analysis method according to the third embodiment executed by the analysis apparatus 10 further includes an image processing step (S81) and a standardization step (S82) in addition to the steps of the first embodiment or the second embodiment. However, FIG. 8 shows a wrinkle state analysis method in which steps (S81) and (S82) unique to the third embodiment are added to the steps of the first embodiment. Hereinafter, each process will be described in detail with a focus on processes having different contents from the first embodiment.

  In the image processing step (S81), the analyzer 10 acquires a surface reflected light image from each polarization image obtained by projecting S-polarized light on the subject's skin and imaging S-polarized light and P-polarized light. Specifically, the analyzer 10 projects a polarization image (SS polarization image) obtained by projecting S polarization and imaging S polarization, and a polarization image (S-S polarization image) obtained by projecting S polarization and imaging P polarization ( The difference from the SP polarization image is taken. The SS polarization image is an image having a strong surface reflection light component on the subject's skin and a small amount of the internal reflection light component, and the SP polarization image is an image having a strong internal reflection light component on the subject's skin. Therefore, by obtaining the difference between the SS polarization image and the SP polarization image, it is possible to obtain a surface reflection light image excluding components (internal reflection light) other than the surface reflection light component. Here, the direction of S-polarized light can be set arbitrarily. On the other hand, the direction of P-polarized light needs to be orthogonal to the direction of S-polarized light. For example, S-polarized light can be set in the vertical direction, and P-polarized light can be set in the horizontal direction. Similarly to the skin luminance image in the first embodiment, it is preferable that each polarization image used here has a linear relationship between the pixel value and the luminance. Therefore, it is preferable that γ correction is not performed. The analysis apparatus 10 can acquire each polarization image from the imaging device, the portable recording medium, another device, or the like via the input / output I / F 3.

  Furthermore, the analyzer 10 further uses a PS polarized image obtained by projecting P polarized light and imaging S polarized light, and a PP polarized image obtained by projecting P polarized light and imaging P polarized light. A surface reflected light image may be acquired. Thereby, a surface reflected light image in which all polarization components are correctly considered can be acquired. However, when using a P-S polarized image and a P-P polarized image, there is a problem that the photographing time becomes long, the subject is easily shifted between images, and more photographing equipment is required. is there. Therefore, according to the comparison between such a problem and the effect of noise suppression, a method using only the combination of the SS polarization image and the SP polarization image as described above is desirable.

  In the normalization step (S82), the analysis apparatus 10 extracts the normalized skin image of the analysis target part from the surface reflected light image obtained in (S81). Specifically, the analysis apparatus 10 detects an analysis target part from the surface reflected light image, and a standardized image (skin image) of the analysis target part based on the detected image region of the analysis target part. To extract. When the resolution, size, and direction of the analysis target part in each polarization image acquired in (S81) are not uniform, as the normalization process, adjustment of the resolution and size of the image, and the direction of the analysis target part in the image are performed. Adjustment etc. may be performed.

FIG. 9 is a diagram illustrating an example of the position of the analysis target part. In the example of FIG. 9, a region surrounded by a white line, that is, a square region below the orbit is detected as an analysis target site.
The method for detecting the analysis target part is not limited. For example, the analysis apparatus 10 can automatically detect the eye and nose region in the surface reflected light image, and can automatically detect the analysis target site from the relative positional relationship between the eye and nose region. As another example, by operating the input device 6, the user may designate a region of the analysis target part in the surface reflected light image and cause the analysis device 10 to generate an image corresponding to the region. . In addition, a region fixed in advance in the surface reflected light image may be set as the analysis target site.

  In the image acquisition step (S21), the analysis apparatus 10 acquires the skin luminance image by deleting the color information from the skin image obtained in (S82). As a method for deleting color information, an average value of RGB values or a weighted average value, or a method of simply using any one of RGB values as a representative value of each pixel can be used. Based on this skin luminance image, (S22) and subsequent steps are executed in the same manner as in the method of the first embodiment.

