KR20050096484A - Decision of occlusion of facial features and confirmation of face therefore using a camera - Google Patents

Abstract

Disclosed is a method of checking a face of a person in an image acquired using a camera and determining whether it is covered by sunglasses or a mask. A method for identifying a face of a person according to the present invention includes the steps of (a) classifying a person's shape and background by using a difference image and face color information, and (b) using a two-dimensional profile in a person's shape region. Separating only the face area of the face; and (c) separating and outputting the area where the eyes, nose and mouth are located in the face area, and (d) determining whether a mask or sunglasses is worn.

According to the present invention, it is possible to determine whether the person displayed on the camera is covering an important part of the face with a mask or sunglasses, and whether the face recognition is possible by clearly determining the positions of the eyes, the nose, and the mouth. There are features that can be.

Description

Decision of Occlusion of Facial Features and Confirmation of Face therefore Using a Camera}

The present invention relates to a method of accurately distinguishing the presence and location of eyes, nose, and mouth for determining facial occlusion using a camera and explaining facial features. The present invention is designed to automatically determine whether a part of the face is intentionally hidden when the face is required to be identified or recognized, and thus can be used for authentication or security purposes.

The present invention relates to a method of determining whether a face is recognized by determining whether a face is covered using a camera.

This technology is used as a tool that allows users to withdraw money only to people who do not cover their faces with sunglasses or masks to check the user's identity in ATMs, which are widely used in banks, etc., or recognizes faces in various services using the Internet. Can be used for this purpose.

Substantially, the method of specifically distinguishing the wearing of sunglasses and a mask is not clearly presented in the related art. However, the following methods have been proposed and used to determine whether or not the face in the process of identifying the features of the face.

1) How to use only face color

It is a way to identify the areas that contain the color of the face and look for eyes, nose and mouth within the area. This method has strong advantages such as face position, rotation or size, but its performance depends on the lighting environment and it is not easy to implement an adaptive model for various lighting environments.

2) Using edge information

This method detects edges in the input image and finds shapes similar to the edges of the eyes and nose in the image. This method has the characteristic of accurately detecting the shape of eyes, nose and lips when the edge image is ideally obtained, but it is difficult to distinguish the edges of facial features among many edges when the background is complex. Adaptive threshold setting is difficult.

3) Use of models by learning neural networks

This method is to learn the features of the face from the input image after learning the features of the face in the neural network. Learning by many different training sets shows better performance, but it is impossible to use all face images of various lighting environments near infinity and requires a large amount of computation, which causes difficulty in real-time implementation.

4) Statistical approach

It shows good performance by detecting face by mathematical modeling using statistical characteristics, but it is not easy to extract general characteristics for various lighting environments during modeling. This is not stable and there is a problem that requires high complexity depending on mathematical modeling.

Therefore, the present invention provides a method for solving a problem that the conventionally proposed methods are not applied to various lighting environments, and provides a method of determining whether a face is covered or hidden within 0.2 seconds when an image of a face is input.

According to an aspect of the present invention, a method for determining whether a face is covered by a mask or sunglasses includes: (a) distinguishing a person's shape and a background by using a difference image and face color information; and (b Separating only the face region of the person from the shape region of the person using a two-dimensional profile; and (c) distinguishing and displaying the areas where the eyes, nose and mouth are located in the face region; and (d) a mask Determining whether or not wearing sunglasses; includes.

In addition, the step (a), (a-1) calculates the difference between the edge information in the input image at the time (a-1) gray-converted t time and the input image at the time (t-1) time of the edge calcination of the same location If the difference is greater than 80, it is determined that the person appeared by valid recognition as a difference; and (a-2) judging an area in which the difference image exists as a human area and separating it from the background.

