KR20090055172A - Method for discriminating thick and thin coating using tongue photo - Google Patents

Abstract

A method for discriminating thick and thin coating using a tongue photo are provided to determine a thick coating and a thin coating based on an objective discrimination function without a light condition and an environmental condition. A brightness correction image and chroma correction image are obtained by correcting the brightness and chroma of an original image(s100). A color value on a maximum horizontal axis is extracted form the correction image, and a critical value is set up based on the extracted color(s200). The tongue coating and tongue body are divided into an image region based on threshold value of the correction image(S300).

Description

Method for Discriminating Thick and Thin Coating Using Tongue Photo}

The present invention relates to a method for classifying larynx and gestation using images, using a tongue image that classifies lingual status as posteriority and gestation by analyzing a planar tongue image without being affected by lighting conditions or surroundings. The present invention relates to a method of judging postpartum condition.

In Chinese medicine, the tongue reflects the physiology and pathology of the human body. It has several physiological functions related to other human organs. In particular, the tongue is said to objectively reflect various physiological and pathological changes in the human body. This part of the human body forms a unified whole through the meridian bleeding that runs the meridians. When a disease occurs, it must affect the normal operation of the bleeding, causing a change in the local area. As a viscera, any abnormal changes in blood donor fluid appear on the tongue through meridian conduction, which is based on oriental medicine theory that it is sensitive to changes and convenient to observe.

Like this, the tongue, which is considered as a mirror that reflects internal organ change in Chinese medicine, is diagnosed mainly by visual information, and the condition of the book is diagnosed using its color, shape, movement, tongue quality and tongue. The moss-covered material on the tongue is used as an important diagnostic factor because it can mask the nature of the scam, the invasion site and the presence or absence of essence through its color, degree of wetting, thickness, shape and coverage. have.

However, despite being a non-invasive and simple diagnostic method, tongue diagnosis is not widely used due to problems of quantification and standardization. Changes in the diagnostic environment, such as the light source, have a significant effect on the diagnosis result, and it is difficult to obtain objective and reproducible results because it is based on the experience and knowledge of the diagnosis person. Recently, with increasing interest in non-invasive diagnostic methods, various studies have been conducted to solve these problems.

Regarding the acquisition of tongue images, in 1986 China raised the issue of false conclusions not only by light sources, temperature, but also by the psychological factors of the physician, and to solve this problem, how to project and shoot light evenly on the tongue under standard light conditions. After that, researches on obtaining more accurate images-how to control the lighting angle and brightness, the attitude of the measurer, how to control the image acquisition environment, the aperture of the camera, the shutter, and the exposure adjustment method Progressed. In addition, research has been conducted on a method of detecting a tongue region necessary for diagnosis from the obtained image. Beginning with the initial passive area detection method, a lot of researches on the automatic area detection algorithm have been conducted in recent years, and researches applying various image segmentation methods have been conducted.

Referring to the related invention, a method of extracting a region of interest from a tongue image and a health monitoring method and apparatus using the tongue image (published in Korean Patent Laid-Open Publication No. 10-2004-0059312, 2004. 07. 05 publication) acquire a tongue image and from there, 5 It extracts the area of interest of dog and compares it with the basic tongue image of the database and monitors the health status. The result of the detection is influenced by the lighting when the basic tongue image or the tongue image is taken. There is a problem that can not determine the thickness.

In addition, the device thickness measurement and vertical image acquisition device and method (Korean Patent Laid-Open Publication No. 10-2006-0053033, published on May 19, 2006) is to obtain a partial vertical image corresponding to the relative intensity of the interference light reflected from the optical probe using a light source Although the device is proposed, it is not necessary to determine the thickness of the tongue through the analysis of the given image. Therefore, it is troublesome to retake the tongue directly, and the thickness of the tongue can be determined by the planar image. There is no problem.

The present invention has been made to solve the above problems of the prior art, and is not affected by the lighting conditions or the surrounding environment at the time of image acquisition, and uses an objected discrimination function to classify images of post-mortem and gestation. The purpose of the present invention is to provide a method for discriminating postpartum condition.

According to an embodiment of the present invention, a method for determining a late condition of a human condition using a tongue image according to the present invention may be performed. Extracting the color values of the maximum horizontal axis from the original image, the brightness correction image, and the saturation correction image, and obtaining a mean and standard deviation, and setting a threshold value from the extracted color values based on the threshold value; A third step of dividing the brightness correction image and the saturation correction image into the tongue area image and the tongue area image; and the horizontal axis color at 1/4, 1/2, and 3/4 of the three tongue area images divided in the third step. In the fourth step of extracting the value and obtaining the mean and standard deviation, a discriminant function using the average or standard deviation calculated in the second and fourth steps is used as a variable. And it comprises a fifth step for determining hutae and Park 's.