[Processing configuration]
FIG. 10 is a diagram conceptually illustrating a processing configuration example of the analysis apparatus 10 in the third embodiment. The analysis apparatus 10 in the third embodiment further includes an image processing unit 18 and a normalization unit 19 in addition to the processing configuration of the first embodiment or the second embodiment. However, FIG. 10 shows a processing configuration in which the image processing unit 18 and the normalization unit 19 unique to the third embodiment are added to the processing configuration of the first embodiment. The image processing unit 18 and the standardization unit 19 are also realized by executing a program stored in the memory 2 by the CPU 1, as with other processing modules.

  The analysis apparatus 10 can execute the above-described wrinkle state analysis method by operating each processing module shown in FIG. That is, the analyzer 10 operates as shown in FIG. In this case, the image processing unit 18 executes the above-described image processing step (S81), and the normalization unit 19 executes the above-described normalization step (S82). Other processing modules are the same as in the first embodiment.

  The analyzer 10 can acquire each polarization image from another computer, a portable recording medium, or the like via the input / output I / F 3. Moreover, the analyzer 10 may further include an imaging device not shown in FIG. 1 connected to the input / output I / F 3. In this case, the analyzer 10 can also acquire each polarization image from the imaging device.

  In the third embodiment described above, the normalization step (S82) and the normalization unit 19 are provided. However, the polarization image is generated in a standardized state, and only the analysis target portion is copied in the polarization image. If it is, the normalization step (S82) and the normalization unit 19 can be omitted.

[Operations and effects in the third embodiment]
In the third embodiment, a surface reflected light image is acquired from the polarization image, and a normalized skin image of the analysis target part is acquired from the surface reflected light image. Then, by deleting the color information from the skin image, a skin luminance image of the analysis target part is acquired, and an estimated visual score is calculated based on the acquired skin luminance image, as in the first embodiment. Is done.

  The surface reflected light image represents a component (unevenness component) derived from the shape of the skin surface excluding color unevenness components such as bears, spots, and freckles. Therefore, according to the third embodiment, the linear texture image extracted from the skin luminance image is further prevented from including a linear component (noise component) derived from other than the shape of the skin surface of the analysis target site. Can do. Thereby, since the linear component intensity | strength of each predetermined angle can be extracted in the state which excluded noise components other than the shape (mainly groove | channel) of the skin surface, a wrinkle state can be analyzed with higher precision. .

  Examples will be given below, and the above-described embodiments will be described more specifically. Each above-mentioned embodiment does not receive any limitation from the following examples.

  The present inventors performed sensory evaluation on a plurality of sample face images, and obtained visual scores of wrinkled states for each subject's skin appearing in each sample face image. In sensory evaluation, three expert evaluators looked at 210 sample face images (photos) showing the entire faces of 210 sample providers, and evaluated the wrinkle state of each sample provider's skin. . In this example, the cheek (lower orbit) portion of the entire face represented by the sample face image is focused on, focusing on “small wrinkles” that depend on the dry state of the epidermis. The condition of wrinkles was evaluated. Further, in this sensory evaluation, in order to unify the evaluation criteria of each professional evaluator, four full-face images representing the four stages of “small wrinkles” were provided as comparison targets. The visual score of all face images with very few wrinkles was set to 1 point, the visual score of all face images with very many wrinkles was set to 4 points, and the visual scores of the two full face images in between were set to 2 points and 3 points. . Each expert evaluator compares each sample face image with 1 point, 1.5 points, 2 points, 2.5 points, 3 points while comparing the four face images with each sample face image. , 3.5 points, or 4 points, respectively. Thereby, the score for three persons was attached | subjected about each sample face image, respectively, and the average of the score for three persons was each determined as the visual score of each sample face image.

  The present inventors perform regression analysis on the 210 skin luminance images including the analysis target regions of the 210 sample providers described above, so that the linear component intensity of one predetermined angle, or two or more We examined whether the visual score could be estimated with high accuracy by linear combination of linear component intensities at a predetermined angle. In this examination, the part focused on the sensory evaluation (left lower eye socket, see FIG. 9) is the analysis target part, and the skin luminance image of this analysis target part (size = 550 × 550 pixels, resolution on the target part = 18) Pixel / mm) is extracted ((S81), (S82), and (S21) in FIG. 8), and a linear texture image at a predetermined angle is extracted from the extracted plurality of skin luminance images (FIG. 8). (S22). Further, the extracted linear texture image is binarized with a predetermined binarization threshold, and the number of white pixels in the obtained binarized image is calculated as the linear component intensity ((( S23)).