Also, in the step (b), (b-1) is a morphological operator which is a morphological operator to connect adjacent pixels with respect to the pixels in which motion is detected and expands the pixels by applying a dilation of 3 pixel intervals, and based on the horizontal axis from the result. Performing a low pass filter to remove unnecessary high frequency components after calculating only pixels having the highest position on the vertical axis, that is, edge pixels of an upper boundary; And (b-2) calculating the difference information with the pixels having 8 pixel intervals from this result, and considers only difference information when the difference information is 20 or more, and when the difference information is generated continuously, the difference among the consecutive difference information. Selecting the largest one and removing the remaining ones, and extracting an area where a face can exist using a pair of pixels having difference information on a horizontal axis; And (b-3) in the extracted areas, the average height is more than 80 and the difference between two average heights is grouped and the average height is small when the combined area width is less than 50 and the continuous area exists. The area is preferably considered to be a noise area and removed.

In addition, in the step (c), the Y-profile is calculated for the area except for the left and right 1/5 of the horizontal length of the face area obtained in the step (b) before (c-1), and the maximum value of the detected profile is obtained. Low frequency filtering to remove the noise component after normalization; and (c-2) calculating the difference (first derivative) of the profile to investigate the inflection point for the peak and finally starting and ending the feature region. Determining; includes.

Also, in the step (d), (d-1) in the step (c), when 1) the lips or the nose are not detected, the midpoint interval between the nose and the lips has a distance of 50% or more of the horizontal length of the face. If the sign of the two diagonal inclinations using the lowest left-right boundary point of the nose element and the lowest horizontal center of the lip element is the same, determining that the mask is worn; and (d-2) the lips and nose are detected. And determining that sunglasses are worn when there is a peak of the horizontal profile for the gray transformation between the peaks of the horizontal profile for the edge transformation.

In addition, the step (e) preferably comprises a step of determining whether the eyes, nose and mouth are in the proper position.

In addition, the apparatus for determining the face is automatically covered by the automatic identification and sunglasses or mask according to the present invention for dealing with the other problem is a face region detection unit for distinguishing the shape and background of the person using the car image and the face color information; A facial feature separator that separates only the face region of the person from the shape region of the person using a two-dimensional profile; A detailed position determination unit of a facial feature for distinguishing an area where eyes, nose and mouth are located in the face area; Determining whether the wearing of the sunglasses and mask to determine whether wearing a mask or sunglasses; And an output unit for determining whether the eyes, the nose, and the mouth exist at an appropriate position and displaying the image.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 illustrates a block diagram of a face identification and occlusion determination apparatus according to an exemplary embodiment of the present invention, and FIG. 2 illustrates a face identification and occlusion determination according to an exemplary embodiment of the present invention performed in the apparatus of FIG. 1. The main steps of the method are shown in a flowchart. 2 is often referenced below.

Referring to FIG. 1, the apparatus for identifying and covering a face according to an exemplary embodiment of the present invention includes a camera unit 10, an image processor 14, and an output unit 12. The image processor 14 includes a face region detector 140. The face area detection unit 140 includes a difference image acquisition unit 142 and a face area determination unit 144 using a face color and a two-dimensional profile. In addition, the image processor 14 includes a face feature separator 146. The facial feature separator 146 includes a Y-axis profile acquirer 148 of the face region and a feature region separator 150 through peak and valley distribution analysis. In addition, the image processing unit 14 includes a detail position determining unit 152 of a facial feature. The detailed position determination portion 152 of the facial features includes the mouth region and the position determination portion 154, the nose region and the position determination portion 156, the eye region and the position determination portion 158, and the placement determination portion of the eye nose ( 160). In addition, the image processing unit 14 includes a sunglasses and a mask wearing decision unit 162. The sunglasses and mask wearing decision portion 162 includes a covering decision portion 164 of the eye feature by the sunglasses and a covering determination part 168 of the mouth feature by the mask. The output unit 12 is a display unit which displays whether sunglasses, masks are worn, and positions of eyes, nose, and mouth on the face.

The operation of the face identification and occlusion determination device as described above will be described.

The camera unit 10 continuously transmits the image I _t (x, y) to the image processing unit 14, and the difference image acquisition unit 142 of the face region detection unit 140. First, the input image I _t (x, y) using the following equation (1) is converted into an edge image E _t (x, y). In Equation 1, COP (i, j) is an operator for Canny edge detection and (*) represents convolution.

[Equation 1] E _t (x, y) = I _t (x, y) * COP (i, j)

In the next step, when N is 20 or more in Equation 2 using the difference between E _t-1 (x, y) and E _t (x, y), it is determined that a person appears in the image.

