JPH11332834A - Surface condition analysis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To numerically analyze the degree of pigmentation of a skin surface in relation with a constitutional factor thereof. SOLUTION: Points of main components in a main component analysis are calculated for pixels from luminance of the pixels of an image taken at 400 nm, 550 nm and 700 to reconstruct the image expressing the points of the pixels by the luminance so that the distribution of melanin is clarified as existing around a base bottom layer and the upper layer thereof. A specified region of interest is determined in a reconstructed image and the points of the main components contained in the region are calculated for a frequency distribution of the points within the region to determine a profile of coloration.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for analyzing a surface state of an object, and more particularly to a method for numerically analyzing the degree of pigmentation on the skin surface.

[0002]

2. Description of the Related Art R, G, B for each pixel of a pigmented image obtained by a CCD camera with a polarizing plate are represented by L ^* , a ^* , b.
^* To convert the degree of pigmentation into a numerical value (Yuko Shibue et al., The 7th Color Engineering Conference, 3-
3, 1990 pp. 55-58) and a method of expressing the color depth of pigmentation by focusing on the value of G (Noriko Hara et al., No. 3)
Proceedings of the 1st SCCJ Research Symposium, 1991, pp.
36-42) have been reported. Evaluation of pigmentation using the total average value of the color difference dE for each pixel of the original image obtained through the TV camera and the image obtained by smoothing the original image (Yuji Taji et al., Cosmetic Approval, 28 , 1994, pp. .147-15
2) and digitization focusing on an ultraviolet image of a pigmented portion (Seiichi Arai et al., Cosmetic Approval, 23 , 1989, pp. 31-42) have also been attempted.

[0003] The applicant of the present invention has disclosed in
No. 9, the main wavelengths of 400 nm, 550 nm, 700
From the combination of the binary images of the spectral images of nm, a melanin image having a different location in the depth direction from the skin surface is extracted, and the color depth and the pigmentation area of the pigmented portion are quantified.

[0004]

However, in each of the above-mentioned methods, the density of the pixel is numerically converted into the color density of the pigment-deposited portion without clearly specifying the constituent factors of the color of the pigment-deposited portion. Originally, a color image formed of only melanin should be reconstructed and subjected to digitization.

Accordingly, it is an object of the present invention to specify a color component of a pigmented portion and reconstruct a color image formed only of melanin to thereby determine the degree of pigmentation on the skin surface in relation to the component. It is an object of the present invention to provide a method that enables numerical analysis.

[0006]

According to the present invention, the luminance of a pixel in a plurality of wavelength bands is determined for each pixel constituting an image of an object by generating an image of the same object in a plurality of wavelength bands. Then, for each pixel, the component score of the main component contained in the spectrum of the reflected light from the object is calculated from the brightness of the pixel in the plurality of wavelength bands, and an image in which the component score of each pixel is represented by the brightness is regenerated. A surface state analysis method including the constituent steps is provided.

Preferably, the method further comprises a step of calculating a component score of a main component included in the frequency distribution from a luminance frequency distribution in a specific region of interest in the reconstructed image. It is.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION
7, pp. 44-51 ", obtained by applying principal component analysis using" Nama no sum of squares and sum of products matrices ", which is one of the multivariate analysis methods, to the spectral reflectance data of the skin of a large number of subjects. The component score M ₁ of the first principal component reflects the degree of coloration due to melanin present around the basal layer, and the component score M ₂ of the second principal component reflects the degree of coloration due to melanin above the basal layer. I've made it clear.

That is, the eigenvectors V ₁ (λ), V ₂ (λ), and V ₃ (λ) of the “raw sum-of-squares / sum-of-products matrix” relating to the spectral reflectances R (λ) of the skin of many subjects. when representing the spectral reflectance of the skin of any subject R a (lambda) and _{R (λ) = M 1 V} 1 (λ) + M 2 V 2 (λ) + M 3 V 3 (λ) (1), M 1 is reflecting the degree of coloration with melanin present around the basal layer, M ₂ became clear that reflects the color degree by melanin in the layers above the basal layer.