The first step may include a first process of converting an original image into an HSV format image, a second process of equalizing a V value histogram of the HSV format image to obtain a brightness correction image, and an S value of the HSV format image. And a third process of obtaining a saturation correction image by stretching the histogram.

In the second step, the maximum horizontal axis is the widest part of the tongue area.

The third step includes a first process of setting a threshold, a second process of dividing an area having a color value above the threshold into a tongue area image, and a region having a color value less than the threshold value into a tongue area area image. This includes the third process.

The fourth step extracts horizontal axis color values at positions 1/4, 1/2, and 3/4 from the top of the tongue area image.

The color value and the threshold value are G values among the RGB color values.

The fifth step is to determine the posterior and the gestational state on the basis of the degree of discrimination of the discriminant function using the standard deviation of the G values calculated in the second and fourth steps as variables.

If the degree of discrimination is greater than a predetermined reference value, it is determined as a posterior condition. If it is smaller than the reference value, it is determined as a condition.

According to the present invention configured as described above, the method of distinguishing the posterior fetus using the tongue image may obtain an objective discrimination function for classifying the posterior fetus and the fetus using an image emphasizing the characteristics of the tongue, and separating the tongue and the tongue. In addition, it is possible to classify with high accuracy because it is not affected by lighting conditions and surrounding environment at the time of image acquisition, and accurate constitution and condition can be diagnosed.

Hereinafter, with reference to the accompanying drawings illustrating a specific configuration and embodiment of the present invention.

In Chinese medicine, the tongue has a thin and transparent mucosal epithelium and is rich in blood vessels and nerves so that the change can be clearly seen, which is considered a very clear sign reflecting various changes in the body. Therefore, the snow (舌 象) acts as a window reflecting the change of the internal organs, and is also used as an important dialectical indicator in various dialectic (八 綱, 六 經, 三焦 衛 氣 營 血).

Diagnosis, which is considered as the main diagnostic factor in tongue diagnosis, is formed by the accumulation of food debris, blood cells, and oral microorganisms between the papillary keratin and the mother's papillae, which is separated from the tongue. Sedimented in the gap is white, thin and glossy (薄 白潤澤) appears, which is called self-determination of the tongue (自決 作用) in other words. If pathological changes occur in the body, the secretion of saliva is suppressed, and it affects self-determination, which leads to a different condition than the normal one. Especially, if the digestive system has a problem, the condition becomes thick and the color varies from yellow to grayish brown. Will change.

Among these, the thickness of the tongue is the second most important diagnostic criterion after the color of the tongue. It is an indicator that can measure the severity or the sheen weight of lesions. It is useful. However, the criteria for dividing the thickness of these tongues are ambiguous, and research is needed to quantify them.

In the present invention, in order to quantitatively distinguish the thickness of the tongue, we examined the method of quantitatively distinguishing the thickness of the tongue from the most common white and white feces.

Through various image transformation and image classification, we extracted significant variables for the distinction between posterior and gestation, and designed discrimination function based on this.

1 is a flowchart illustrating a method for determining the appearance of a larynx using a tongue image according to the present invention.

Referring to FIG. 1, the present invention first requires an original image in which only a tongue area is displayed. In order to obtain an objective tongue image, the present invention uses a digital tongue system.

FIG. 2 is a diagram illustrating an appearance of a digital photographing system, and FIG. 3 is a diagram illustrating a face image photographed by the digital photographing system.

2 is a device designed to obtain a snowy image using a standardized light source and a digital camera and to obtain a more accurate image through color correction, which effectively forms a dark room when the eyepiece is fixed to the eyepiece. Ergonomically designed and standardized light sources use strobe lighting with color temperature (5500K) close to sunlight.

However, it is possible to distinguish the posterior and the fetus by applying the present invention even if the image is not taken by the digital tongue system, and the images used in the present invention are not limited to those described above, and all images are taken when the tongue part of the tongue is photographed. The present invention can be applied.

In the present invention, 720 subjects were recruited from an oriental hospital and a cooperating institution to acquire a facial image including the tongue as shown in FIG. 2, and only the image of the tongue was extracted therefrom.

Since only the tongue area is required for diagnosis, the image extracted only the tongue part becomes the original image in the present invention. That is, it is preferable that the parts other than the tongue are black, and lips, jaws, etc. other than the tongue are not visible.