A Gabor filter corresponding to the predetermined angle θ was used for extracting the linear texture image. In general, the Gabor filter is represented by the following equation. Appropriate values were set for the other parameters λ, φ, σ, and γ that determine the shape of the Gabor filter according to the imaging environment of the sample face image (λ = 10, φ = 0, σ = 4, γ = 1). The details of the Gabor filter are known as described in Patent Document 2.

  And while changing the combination of the binarization threshold value for obtaining the linear component strength and the target angle θ, the present inventors changed the correlation coefficient between the linear component strength of the angle θ and the visual score. A combination of an angle and a threshold value that can be calculated sequentially to obtain a linear component intensity highly correlated with the visual score was searched. In the present embodiment, Gabor filters corresponding to 36 predetermined angles from 0 degrees to 175 degrees set with an angle width of 5 degrees were sequentially used. In addition, this angle notation is a notation when the vertical reverse direction is 0 degree, the vertical direction is 180 degrees, and it is shown to increase clockwise toward the face.

  FIG. 11 is a graph showing the correlation between the angle of the linear component intensity and the binarization threshold value and the visual score. In FIG. 11, the horizontal axis indicates the angle, the vertical axis indicates the correlation coefficient, and the binarization threshold is distinguished by the line type. From the result of FIG. 11, a combination of the binarization threshold “95” and the angle “135 degrees” in which the absolute value of the correlation coefficient is maximized is detected. It is possible to obtain a linear component intensity indicating that the absolute value of the relation number is about 0.7). In this embodiment, the combination of the binarization threshold value and the angle at which the absolute value of the correlation coefficient is maximized is detected as described above, but the combination in which the absolute value of the correlation coefficient indicates the periphery of the maximum value is used. Also good. For example, the binarization threshold value may be set in a range of 63 to 127, and the angle may be set in a range of 130 degrees to 140 degrees. Further, although the optimum value of the binarization threshold in this embodiment is 95, the brightness of the linear texture image changes when the shooting conditions such as the brightness of the illumination, the shutter speed, and the aperture change. Since the optimum value of the binarization threshold also changes, the binarization threshold changes according to the shooting conditions. The reason why the correlation coefficient is negative in FIG. 11 is that the linear component intensity is represented by the number of white pixels, that is, the size of the portion other than the groove. Further, as described above, since an angle of 5 degrees is used, in FIG. 11, the data for the step width is linearly interpolated.

  Thus, it was verified that the visual score can be estimated with high accuracy by the linear component intensity of one angle “135 degrees” in the environment of the present embodiment. Furthermore, the present inventors examined whether the visual score could be estimated with higher accuracy by further adding linear component intensities at other angles.

  FIG. 12 is a graph showing the relationship between the added angle and the degree-of-freedom-adjusted determination coefficient in the regression analysis in which another angle is added to the angle “135 degrees”. According to FIG. 12, the degree-of-freedom-adjusted determination coefficient is maximized at an angle “85 degrees”. Also, the degree of freedom adjusted determination coefficient is increased (0.4502 to 0) by adding the linear component intensity of angle “85 degrees” as compared with the case where only the linear component intensity of angle “135 degrees” is used. To 4607). Thereby, it was confirmed that the estimation accuracy of the visual score is improved by adding the linear component strength of the angle “85 degrees” to the linear component strength of the angle “135 degrees”.

  FIG. 13 is a graph showing the relationship between the added angle and the degree-of-freedom-adjusted determination coefficient in the regression analysis in which another angle is added to the angles “135 degrees” and “85 degrees”. According to FIG. 13, any degree-of-freedom adjusted determination coefficient is lower than when the linear component intensities of angles “135 degrees” and “85 degrees” are used (0.4607). As a result, it was confirmed that if the linear component intensities of angles “135 degrees” and “85 degrees” were used, the visual score could be estimated with sufficient accuracy without using the linear component intensities of more angles. .