In the present invention, the two-dimensional distribution of melanin existing in the basal layer and the upper layer is first clarified by applying this method to each of the pixels constituting the image of the skin. That is, for example, as described in
The use of the trispectral camera disclosed in Japanese Patent No. 9839 makes it possible to obtain reflected light images at λ = 400 nm, 550 nm, and 700 nm. R (λ) (λ = 400, 550, 7) converted from each pixel value constituting this image
00) and the eigenvectors V ₁ (λ), V ₂ (λ), V
₃ By solving the ternary system obtained by substituting the values of λ = 400 nm, 550 nm, and 700 nm of (λ) into the equation (1), the values of M ₁ and M ₂ can be obtained for each pixel.

First, the eigenvectors V ₁ (λ), V
To determine ₂ (λ), V ₃ (λ), 75 whites and 12 blacks, measured by Hitachi Color Analyzer 607,
From the spectral reflectance data of the cheeks of a total of 150 women (20's to 50's) of 63 yellow people (Japan and Southeast Asia), spectral reflectance data at 31 wavelength points at 400 nm to 700 nm, at 10 nm intervals, are extracted between wavelengths. The eigenvalues and cumulative contributions of the raw sum of squares and sum of products matrix were calculated. Table 1 shows the obtained eigenvalues and cumulative contribution ratios, and FIG. 1 shows the eigenvectors. Table 2 shows values of eigenvectors at 400 nm, 550 nm, and 700 nm.

[0012]

[Table 1]

[0013]

[Table 2]

In view of the cumulative contribution, the skin spectral reflectance R
The magnitude of (λ) can be almost completely described by equation (1). Therefore, if skin spectral reflectances of 400 nm, 550 nm, and 700 nm are obtained, M ₁ , M ₂ , and M ₃ can be calculated by solving simultaneous equations. When the pigmented part was photographed with a prototype trispectral camera, the main wavelength was 400n.
The luminance I ₂ of each pixel of the m, 550 nm, and 700 nm spectral images is obtained.

A method of converting the luminance I ₂ of each pixel into a spectral reflectance R ₂ using a luminance calibration tile described in Japanese Patent Application Laid-Open No. 7-19839 will be described with an example of a main wavelength of 400 nm. Brightness calibration tile GRAY measured in advance
NO. 1000, GRAYNO. When the spectral reflectance of 3000 is 7.4% and 27.0%, respectively, and the luminance at the time of calibration is 40 and 250, respectively, the luminance I ₂ of each pixel and the spectral reflectance are calculated using these values. Formula (2) representing the relationship with R ₂ is created.

[0016]

(Equation 1)

By substituting I ₂ = 250 and R ₂ = 27.0 into equation (2) to obtain α, equation (3) is formed. Since then
The luminance I ₂ of each pixel of the obtained spectral image is substituted into Expression (3), and the spectral reflectance R ₂ for each pixel is obtained.

[0018]

(Equation 2)

Individual spectral images at main wavelengths of 550 nm and 700 nm
For each pixel, a conversion equation was similarly obtained. here
In the spectral image of main wavelength 400nm, 550nm, 700nm
The spectral reflectance of each pixel in R₄₀₀, R
₅₅₀, R₇₀₀And this value was set to 400 nm, 5 shown in Table 2.
Equation using the values of eigenvectors at 50 nm and 700 nm
Substituting into (4), the coloration of the pixel by melanin
Corresponding component score M₁, M _TwoCan be calculated.

R₄₀₀= 0.0806M₁+ 0.1347M_Two+ 0.2897M_Three R₅₅₀= 0.1333M₁+ 0.2054M_Two-0.2238M_Three (4) R₇₀₀= 0.2757M₁-0.3031M_Two-0.1417M_Three Skin spectral reflectance of 150 women measured this time
M obtained from₁, M _TwoAre 265.
3, 26.5, and the minimum values are 91.2, -19.5.
Was.

[0021] divided into a value of 256 gradations of M ₁ about 265.3 to 91.2, with a numerical value that represents a color by melanin, and the luminance for visual display by the image. That is, the M ₁ derived from the pixel into Equation (5) to determine the intensity I _M1 corresponding to M ₁ of the pixel.