The original image in which the extracted body appears is a 24-bit BMP image of RGB and has an image resolution of 1280 × 960, but an image of various formats or various resolutions may be used.

According to the present invention, a method for determining a late condition of a fetus using a tongue image is obtained by correcting an original image photographed by a body region through a preprocessing process. This is because the color or brightness of the original image varies depending on the lighting conditions and the surrounding environment when shooting, and not only to remove the influence of the lighting or the surrounding environment, but also to make more accurate discrimination using the corrected image. . In other words, the corrected image together with the original image is also used to determine the post-natal status is more accurate.

In detail, the original image is corrected for brightness and saturation (S100).

This process will be described in more detail with reference to FIGS. 4 to 5C. 4 is a flowchart illustrating a process of obtaining a corrected image by correcting an original image, FIG. 5A is an original image, FIG. 5B is a brightness correction image with corrected brightness values, and FIG. 5C is a saturation correction image with corrected saturation values. This is the illustrated figure.

Referring to FIG. 4, in the correcting an image in the present invention, first, an original image is converted into an HSV format (S110).

Figure 5a is the original image showing only the body portion, the upper and lower two images are slightly different in brightness depending on the shooting state.

H (Hue) represents hue, S (Saturation) represents saturation, and V (Value) represents brightness.

Next, the brightness value (V value) histogram of the original image is equalized to obtain a brightness corrected image (S120).

In order to correct images with different brightness, the histogram of the V value representing brightness in HSV is smoothed to widen the brightness distribution biased to a specific part so as to have a uniform brightness value overall.

FIG. 5B is a brightness correction image obtained by smoothing the image of FIG. 5A. Looking at the two images at the top and bottom, the original image at the top was dark and the original image at the bottom was relatively bright, but as the result of the brightness smoothing, the two images had similar brightness.

Next, the histogram of the S value of the HSV is expanded to emphasize the saturation values showing differences in tongue and tongue (S130). The S value expansion may be performed on the brightness correction image obtained by the original image or by smoothing the preceding brightness.

Referring to FIG. 5C, the white portion is whiter and the red portion is corrected redder. The white part is the tongue with the tongue, and the red part is the tongue of the tongue.

It can be seen that the tongue is more clearly observed by the correction.

Next, the maximum horizontal axis color value is extracted from the original image, the brightness correction image, and the saturation correction image (S200).

Here, the maximum horizontal axis refers to the horizontal line of the largest horizontal axis in the body image, and extracts color values existing in the horizontal line along the line.

6A is a diagram illustrating a maximum horizontal axis W0 in an original image, and FIG. 6B is a graph illustrating color value distributions obtained by decomposing color values of the maximum horizontal axis into R, G, and B values.

Referring to FIG. 6B, since the tongue has a reddish color as a whole, it can be seen that the R value is thickly distributed.

The average and standard deviation of the extracted color values are obtained.

Next, a threshold value is set based on the horizontal axis color value analyzed above to divide the tongue area and the tongue area (S300).

7 is a diagram specifically illustrating a process of dividing a tongue area and a tongue area according to the present invention.

Referring to FIG. 7, first, a predetermined threshold value is set within the color value range of the largest horizontal axis (S310). This threshold is the boundary that separates white from red.

The threshold value is compared with the color value of each pixel (S320), and when the color value of the corresponding pixel is greater than or equal to the threshold value, it is determined to be close to white and classified as a setting area (S330). In step S340, the cataract is classified into a lingual area.

8 is a diagram illustrating a tongue area image and a tongue area image.

(A) of FIG. 8 is a tongue area image of an original image, and (b) is a tongue area region image of an original image.

(C) of FIG. 8 is a tongue area image of the chroma correction image, and (d) is a tongue area image of the chroma correction image.

As described above, the process of dividing the lingual region and the lingual region is performed on both the original image, the brightness correction image, and the saturation correction image.

Next, the original image, the brightness correction image, and the saturation correction image are separated into tongues and tongues, and color values of the horizontal axis are extracted from 1/4, 1/2, and 3/4 of the tongue area image among the two images (S400). ).

9A is a diagram showing horizontal axes W1, W2, and W3 of 1/4, 1/2, and 3/4 points in the tongue area image, and FIG. 9B shows 1/4, 1/2, and 3/4 points. The graph shows the color values on the horizontal axis.

Each dividing point of FIG. 9A divides the top and bottom of the setting area in full length. It is not based on the entire body but only on the area of the tongue.