From the above, the single regression equation (1) having the linear component intensity at the angle “135 degrees” as the explanatory variable and the visual score as the objective variable, and the two linear component intensities at the angles “135 degrees” and “85 degrees” are obtained. The multiple regression equation (2) using the visual score as the explanatory variable as the explanatory variable was obtained.
S = −0.00002543 · X ₁₃₅ +8.652 (1)
S = −0.00003068 · X ₁₃₅ + 0.000009301 · X ₈₅ +7.484 (2)

FIG. 14 is a graph showing the relationship between the estimated visual score S calculated by the single regression equation (1) and the visual score obtained by sensory evaluation. FIG. 15 is a graph showing the relationship between the estimated visual score S calculated by the multiple regression equation (2) and the visual score obtained by sensory evaluation. 14 and 15, the horizontal axis indicates the estimated visual score S, and the vertical axis indicates the visual score. 14 and 15, it was confirmed that there is a correlation between the estimated visual score obtained from the regression equation and the visual score obtained by sensory evaluation. Further, the coefficient of determination indicated by FIGS. 14 and 15 (R ^2), that a high visual score better is the accuracy of the regression equation than the single regression equation (1) (2) is estimated was confirmed.

  In Example 1 described above, the number of white pixels in the binarized image is used as the linear component intensity. Hereinafter, an example in which the ratio of the number of black pixels to the total number of pixels is used as the linear component intensity will be described as a second embodiment. However, the linear component intensity of the present invention may be the number of black pixels, or the ratio of the number of white pixels to the total number of pixels.

The following regression equations (3) and (4) were obtained in Example 2 by using the same method and environment as in Example 1 except for the method of calculating the linear component intensity. And according to each regression formula (3) and (4), it was confirmed that a visual score can be estimated with the substantially the same precision as Example 1. FIG.
S = 7.693 · X ₁₃₅ +0.959 (3)
S = 9.281 · X ₁₃₅ -2.814 · X ₈₅ +1.017 (4)

In Example 2, the validity of the wrinkle state analysis method in each of the above-described embodiments is further targeted for the analysis target parts of two sample providers # 1 and # 2 that have the same visual score in the sensory evaluation. Verified. In this verification, regression equations (3) and (4) were used.
FIG. 16 is a diagram illustrating skin luminance images of the analysis target parts of the sample providers # 1 and # 2. Compared to sample provider # 1, the surface to be analyzed of sample provider # 2 has more irregular shapes on the skin surface, but according to sensory evaluation by a professional evaluator, the state of “small wrinkles” Both had the same visual score.

  FIG. 17 is a table comparing the linear component intensity and the estimated visual score obtained from the skin luminance images of the sample providers # 1 and # 2. As shown in FIG. 17, the difference between the estimated visual score calculated from two images having the same visual score (2.5) and the actual visual score (2.5) is only when 135 degrees is used ( -0.18), it becomes smaller when using two angles of 135 degrees and 85 degrees (-0.04), and according to the wrinkle state analysis method in each of the embodiments described above, the sensory evaluation It was verified that an estimated visual score similar to the visual score can be obtained.

  FIG. 18 is a diagram illustrating an example in which linear texture images of 135 degrees and 85 degrees for sample providers # 1 and # 2 are output side by side. In FIG. 18, the binarized image is output as a linear texture image. From the output of FIG. 18, it is visually recognized that the sample provider # 2 has a greater linear component intensity at both angles than the sample provider # 1. On the other hand, both the visual score and the estimated visual score are the same or approximate. That is, it is estimated that information close to the visually recognized wrinkle state is extracted by canceling the linear component of 135 degrees having a high correlation with the visual score by the other linear component of 85 degrees. This is connected to the fact that the sign of the 85 degree coefficient is inverted from the 135 degree coefficient in the multiple regression equations (2) and (4).

  FIG. 19 is a diagram illustrating an output example of a superimposed image of 135 degree and 85 degree linear texture images for sample providers # 1 and # 2. The upper two images are 135 degree linear texture images extracted from the skin luminance image by applying the Gabor filter. The lower two images are obtained by applying a Gabor filter and binarization processing to a 135 degree linear texture image extracted from a skin luminance image by applying the Gabor filter. This is an image on which a (binarized image) is superimposed.