Equation 2 D _t (x, y) = E _t-1 (x, y) -E _t (x, y)

2, the camera 202 continuously provides an image, obtains the difference image in the difference image acquisition process 204, and determines that the person appeared in the image in the determination process 210 of the appearance of the person. Repeat until appropriate.

The difference image D _t (x, y) shown in FIG. 3C obtained by the difference image acquisition unit 142 of FIG. 1 is transmitted to the face region determination unit 144. The face region determination unit 144 fills the empty space between the edges through the opening operation of the morphology of Equation 3 to express the human region at D _t (x, y). In Equation 3, MDt (x, y) represents a difference image with edges connected, MOOP (i, j) represents an open operator of a morphology, and (*) represents a convolution.

Equation 3 MDt (x, y) = D _t (x, y) * MOOP (i, j)

4 (a) shows MD _t (x, y) 402 obtained by applying Equation 3 to D _t (x, y). In the face region determination unit 144 of FIG. 1, the two-dimensional profile in FIG. 2 is used for the MD _t (x, y). The face region is classified through the process of predicting the face region 212 and the process of identifying the face region using the color information 214. 4 (b) shows the result of applying the X-axis profile in the process 212 of applying the two-dimensional profile. In FIG. 4B, the upper horizontal part 404 shows the upper limit of the face area, and the lower horizontal part 406 shows the lower limit of the face area. 4 (c) shows the result of applying the Y-axis profile in the process 212 of applying the two-dimensional profile. In FIG. 4C, the left peak 408 shows the left limit of the face area and the right peak 410 shows the right limit of the face area. (404), (406), (408) and (410). An example of the face region determined by applying the face separator 216 of FIG. 2 using face color information in the region defined by FIG. 4 is shown as a rectangle 412 in FIG.

The face region 412 determined by the face region determination unit 146 is transmitted to the face feature separation unit 146 of FIG. 1. The facial feature separator 146 acquires the Y-axis profile of the face region at the position where the eyes, nose and mouth are located (148), analyzes the peak and valley of the profile (150), and horizontally separates the face region 412. . The starting point and the ending point of the horizontally separated region are determined by the following algorithm corresponding to the process 220 of FIG. 2. In the equation, x (i) represents the Y-axis profile, (i) starts at the bottom of the face region and x (i) represents the sum of the brightness of the I-th column.

[Start, end, peak, valley judgment algorithm]

Procedure 1. Set the point satisfying Equation 4 as the starting point of the region.

[Equation 4]

x (i-1) == x (i) && x (i + 1)> 0

Step 2. Set the point satisfying Equation 2 as the end point of the region.

[Equation 5]

x (i-1)> x (i) && x (i + 1) == 0

Step 3. Set the peak to satisfy the equation (6).

[Equation 6]

x (i-1) <t (i)> = t (i + 1)

Step 4. Determine the valley that satisfies Equation 7.

[Equation 7]

[x (i-1)> x (i) <x (i + 1)] AND [x (i-1)> x (i) == x (i + 1)]

FIG. 5A shows the Y-axis profile of the determined face region 412. An example of a viewpoint, an end point, a peak, and a valley obtained as a result of applying the feature point finding process 220 is illustrated in (b) of FIG. 5 (b) for the viewpoint 502, valley 504, peak 506, and end point 508. In order. In FIG. 5C, the result of dividing the face area 412 using the valley and the end point is shown using the red line 510, and the end point of the area is illustrated using the green line 512. Doing. In this way, the face area 412 is divided into seven areas of RC = {(R1), (R2), (R3), (R4), (R5), (R6), and (R7)}.

The set RC of the facial feature areas separated by the application of the facial feature separator 146 of FIG. 1 is transmitted to the detail position determining unit 152 of the facial feature. The delivered facial feature region, the mouth region for the RC, and the position determining unit 154 include Algorithm 1.

Algorithm 1: Lip Detection Algorithm

Procedure 1. In the edge image E _t (x, y), 10 initial regions of the lip candidates having an average of 12 or more brightness values of 4 or more pixels are selected for an area of 50% or less of the vertical area of the face area.