[0022]

(Equation 3)

Similarly, M ₂ was substituted into equation (6), and the luminance I _M2 corresponding to M ₂ of the pixel was calculated.

[0024]

(Equation 4)

The above-described processing is performed for each pixel, and an M ₁ image reflecting the degree of coloration due to melanin present around the basal layer is obtained from I _M1 by using melanin above the basal layer from I _M2. It was reconstituted M ₂ image reflecting the color degree. For example, rather Aoguroi pigmentation portion that is seen in the color photograph, but its presence in the M ₁ image can be confirmed,
Not seen the evidence in the M ₂ image. This is thought to be because the melanin that forms this pigmentation is relatively deep from the skin surface.

Next, to quantify the degree of coloration, a region of interest is set in an arbitrary region of the photographed image. For example, 1
A square composed of 92 × 192 pixels is defined as the size of the region of interest. Hereinafter referred methods for quantifying the degree of coloration with melanin present around the basal layer of M ₁ images. Since the image of M ₁ is created using the luminance I _M1 corresponding to the component score M ₁ of each pixel, the luminance I _{M1 of} each pixel is used as a parameter for quantifying the degree of coloration.
Is used.

When measuring the spectral reflectance of the skin with a spectrophotometer, the average spectral reflectance within the area of the skin pressed against the opening of the integrating sphere is measured. Therefore, similarly to the measurement by the spectrophotometer, the luminance I _{M1, j} of all pixels j in the region of interest is used.
Is calculated by the equation (7) to obtain an average pigmentation density m ₁ . In this case The problem is also the same as the value of the average pigmentation density m ₁ by melanin, the frequency distribution of the individual luminance I _M1 is the different cases exist.

Therefore, the frequency distribution curve of the luminance I _M1 is quantified and displayed together with the melanin concentration m ₁ to express a more precise coloration state due to melanin. The method is described below. The luminance I _M1 in the region of interest of the 250 sets of three-wavelength spectral images taken of the pigmented part of the cheeks of Japanese women (20s to 50s) is divided into 10 ranks, and the average of the frequency for each rank FIG. 2 is obtained by obtaining the values and writing the average frequency distribution. Table 3 shows details of the ranking.

In this way, if the pigmentation is thin as a whole, a frequency distribution curve with a large slope and a rightward slope is obtained, and if the pigmentation as a whole is dark, a curve component ascending to the left is emphasized. In fact, in addition to these large trends, the reflection of intermediate concentrations is also folded into the form, forming the final frequency distribution. Therefore, if the form of the frequency distribution in FIG.
A numerical value is obtained that describes the structure of the color depth that is visually perceived. The inventors of the present application have already reported a method of digitizing such a frequency distribution curve (Yukiko Kawaguchi, Cosmetic Sci., 28 , 1994, pp. 153-162). Adopted as a method for. That is, for a color image formed by melanin reconstructed from the obtained spectral image including 250 sets of pigmented portions, a sum of squares and a product between frequencies related to the frequency of the frequency distribution of the luminance I _M1 corresponding to FIG. Eigenvalue analysis of the union matrix was performed. Table 3 shows the ranking of the luminance and the obtained eigenvalues, eigenvectors, and cumulative contribution ratios.

[0030]

[Table 3]

Taking up the first to fifth principal components,
Almost 100% of the frequency distribution of luminance I _M1 for each pixel
Can be explained. The first main component corresponds to a quantitative difference in a portion where coloration due to melanin is light, the second main component corresponds to a quantitative difference in a dark portion, and the third main component corresponds to a quantitative difference in a medium coloration portion. As a result, 98% of the frequency distribution can be expressed by the first to third principal components. The remaining 2% is the fourth to fifth
This is explained by the quantitative difference in the frequency distribution corresponding to the eigenvectors related to the principal components.

Hereinafter, the degree of coloration by melanin represented by the first to fifth main components will be represented by the component scores Z _M11 , Z _M12 , Z _M13 , Z _M14 and Z _M15 . The obtained Z _M11 , Z _M12 , Z _M13 , Z _M14 and Z _M15 are 0
The dynamic range is expanded so that a value between 100 and 100 can be displayed. Table 4 shows the maximum and minimum component scores of the first to fifth principal components.