  In each of the above-described examples, the cheek (lower orbit) portion as an analysis target portion was used, and the visual score mainly reflected a “small wrinkle” state depending on the dry state of the epidermis. Therefore, it can be said that the estimated visual score obtained in each of the above-described embodiments is an index value indicating the visual impression of the state of “small wrinkles”. As described above, “wrinkles” may be classified into a plurality of types depending on the cause of occurrence, such as aging, dryness, and decrease in dermal function. Therefore, in the sensory evaluation of professional evaluators, other types of wrinkles By evaluating to the center, it is also possible to calculate an estimated visual score indicating the visual impression of the type of wrinkles other than “small wrinkles”. For example, if a visual score that mainly reflects the state of `` dermis wrinkles '' caused by a decrease in dermal function is obtained by sensory evaluation, an estimated visual score indicating the visual impression of the `` dermis wrinkle '' state can also be calculated. Is possible.

  Thereby, a different regression equation is prepared for each wrinkle type in advance, and an estimated visual score representing each type of wrinkle state is calculated by switching the regression equation according to the type of wrinkle to be analyzed. It is also possible to adopt. For example, when counseling a skincare product that restores dryness of the epidermis, an estimated visual score indicating the appearance of “small wrinkles” is calculated. A form of calculating the score is conceivable.

  In each of the above-described examples, the left cheek (lower orbit) is the analysis target site. This is because “small wrinkles” that depend on the dry state of the epidermis are the main analysis objects, and the lower part of the orbit is easy to dry, and it is easy to make a difference for each person. Therefore, the right cheek (lower eye socket) may be the analysis target site. In this case, for example, 225 degrees is used instead of 135 degrees, and 275 degrees is used instead of 85 degrees. Similarly to the above, this angle notation is a notation in which the vertical reverse direction is 0 degrees, the vertical direction is 180 degrees, and increases clockwise toward the face. Further, since 225 degrees and 275 degrees are angles when the face is considered to be bilaterally symmetric, each angle may actually be slightly shifted. As can be seen from the Gabor filter equation (Equation 1), the shape of the filter is the same even if the angle θ is shifted by 180 degrees. For example, the same effect can be obtained by using 45 degrees instead of 225 degrees. Is obtained.

  Moreover, an estimated visual score may be calculated for each of the left and right cheeks. In this case, in the line detection step (S22), the analysis apparatus 10 specifies one or more predetermined angles corresponding to the analysis target part from a plurality of use angle information determined in advance for the plurality of analysis target parts. To do. That is, the analysis device 10 specifies 135 degrees, or 135 degrees and 85 degrees when the left cheek is the analysis target part, and 225 degrees or 225 degrees when the right cheek is the analysis target part. And 275 degrees. Further, one analysis target part may be selected from a plurality of different analysis target parts such as around the eyes, cheeks, around the mouth, neck, corners of the eyes, and forehead. Also in this case, since the direction (angle) that affects the impression of the appearance of the wrinkled state is different for each part, the use angle information may be determined in advance for each analysis target part. For example, when the forehead is the analysis target part, one or more predetermined angles including the horizontal angle are specified, and when the neck is the analysis target part, one or more including the vertical angle is specified. A predetermined angle is specified. Then, in the calculation step (S24), the analysis apparatus 10 identifies a regression equation corresponding to the analysis target part from a plurality of regression equations determined in advance for the plurality of analysis target parts, and the identified regression An estimated visual score is calculated using the formula.

  In addition, in the some flowchart used by the above-mentioned description, although several process (process) is described in order, the execution order of the process performed by each embodiment is not restrict | limited to the order of the description. In each embodiment, the order of the illustrated steps can be changed within a range that does not hinder the contents. Each process may include contents executed by a person. For example, the skin luminance image acquired in (S21) may be read by the analysis device 10 from a file selected by a human operation on the analysis device 10. In (S82), when the analysis target part is cut out from the acquired skin luminance image, the position of the analysis target part may be designated by a human operation on the analysis apparatus 10. Moreover, each above-mentioned embodiment can be combined in the range in which the content does not conflict.