Step 2. For 8-neighboring pixels around each initial region, the same region includes pixels that are 50% or more of the brightness of the starting point.

Process 3. When the extended regions are neighboring and the following conditions are met, they merge into one region.

Condition 1) Combined Total Horizontal Length <70% of Face Length

Condition 2) Combined Aspect Ratio〉 = 2.0

Condition 3) Low Brightness Element / High Brightness Element〉 50%

Procedure 4. From these areas, the area with the color most similar to the lip color is selected as the lip area.

In FIG. 6, (a) shows an application example of the process 1 and 2 of the algorithm 1. In the RC, 10 initial regions of the lip candidates were detected and illustrated for (R1), (R2), (R3), and (R4), which correspond to an area where the center of the vertical region of the face region 412 is 50% or less. 6 (a) to 602 show five initial regions, and 604 shows the remaining five regions. (B) of FIG. 6 shows an application example of Process 3 of Algorithm 1. FIG. The areas marked (602) are close to the perimeter and are removed with noise, and the baths marked (604) are integrated into three independent mouth features, (604), (606), and (608), applying condition 3 in condition 1 . In FIG. 6, (d) shows the result of determining by mouth that the characteristic having the most characteristic of the lip color among the 604, 606, and 608 is shown.

Further, in the detailed position determination unit 152, the area of the nose with respect to the transferred facial feature region, RC, and the position determination unit 156 include algorithm 2.

Algorithm 2: Nose Detection Algorithm

Procedure 1. In the face region of G _t-1 (x, y) which is gray converted from the input image I _t (x, y), the center of the vertical region is 4 or more pixel brightness values for the region ranging from 16 to 75%. Eight initial candidates with a mean of 12 or more are selected.

Step 2. Include pixels in the region that are at least 90% of the brightness for 8-neighbor pixels around each initial region.

Process 3. The expanded areas merge into one area when the following conditions are met.

Condition 1) The difference between the brightness of the two elements 〈40% of the high brightness factor

Condition 2) Combined horizontal length 〈1/4 of face length

Condition 3) The combined midpoint is within 3/5 of the central part of the face

Condition 4) The combined length 〈1/5 of the face length

Aspect ratio of mating length> 1.3

Procedure 4. In the expanded area, select the candidates that meet the following conditions.

Condition 1) The horizontal length is 〈1/3 of the horizontal length of the face

Condition 2) The element must be in the center 3/5 of the horizontal position of the face area.

Condition 3) Horizontal Length 〈Horizontal Length / 10 Faces

0.4 〈Horizontal Length / Vertical Length 〈= 0.7

Process 5. The candidates for the nose are finally determined to satisfy the following conditions.

Condition 1) Detects two of the highest average brightness of pixels

Condition 2) Merging elements containing both under the following conditions

2-1) Coupling Length 〈3/10 of Face Length

2-2) The combined length 〈2/10 of the face length

Horizontal length / vertical length〉 1.3

Horizontal length 〈face horizontal length / 10 and horizontal length / vertical length〉 0.7

In FIG. 7, (a) shows four co-candidate regions 702 obtained through the processes 1 and 2 of Algorithm 2. In FIG. Since the nose is located at the center of the face, there may be a case where all features of the face are included as candidates. The nose regions 704 obtained by applying the processes 3, 4, and 5 of the algorithm 2 to these ko candidate regions are shown in (b) of FIG. 7.

If the facial features corresponding to the mouth and the eye are not defined in Algorithm 1 and Algorithm 2, the tag is set to indicate that the mask is masked by moving to the mask verification process 232 in the extraction and extraction process 230 of FIG. 2. Let's do it. When the facial features corresponding to the mouth and eyes are obtained, the eye extraction process 242 of FIG. 2 corresponding to the area and position determination unit 158 of the eye of FIG. 1 defined by the following algorithm 3 is performed.

Algorithm 3: Eye Detection Algorithm

Procedure 1. Four or more areas of the face of G _t-1 (x, y) that are gray converted from the input image I _t (x, y) within three times the distance between the nose and lips above the nose Eight initial candidate areas for which the average value of the pixel brightness values are 12 or more are selected.