[0033]

[Table 4]

In general, the maximum value of the i-th principal component score is Z
_{M1i, max,} and the minimum value Z _{M1i, min,} and when the component score of the image and Z _M1i, component scores after the dynamic range expansion Z _{M1 i, D} is obtained from equation (8).

[0035]

(Equation 5)

Using the previously obtained m ₁ and Z _{M1i, D} for the first to fifth main components _, the profile of FIG. 3 is drawn to show the pigmentation state. FIG. 3 shows a profile relating to coloration due to melanin existing around the basal layer. For even coloration with melanin present in the upper layer than the base layer, using the luminance I _M2 obtained by the formula (6), the average pigmentation concentration m ₂ and Z for profile creation
_{M21, D} , _{ZM22, D} , _{ZM23, D} , _{ZM24, D} , _{ZM25, D} were obtained. FIG. 4 shows the results. In this case, as shown in Table 5,
The frequency distribution of the luminance I _M2 that can be explained by the first to fifth principal components is 99.94%.

Between the frequencies of the frequency distribution of I _M2 ,
FIG. 5 shows eigenvectors obtained from the product-sum matrix.

[0038]

[Table 5]

As described above, the usefulness of the pigmentation-compatible cosmetics can be obtained by reconstructing and displaying a color image by melanin having different relative depths of the existing places, quantifying the coloration, and profiling the pigmentation concentration state. In addition, an increase in pigmentation with age can be revealed. next,
For skin that has been tanned with an artificial light source, is it possible to reconstruct a color image using melanin with different relative depths and quantify the pigmentation density to evaluate the recovery of pigmentation? We verified whether or not. Table 6 shows the conditions of the sunburn test.

[0040]

[Table 6]

FIG. 6 shows the color profile of melanin present around the basal layer composed of the values obtained from Equations (6) and (7). FIG. 7 shows the color profile of melanin above the basal layer. Indicates a profile. The solid line illustrates the state before sunburn, the dotted line illustrates the state after one week of sunburn, and the dashed line illustrates the state after one month. As a result, a state in which the coloration due to melanin decreases in the order of one week after sunburn, one month after sunburn, and before sunburn is reflected in the image. Also, in a profile created using a numerical index, a reduction in coloration can be seen.

In this case, because of temporary pigmentation due to sunburn, it is understood from images and numerical values that both melanin existing around the basal layer and melanin existing above the basal layer are reduced. You need to be able to do it, but the results meet that requirement. The above suggests that the method proposed in the present application is sufficiently practical for the evaluation of recovery of pigmentation.

[0043]

As described above, according to the present invention, the degree of pigmentation on the skin surface can be numerically analyzed in relation to its constituent factors.

[Brief description of the drawings]

FIG. 1 is a diagram showing eigenvectors of a raw sum of squares and sum of products matrix between wavelengths calculated from spectral reflectance data.

FIG. 2 is a graph showing an average frequency distribution of luminance I _M1 corresponding to a score M _{1 of a} first principal component.

FIG. 3 is a diagram showing a profile of coloration due to melanin present around the basal layer.

FIG. 4 is a view showing a color profile of melanin present above a basal layer.

FIG. 5 is a diagram showing eigenvectors obtained from a frequency distribution of luminance I _M2 .

FIG. 6 is a diagram showing a profile of coloration due to melanin present around the basal layer.

FIG. 7 is a diagram showing a profile of coloration due to melanin present above a basal layer.

Claims (2)

[Claims] An image of a plurality of wavelength bands is generated for the same object to determine the luminance of the pixels in the plurality of wavelength bands for each pixel constituting the image of the object. A surface state including each step of calculating a component score of a main component included in a spectrum of reflected light from an object from luminance of a pixel in a wavelength band, and reconstructing an image in which the component score of each pixel is represented by luminance. analysis method. 2. The method according to claim 1, further comprising: calculating a component score of a principal component included in the frequency distribution from a frequency distribution of luminance in a specific region of interest in the reconstructed image. .