  Part or all of the above contents can also be specified as follows. However, the above-mentioned content is not limited to the following description.

<1> An image acquisition step of acquiring a skin luminance image of the analysis target part;
A line detection step of extracting one or more linear texture images for one or more predetermined angles corresponding to the analysis target site from the skin luminance image acquired in the image acquisition step;
An intensity extraction step of extracting the linear component intensity of the one or more predetermined angles from the one or more linear texture images extracted in the line detection step;
Obtained by regression analysis on a plurality of sample face images, the relational expression between the linear component intensity at one or more predetermined angles and the visual score of wrinkles on the face image, and extracted in the intensity extraction step Using the linear component intensity, a calculation step of calculating an estimated visual score of skin wrinkles of the analysis target part;
Wrinkle state analysis method including

<2> The relational expression is a single regression equation having the linear component intensity of the first predetermined angle having the highest correlation with the visual score as an explanatory variable and the visual score as an objective variable, or the first predetermined angle. It is a multiple regression equation with the linear component intensities at two predetermined angles including the explanatory variable and the visual score as the objective variable.
The wrinkle state analysis method according to <1>.
<3> An information acquisition step of acquiring age information of the skin subject from which the estimated visual score is calculated in the calculation step;
An output step of outputting a representative visual score corresponding to the age information acquired in the information acquisition step and the estimated visual score calculated in the calculation step;
The wrinkle state analysis method according to <1> or <2>, further comprising:
<4> An information acquisition step of acquiring age information of the skin subject from which the estimated visual score is calculated in the calculation step;
An output that outputs the representative linear texture image corresponding to the age information acquired in the information acquisition step and the linear texture image extracted in the line detection step side by side for each predetermined angle corresponding to the analysis target region Process,
The wrinkle state analysis method according to any one of <1> to <3>, further including:
<5> An output step of outputting an image in which a plurality of linear texture images extracted in the line detection step are superimposed and linear components are represented in different colors for each image,
The wrinkle state analysis method according to any one of <1> to <4>, further including:
<6> The intensity extraction step binarizes the one or more linear texture images extracted in the line detection step using a predetermined binarization threshold, and a binary value obtained by binarization processing Extracting the linear component intensity of the one or more predetermined angles using the area of one value of the digitized image;
The wrinkle state analysis method according to any one of <1> to <5>.
<7> The line detection step applies the one or more linear texture images to the skin luminance image by applying a Gabor filter corresponding to the one or more predetermined angles corresponding to the analysis target region. Extract,
The wrinkle state analysis method according to any one of <1> to <6>.
<8> An image processing step of acquiring a surface reflected light image from each polarized image obtained by projecting S-polarized light on the subject's skin and imaging S-polarized light and P-polarized light;
A normalization step of extracting a normalized skin image of the analysis target site from the surface reflected light image;
Further including
The image acquisition step acquires the skin luminance image by deleting color information from the skin image.
The wrinkle state analysis method according to any one of <1> to <7>.
<9> The analysis target site is any one of cheeks, eye circumferences, mouth circumferences, and neck muscles, in which fine wrinkles are likely to occur.
The wrinkle state analysis method according to any one of <1> to <8>.
<10> The line detection step includes specifying the one or more predetermined angles corresponding to the analysis target part from a plurality of use angle information predetermined for a plurality of analysis target parts,
In the calculation step, the analysis target is selected from a plurality of relational expressions between a linear component intensity at one or more predetermined angles and a visual score of wrinkles on a face image, which are predetermined for the plurality of analysis target parts. Identifying the relational expression corresponding to the site,