Step 2. Include pixels in the region that are at least 90% of the brightness for 8-neighbor pixels around each initial region.

Step 3. Calculate the centerline using the highest horizontal center of the nose element and the lowest horizontal center of the lip element, and calculate two diagonal lines using the lowest left-right boundary point of the nose element and the lowest horizontal center of the lip element. Among the areas obtained, the areas existing between the diagonal lines are selected as the candidates for operation.

Process 4. Threshold of the cabinet formed by the two eyes and lips of the normal face by experiment by creating an inverse triangle combination of eye-mouth among the ratio of horizontal length / vertical length in the range of (1.0 ~ 7.5) Determine the eye by examining the combination with the angle that most closely resembles The distance between the midpoints of only two eyes should be less than 1.5 times the length of the diagonal of the lips.

In FIG. 8A, the centerline 806 of an area in which eye 3 may be calculated by using the positions of the nose 808 and the mouth 810 obtained through Algorithms 1 and 2 in Step 3 of Algorithm 3 is described. ), Left diagonal 804 and right diagonal 804. In addition, (b) of FIG. 8 shows eye candidate regions 812 obtained through steps 1 and 2. FIG. FIG. 8C illustrates the left eye 814 and the right eye 816 obtained by applying the process 4 to the eye candidate regions 812 obtained through the process 3.

When the eye is detected using the algorithm 3 in the eye detection process 242 of the flowchart of FIG. 2, the positions of the eyes, the nose, and the mouth are output to the output unit 12 of FIG. 1. If the eye is not detected, the sunglasses determination process 244 is executed in the flowchart of FIG. 2. The process is processed in the wearing decision portion 162 of the sunglasses and mask in FIG.

In the flowchart of FIG. 2, the mask determination process 232 is detected when the midpoint distance between the nose and the lips has a distance of 50% or more of the horizontal length of the face and is incorrectly detected by the lips or the nose, but within three times the vertical distance between the nose and the lips. Vertical zone distinction by the horizontal profile of the edges of the edges of the two diagonal slopes, where no eye candidates exist within the three zones above the nose and using the lowest left-right boundary of the nose element and the lowest horizontal center of the lip element. If the signs are the same, it is determined that a mask is worn. In FIG. 9, the lower two Y-axis profiles 902 and 904 correspond to the upper human regions 906 and 908, respectively. An example of the result of finally determining whether the mask is worn is shown in 914.

In the flowchart of FIG. 2, the sunglasses judging process 244 determines that the sunglasses are worn when the peak of the horizontal profile for the gray conversion exists between the peaks in the Y-axis profile. However, the test area is limited to a range within three times the vertical distance between the nose and the lips and three areas on the nose. Sunglasses have a certain area such as an ellipse because the lens part has a dark characteristic when converting gray. When the edge is used, the value exists at the boundary of the lens or the frame of the lens, and the value is not detected much in the sunglasses. Profiles have different characteristics. In FIG. 10, using the Y-axis profile 1008 that determines whether there is a peak of a horizontal profile for gray conversion based on the Y-axis profile 1004 of the lower left side, the sunglasses may be displayed in the face region 1004 wearing sunglasses. The peak line passes through the center. The face region 1014 illustrates a case where the eyes of 1012 are divided into two.

In the sunglasses and mask wearing decision unit 162 of FIG. 1, the mask wearing decision process 232 of FIG. 2 and the sunglasses wearing decision process 244 of FIG. 2 are passed to the output unit 12 of FIG. 1. (250) The results are shown as in Figure 11 (a) and (b). The determination result of wearing the sunglasses is indicated by Sunglass Image !!! 1104 in FIG. 11A and the result of wearing the mask is indicated by Mask Image !!! 1108 in FIG. 11B. In the case of passing through the eye extraction process 242 of FIG. 2, the position measurement result 260 of the eyes, nose, and mouth of FIG. 2 is output to the output unit 12 of FIG. 1. In FIG. 11C, the determination result of the face is displayed as Normal Face 1110, and the left eye 1112, the right eye 1114, the nose 1116, and the mouth 1118 are respectively shown in quadrangles. do.