The average and standard deviation of the extracted color values are obtained.

In the case of white fetus, the area where white color is thicker and thicker than that of fetuses is wider and thicker, so there is a difference in SV, saturation value of HSV between fetuses and posterior fetuses. have. In particular, due to the peculiar nature of the taetae, the difference in the standard deviation rather than the average value.

The images were analyzed to make a discriminant function using the mean and standard deviation of each color value obtained above.

A threshold was specified after comparing the G-values of the areas where the appearance was obvious and those of the other areas in the image extracted into the tongue area. We compared the area and distribution of the area where G value is above a certain value by separating the setting area based on the threshold value.

FIG. 10A is a diagram illustrating an image of an original lingual area and a threshold value of a G value of 180 when a white poo is 180, and FIG. 10B is a diagram of an original gestation of a white larynx. When the threshold value of the region image and the G value is 180, the image when the threshold value of the G value is 200 is illustrated.

As a result of the analysis, it can be seen that the range of the larynx is larger than the threshold value of the larynx in general.

The present invention provides 24 variables (original image, saturation corrected image, and each average and standard deviation of the maximum horizontal axis color values of the brightness corrected image, 1/4, 1/2 of the original image, saturation corrected image, and brightness corrected image). , The mean and standard deviation of the horizontal axis color values of 3/4 points) were used to find the significant variables in the classification and derived the discriminant through statistical techniques.

Discriminant analysis used in the analysis is one of the multivariate statistical techniques, and finds discriminant variables (independent variables) of equal or analogous scales that can explain categorical group variables (dependent variables). It is a method to find out the group to which each object belongs by substituting the characteristics of each object to be classified.

In the present invention, the discrepancy function was obtained by using the standard deviation of the G value as a variable since the difference in the standard deviation rather than the average value was due to the characteristic of the sparse fetus.

The discrimination function thus created is shown in Equation 1 below, and Equation 2 is a discrimination function created using an average.

Is the discrimination coefficient of the scalar value, A, B and C are the discrimination coefficients of the 3x1 vector value, and D is the discrimination degree.

The standard deviation of the maximum horizontal axis G value and the standard deviation of the G-value of the brightness correction image are obtained through the standardized discriminant coefficient representing the relative importance of each independent variable obtained from various tongue images through the statistical program. It was found that the explanatory power was high in distinguishing the posterior and the young, and among them, the standard deviation of the G value corresponding to the maximum horizontal axis was 1.037, and the standard deviation of the G value of the quarter point of the brightness-adjusted image was -0.512. It was found that the standard deviation of the horizontal axis G value was the most explanatory. Using these two variables, we can find the standardized canonical discrimination function that can effectively distinguish between posterior and gestational expressions. The discriminant function is a linear combination of two variables.

Here, α and β are the coefficients of discrimination of scalar values, and D is a discrimination degree.

The output D value is obtained by inputting the G standard deviation of the maximum horizontal axis and the G standard deviation of the quarter obtained after brightness correction using the above-described discrimination function, and the D value is a predetermined reference value. If it is larger than, it is classified as posterior.

In one embodiment of the present invention, the reference value can be automatically obtained through a statistical program. In the experiment of the present invention, the reference value was calculated as 0.806.

11 is a diagram showing the results of examining the frequency distribution of D values of 52 females and females who were obtained in the experiment using the discriminant function.

Referring to FIG. 11, the center point of the posterior condition distribution is 1.908 and the center point of the park condition distribution is -0.297, which shows that the ranges of the two tongues are relatively distinct, and it is confirmed that there are few overlapping distributions.

Table 1 below is a table comparing the results of the diagnosis using the method of determining the posterior fetal palpability using the tongue image according to the present invention and the results diagnosed by a traditional Chinese medicine doctor.

Expected Results all Park Tae After Chinese medicine diagnosis Park Tae 45 43 (95.6%) 2 (4.4%) 45 Whoa 7 2 (28.6%) 5 (71.4%) 7

To see the accuracy of the discriminant function, we compared the results of the discriminant function with the diagnosis of the Chinese medicine. As shown in [Table 1], the results of the discriminant function classified 43 patients as Park Tae for 45 patients diagnosed with Park Tae-tae, and in the case of post-mortem, 5 of 7 patients diagnosed by Oriental Medicine were classified as posterior. Could. This shows that the overall prediction rate is 92.3%.

As described above with reference to the drawings, but the method of determining the larynx fetuses using the tongue image according to the present invention, the present invention is not limited by the embodiments and drawings disclosed herein, within the scope of the technical idea is protected Can be applied.