The wrinkle state analysis method according to any one of <1> to <9>.
<11> The analysis target site is a left cheek or a right cheek,
When expressing the one or more predetermined angles so that the vertical reverse direction is 0 degree, the vertical direction is 180 degrees, and increases clockwise toward the face,
When the analysis target part is a left cheek, the one or more predetermined angles corresponding to the analysis target part are 135 degrees, or 135 degrees and 85 degrees,
When the analysis target part is a right cheek, the one or more predetermined angles corresponding to the analysis target part are 225 degrees, or 225 degrees and 275 degrees.
The wrinkle state analysis method according to any one of <1> to <10>.
<12> Image acquisition means for acquiring a skin luminance image of the site to be analyzed;
Line detection means for extracting one or more linear texture images for one or more predetermined angles corresponding to the analysis target site from the skin luminance image acquired by the image acquisition means;
Intensity extracting means for extracting the linear component intensity at the one or more predetermined angles from the one or more linear texture images extracted by the line detecting means;
Obtained by regression analysis for a plurality of sample face images, the relational expression between the linear component intensity at one or more predetermined angles and the visual score of wrinkles on the face image, and extracted by the intensity extracting means A calculation means for calculating an estimated visual score of skin wrinkles of the analysis target site using the linear component intensity;
A wrinkle state analyzer.
<13> The relational expression is a single regression equation having the linear component intensity of the first predetermined angle having the highest correlation with the visual score as an explanatory variable and the visual score as an objective variable, or the first predetermined angle. It is a multiple regression equation with the linear component intensities at two predetermined angles including the explanatory variable and the visual score as the objective variable.
The wrinkle state analyzer according to <12>.
<14> Information acquisition means for acquiring age information of the skin subject from which the estimated visual score is calculated;
Output processing means for outputting a representative visual score corresponding to the age information acquired by the information acquisition means and the estimated visual score calculated by the calculation means;
The wrinkle state analyzer according to <12> or <13>, further comprising:
<15> Information acquisition means for acquiring age information of the skin subject from which the estimated visual score is calculated;
An output that outputs the representative linear texture image corresponding to the age information acquired by the information acquisition unit and the linear texture image extracted by the line detection unit side by side for each predetermined angle corresponding to the analysis target region Processing means;
The wrinkle state analyzer according to any one of <12> to <14>, further comprising:
<16> Output processing means for outputting a plurality of linear texture images extracted by the line detection means, and outputting an image in which linear components are represented in different colors for each image;
The wrinkle state analyzer according to any one of <12> to <15>, further comprising:
<17> Image processing means for acquiring a surface reflected light image from each polarized image obtained by projecting S-polarized light onto the subject's skin and imaging S-polarized light and P-polarized light;
Normalization means for extracting a normalized skin image of the analysis target site from the surface reflected light image;
Further comprising
The image acquisition means acquires the skin luminance image by deleting color information from the skin image.
The wrinkle state analyzer according to any one of <12> to <16>.
<18> The line detection unit specifies the one or more predetermined angles corresponding to the analysis target part from a plurality of use angle information predetermined for a plurality of analysis target parts,
The calculation means determines the analysis target from a plurality of relational expressions between a linear component intensity at one or more predetermined angles and a visual score of wrinkles on a face image, which are predetermined for the plurality of analysis target parts. Identifying the relational expression corresponding to the site;
The wrinkle state analyzer according to any one of <12> to <17>.
<19> A program that causes at least one computer to execute the wrinkle state analysis method according to any one of <1> to <11>.