As described above, according to the present invention, it is possible to determine whether the face of the person appearing in the camera is covered with sunglasses or a mask, and it is possible to check the exact positions of eyes, nose and mouth. In addition, in addition to sunglasses or masks, the face features such as eyes, nose, mouth, etc. disappeared by covering the face with the hand can be determined and can be used in various tools for verifying and authenticating the face.

1 is a block diagram showing the structure of the image pickup device and algorithm of the face identification and occlusion determination method according to an embodiment of the present invention.

FIG. 2 is a flow chart showing the main steps of a method for face identification and occlusion determination according to an embodiment of the present invention performed in the apparatus structure of FIG.

3 is a view for explaining a process of determining the appearance of a person in the edge difference image.

4 is a diagram illustrating a process of preprocessing an edge difference image through morphology opening and filtering and finding a face region through a two-dimensional profile.

FIG. 5 is a diagram illustrating a process of finding, from the Y-axis profile, an area where eyes, nose, and mouth are expected to exist in a face area.

6 is a diagram illustrating a process of finding a lip by searching a region in which the lip exists.

7 is a diagram illustrating a process of finding a lip by searching a region where a nose exists.

8 is a diagram illustrating a process of defining an area in which an eye exists using the position information of the lips and the nose and searching the inside to find the eye.

FIG. 9 is a diagram illustrating a process of determining whether a mouth is covered by a mask using a Y-axis profile.

FIG. 10 is a diagram illustrating a process of determining whether sunglasses are covering an eye in an area where an eye exists by using a Y-axis profile.

11 is a diagram illustrating an example of an output screen.

Claims (5)