1 is a flow chart illustrating a method for determining the appearance of a larynx using a tongue image according to the present invention;

2 is a view showing the appearance of the digital snow-collecting system;

3 is a view showing a facial image photographed by the digital photographic system;

4 is a flowchart illustrating a process of obtaining a corrected image by correcting an original image;

5A is a diagram illustrating an original image;

5B is a diagram illustrating a brightness correction image in which brightness values are corrected.

5C is a diagram illustrating a saturation correction image in which a saturation value is corrected;

6A is a diagram illustrating a maximum horizontal axis W0 in an original image;

6B is a graph showing a color value distribution obtained by decomposing a color value of a maximum horizontal axis with respect to R, G, and B values;

7 is a view specifically illustrating a process of dividing a tongue area and a tongue area according to the present invention;

8 is a diagram illustrating a tongue area image and a tongue area image;

FIG. 9A is a diagram illustrating horizontal axes W1, W2, and W3 of positions 1/4, 1/2, and 3/4 in a tongue area image.

9B is a graph showing color values in the horizontal axis at 1/4, 1/2, and 3/4 points;

10A is a diagram illustrating an image of an original lingual region and a threshold value of a G value of 200 when a white poo is 180 and a threshold value of a G value of 200;

FIG. 10B is a view illustrating an image of an original tongue area image and a case where a threshold value of the G value is 200 when the white tongue is 180 and the threshold value of the G value is 180.

11 is a diagram showing the results of examining the frequency distribution of D values of 52 females and females who were obtained in the experiment using the discriminant function.

Claims (11)

A first step of obtaining a brightness correction image and a saturation correction image by correcting the original image photographing the body part with respect to brightness and saturation, respectively; Extracting a color value of a maximum horizontal axis from the original image, the brightness correction image, and the saturation correction image to obtain an average and a standard deviation; Setting a threshold value from the extracted color values and dividing the original image, the brightness correction image, and the saturation correction image into a tongue area image and a tongue area image based on the threshold value; A fourth step of extracting horizontal axis color values from the 1/4, 1/2, and 3/4 points of the three tongue area images divided in the third step, and obtaining the average and standard deviation; And a fifth step of determining a posterior condition and a gestation using a degree of discrimination of a discriminant function using the average or standard deviation calculated in the second and fourth steps as variables. The method according to claim 1, The first step may include a first step of converting an original image into an HSV format image; Obtaining a brightness correction image by equalizing a V value histogram of the HSV format image; And a third step of obtaining a saturation correction image by stretching the histogram of the S value of the HSV format image. The method according to claim 1, In the second step, the maximum horizontal axis is a posterior feces diagnosis using the tongue image, characterized in that the widest width of the tongue area. The method according to claim 1, The third step may include a first step of setting a threshold; A second step of dividing an area having a color value greater than or equal to the threshold value into an area image; And a third process of dividing an area having a color value below the threshold into a lingual area image. The method according to claim 1, The fourth step is to determine the post-mortem condition using the tongue image, characterized in that the horizontal axis color values are extracted at the 1/4, 1/2, 3/4 point from the top of the tongue area image. The method according to any one of claims 1 to 5, The color value and the threshold value is a post-mortem condition determination method using the tongue image, characterized in that the G value of the RGB color value. The method according to claim 1, The fifth step may include a fifth step of determining post-mortem and gestation on the basis of the degree of discrimination of the discriminant function using the standard deviation of the G values calculated in the second and fourth steps as variables. Determination method of postpartum condition. The method according to claim 1, The determination function of the fifth step is a post-mortem condition determination method using the tongue image, characterized in that the following equation.

(Where α is the coefficient of discrimination of scalar values, A, B and C are the coefficients of discrimination of 3x1 vector values, and D is the degree of discrimination) The method according to claim 1, The determination function of the fifth step is a post-mortem condition determination method using the tongue image, characterized in that the following equation.

(Where α is the coefficient of discrimination of scalar values, A, B and C are the coefficients of discrimination of 3x1 vector values, and D is the degree of discrimination) The method according to claim 1, The determination function of the fifth step is a post-mortem condition determination method using the tongue image, characterized in that the following equation.

(Where α and β are the coefficients of discrimination of scalar values, and D is the degree of discrimination) The method according to claim 8 or 9, The method of judging fetuses using the image of the tongue is to determine the feces of the posterior feces when the degree of discrimination D is greater than a predetermined reference value, and to determine the feces of feces when the difference is smaller than the reference value.

Images