1 CPU
2 Memory 5 Output device 6 Input device 10 Wrinkle state analyzer (analyzer)
DESCRIPTION OF SYMBOLS 11 Image acquisition part 12 Line | wire detection part 13 Intensity extraction part 14 Calculation part 16 Information acquisition part 17 Output processing part 18 Image processing part 19 Normalization part

Claims (13)

An image acquisition step of acquiring a skin luminance image of the analysis target part;
A line detection step of extracting one or more linear texture images for one or more predetermined angles corresponding to the analysis target site from the skin luminance image acquired in the image acquisition step;
An intensity extraction step of extracting the linear component intensity of the one or more predetermined angles from the one or more linear texture images extracted in the line detection step;
Obtained by regression analysis on a plurality of sample face images, the relational expression between the linear component intensity at one or more predetermined angles and the visual score of wrinkles on the face image, and extracted in the intensity extraction step Using the linear component intensity, a calculation step of calculating an estimated visual score of skin wrinkles of the analysis target part;
Wrinkle state analysis method including The relational expression is a single regression equation having the linear component intensity of the first predetermined angle having the highest correlation with the visual score as an explanatory variable and the visual score as an objective variable, or two including the first predetermined angle. It is a multiple regression equation with the linear component intensity at a predetermined angle as an explanatory variable and the visual score as an objective variable.
The wrinkle state analysis method according to claim 1. An information acquisition step of acquiring age information of the skin subject for which the estimated visual score was calculated in the calculation step;
An output step of outputting a representative visual score corresponding to the age information acquired in the information acquisition step and the estimated visual score calculated in the calculation step;
The wrinkle state analysis method according to claim 1 or 2, further comprising: An information acquisition step of acquiring age information of the skin subject for which the estimated visual score was calculated in the calculation step;
An output that outputs the representative linear texture image corresponding to the age information acquired in the information acquisition step and the linear texture image extracted in the line detection step side by side for each predetermined angle corresponding to the analysis target region Process,
The wrinkle state analysis method according to claim 1, further comprising: An output step of outputting an image in which a plurality of linear texture images extracted in the line detection step are superimposed and linear components are represented in different colors for each image;
The wrinkle state analysis method according to any one of claims 1 to 4, further comprising: In the intensity extraction step, the one or more linear texture images extracted in the line detection step are binarized using a predetermined binarization threshold, and a binarized image obtained by the binarization processing is processed. Extracting the linear component intensity of the one or more predetermined angles using the area of one value,
The wrinkle state analysis method according to any one of claims 1 to 5. The line detection step extracts the one or more linear texture images by applying a Gabor filter corresponding to the one or more predetermined angles corresponding to the analysis target site to the skin luminance image.
The wrinkle state analysis method according to any one of claims 1 to 6. An image processing step of acquiring a surface reflected light image from each polarization image obtained by projecting S-polarized light on the subject's skin and imaging S-polarized light and P-polarized light;
A normalization step of extracting a normalized skin image of the analysis target site from the surface reflected light image;
Further including
The image acquisition step acquires the skin luminance image by deleting color information from the skin image.
The wrinkle state analysis method according to any one of claims 1 to 7. The site to be analyzed is any one of cheeks, around the eyes, around the mouth, and the neck, where wrinkles are likely to occur.
The wrinkle state analysis method according to any one of claims 1 to 8. The line detection step includes specifying the one or more predetermined angles corresponding to the analysis target part from a plurality of use angle information predetermined for a plurality of analysis target parts,
In the calculation step, the analysis target is selected from a plurality of relational expressions between a linear component intensity at one or more predetermined angles and a visual score of wrinkles on a face image, which are predetermined for the plurality of analysis target parts. Identifying the relational expression corresponding to the site,
The wrinkle state analysis method according to any one of claims 1 to 9. The analysis target site is a left cheek or a right cheek,
When expressing the one or more predetermined angles so that the vertical reverse direction is 0 degree, the vertical direction is 180 degrees, and increases clockwise toward the face,
When the analysis target part is a left cheek, the one or more predetermined angles corresponding to the analysis target part are 135 degrees, or 135 degrees and 85 degrees,
When the analysis target part is a right cheek, the one or more predetermined angles corresponding to the analysis target part are 225 degrees, or 225 degrees and 275 degrees,
The wrinkle state analysis method according to any one of claims 1 to 10. Image acquisition means for acquiring a skin luminance image of the analysis target part;
Line detection means for extracting one or more linear texture images for one or more predetermined angles corresponding to the analysis target site from the skin luminance image acquired by the image acquisition means;
Intensity extracting means for extracting the linear component intensity at the one or more predetermined angles from the one or more linear texture images extracted by the line detecting means;
Obtained by regression analysis for a plurality of sample face images, the relational expression between the linear component intensity at one or more predetermined angles and the visual score of wrinkles on the face image, and extracted by the intensity extracting means A calculation means for calculating an estimated visual score of skin wrinkles of the analysis target site using the linear component intensity;
A wrinkle state analyzer.   The program which makes at least one computer perform the wrinkle state analysis method of any one of Claim 1 to 11.

Images