In a method of determining whether a face is covered by sunglasses or a mask using a camera and determining the position of eyes, nose and mouth that are characteristic of the face, (a) classifying a person's shape and background by using the difference image and face color information; (b) separating only the face region of the person from the shape region of the person using a two-dimensional profile; (c) distinguishing and displaying an area where eyes, a nose, and a mouth are located in the face area; (d) determining whether a mask or sunglasses is worn; The method of claim 1, wherein step (a) comprises: (a-1) Calculate the difference between the edge information in the input image at gray time t and the input image at the previous (t-1) time. Determining that a person has appeared; And and (a-2) judging an area in which the difference image exists as a human area and separating the area from the background. According to claim 1, wherein step (b) is, (b-1) In order to connect adjacent pixels with respect to the motion detected pixels, the pixels are extended by applying a dilation of three pixel intervals, which are morphological operators, and pixels having the highest position on the vertical axis with respect to the horizontal axis from the result. In other words, after calculating only the edge pixels of the upper boundary, performing a low pass filter to remove unnecessary high frequency components; And (b-2) From this result, the difference information with the pixels having 8 pixel intervals is calculated. Only when the difference information is 20 or more, the difference information is regarded as valid difference information. Selecting the largest one and removing the remaining ones, and extracting an area where a face can exist using a pair of pixels in which difference information exists on a horizontal axis; And (b-3) Groups where the average height is more than 80 in the extracted area and the ratio of the two average heights is a certain ratio, the area where the average height is small when the width of the combined area is less than 50 and when there is a continuous area. Removing the face area as a noise area; and removing the face area from the input image. The method of claim 1, wherein step (c) comprises: (c-1) Calculate the Y-profile for the area except the left and right 1/5 of the horizontal length of the face area obtained in step (b) above, normalize to the maximum value of the detected profile, and then remove the noise component. Low frequency filtering; And (c-2) calculating the difference (first derivative) of the profile to investigate the inflection point for the peak and finally determining the start point and the end point of the feature region by the following algorithm; [Start, end, peak, valley judgment algorithm] Procedure 1. Set the point satisfying Equation 1 as the starting point of the region. [Equation 1] x (i-1) == x (i) && x (i + 1)> 0 Step 2. Set the point satisfying Equation 2 as the end point of the region. [Equation 2] x (i-1)> x (i) && x (i + 1) == 0 Step 3. Set the peak to satisfy the equation (3). [Equation 3] x (i-1) <t (i)> = t (i + 1) Step 4. Determine the valley that satisfies Equation 4. [Equation 4] [x (i-1)> x (i) <x (i + 1)] AND [x (i-1)> x (i) == x (i + 1)] (c-3) detecting areas that can be determined by eyes, nose, and lips in each of the divided areas using the following algorithms; determining the presence and location of the mouth, nose, and eyes; ; And [Lip Detection Algorithm] Process 1. For the region where the center of the vertical region of the face region is 50% or less in the edge image, 10 initial regions of the lip candidates having an average of brightness values of 4 or more pixels are 12 or more. Step 2. For 8-neighboring pixels around each initial region, the same region includes pixels that are 50% or more of the brightness of the starting point. Process 3. When the extended regions are neighboring and the following conditions are met, they merge into one region. Condition 1) Total combined horizontal length 〈70% of face length Condition 2) Combined Aspect Ratio〉 = 2.0 Condition 3) Low Brightness Element / High Brightness Element> 50% Procedure 4. From these areas, the area with the color most similar to the lip color is selected as the lip area. Nose Detection Algorithm Procedure 1. Select 8 initial candidate regions with average of 4 or more pixel brightness values of 12 or more in the center of the vertical region in the gray-converted image. Step 2. Include pixels in the region that are at least 90% of the brightness for 8-neighbor pixels around each initial region. Process 3. The expanded areas merge into one area when the following conditions are met. Condition 1) The difference between the brightness of the two elements 〈40% of the high brightness factor Condition 2) Combined horizontal length 〈1/4 of face length Condition 3) The combined midpoint is in the range 3/5 of the central part of the face Condition 4) Coupling length 〈1/5 of face length Aspect ratio of mating length> 1.3 Procedure 4. In the expanded area, select the candidates that meet the following conditions. Condition 1) The horizontal length is 〈1/3 of the horizontal length of the face Condition 2) The element must be in the center 3/5 of the horizontal position of the face area. Condition 3) Horizontal Length 〈Horizontal Length / 10 Faces 0.4 〈Horizontal Length / Vertical Length 〈= 0.7 Process 5. The candidates for the nose are finally determined to satisfy the following conditions. Condition 1) Detects two of the highest average brightness of pixels Condition 2) Merging elements containing both under the following conditions 2-1) Coupling Length 〈3/10 of Face Length 2-2) The combined length 〈2/10 of the face length Horizontal length / vertical length〉 1.3 Horizontal length <Face horizontal length / 10 / Horizontal length / Vertical length> 0.7 [Eye Detection Algorithm] Procedure 1. Select eight nose candidate areas with an average of 4 or more pixel brightness values of 12 or more for areas that are within 3 times the nose and lips gap above the nose. Step 2. Include pixels in the region that are at least 90% of the brightness for 8-neighbor pixels around each initial region. Step 3. Calculate the centerline using the highest horizontal center of the nose element and the lowest horizontal center of the lip element, and calculate two diagonal lines using the lowest left-right boundary point of the nose element and the lowest horizontal center of the lip element. Among the areas obtained, areas existing between these diagonal lines are selected as eye candidates. Process 4. Threshold of the cabinet formed by two eyes and lips of a normal face by experiment by creating an inverse triangle combination of eye-mouth in the ratio of horizontal length / vertical length in the range of (1.0 ~ 7.5) in the candidate area Determine the eye by examining the combination with the angle that most closely resembles The distance between the midpoints of only two eyes should be less than 1.5 times the length of the diagonal of the lips. (c-4) determining the occlusion by displaying the sunglasses image and the mask image, and displaying the positions of eyes, nose, and lips on the screen while displaying the normal face in the case of a normal face. The method of claim 1, wherein step (d) (d-1) In step (c) above, 1) when no lip or nose is detected, when the midpoint of the nose and the lip has a distance of 50% or more of the horizontal length of the face, the lowest left-right of the nose element Determining that a mask is worn when two diagonal gradients using the boundary point and the lowest horizontal center of the lip element are the same; And (d-2) Determining that the lips and nose is detected and wearing a sunglasses when the peak of the horizontal profile for the gray transformation is present between the peak of the horizontal profile for the edge transformation; including a mask and sunglasses How to determine whether the face is covered by using.