JPH0877334A - Automatic feature point extracting method for face image - Google Patents

Abstract

PURPOSE: To automatically and easily extract the feature point of a face image at high speed by changing the algorithm of feature point detection corresponding to the part of the face and performing the detection in the optimum order of detection. CONSTITUTION: A head apex 1, right eyebrow left end point 2, right eyebrow right end point 3, right eyebrow upper end point 4, right eyebrow lower end point 5, left eyebrow left end point 6, left eyebrow right end point 7, left eyebrow upper end point 8, left eyebrow lower end point 9, right eye left end point 10, right eye right end point 11, right eye upper end point 12, right eye lower end point 13, left eye left end point 14, left eye right end point 15, left eye upper end point 16, left eye lower end point 17, nose left end point 18, nose right end point 19, nose lower end point 20, lips left end point 21, lips right end point 22, lips upper end point 23, lips center end point 24, lips lower end point 25, and chin lower end point 32 or the like are used as the feature points of the face image. When calculating these feature points, binarization processing is executed within respective candidate areas in the order of the lips, eyes and eyebrows, the feature points are extracted from the extracted object areas, the feature points are decided by edge processing within the candidate areas for the other parts, and a feature point group can be automatically extracted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for automatically extracting a feature point of a face image. More specifically, a candidate area for each part of the face image is set, and a threshold value is automatically set within this area. The present invention relates to a method of automatically extracting feature points of a face image by performing a binarization process.

[0002]

2. Description of the Related Art Since human faces are rich in changes and movements and rich in facial expressions, many studies have been conducted as having a great role as a medium for transmitting information between humans and computers and between humans and computers. For example, facial expressions play a major role as information transmission elements when people communicate with each other without using words.

On the other hand, each person's face is different,
It can be useful information to characterize each individual or group (eg, racial group).

If such face image information is handled as a bitmap image, which is considered as a set of small points, the amount of information becomes very large, but so-called "characteristic points" are taken out only from the difference. If you pay attention to, you can express the whole with a small amount of information, and it will be very effective for electronic or mechanical processing.

A face image captured by a CCD camera, a scanner or the like is stored in a memory as bitmap information or displayed on a display. The method that has been used for a long time to determine the feature points is to manually determine the feature points by operating the mouse or keyboard while looking at the display on which the face image is displayed. However, this method requires a great deal of time and labor and is not practical.

[0006] In recent years, some attempts have been made to automate the extraction of feature points of a face image due to the development of image processing.

These conventional feature point extraction methods include H
There are a method of attempting to extract feature points of a face from an image including a person by the VS color system, and a method of performing feature point extraction by automatically setting a threshold value according to an area change using the YIQ color system.

[0008]

However, these methods have a problem in that when extracting feature points, the detection accuracy is remarkably lowered depending on the part because the algorithm is fixed even if the detected part on the face is different. .

The present invention has been made in view of the above points, and provides a method capable of automatically and quickly and easily extracting feature points of a face image, which has been spending a lot of time and labor in the past. With the goal.

Another object of the present invention is to automatically extract a feature point group suitable for various applications such as facial expression synthesis using facial feature points, facial expression transformation, facial expression recognition, personal identification, and average value synthesis. Is to provide.

[0011]

In order to achieve the above object, the present invention provides, as feature points of a face image, a head vertex, a right eyebrow left end point, a right eyebrow right end point, a right eyebrow upper end point, a right eyebrow lower end point. , Left eyebrow left endpoint, left eyebrow right endpoint, left eyebrow upper edge point, left eyebrow lower edge point, right eye left edge point, right eye right edge point, right eye upper edge point, right eye lower edge point, left eye left edge point, left eye right edge point, left eye upper edge point, left eye Bottom point, left nose endpoint,
Right nose, bottom nose, left lip, right lip, top lip,
Lip center point, lower lip point, nose left side face contour point, nose right side face contour point, lip left side face contour point, lip right side face contour point, chin left side face contour point, chin right side face contour point, lower jaw point, When obtaining these feature points, binarization processing is performed in each candidate area in the order of lips, eyes, and eyebrows, feature points are extracted from the extracted target area, and other parts are extracted in the candidate area. Feature points are determined by edge processing.

[0012]

With the above-described structure, the present invention changes the feature point detection algorithm depending on the face part and performs detection in the optimum detection order.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings.

When attention is paid to the shape of the face, the information representative of the characteristics of each individual is the information on the constituent parts such as the facial contour line, eyes, lips, and eyebrows.

The characteristic information is as follows: -Effective information for characterizing each individual-The characteristic points are determined from the viewpoint of information that can be extracted relatively easily by the image processing method, as shown in FIG.

These 32 points are considered to be the minimum points indicating the characteristic points of the face, but if a smaller number of points is required and some trouble can be tolerated, 26-27, 3
The 0-31 pair or 28-29, 30-31 pair can be omitted.

When considering a series of processes from the input of a face image to the extraction of feature points, the problem of which part to extract first appears. For that part, conditions such as: -There are few individual differences in color and shape-It is in a position where it can be smoothly guided to the extraction process of other parts-Extraction is as easy as possible.

The lips can be considered as a part satisfying these conditions. There are few individual differences in the lips, and the eyes are based on the lips.
Since there is a positional relationship in which important features such as the nose and chin can be extracted, and the region is specialized by operating the RGB system like the Q axis of the YIQ color system, extraction is relatively easy.

Normally, the image signal is given in RGB system, and many image devices and the like also assume this RGB, which is convenient for approaching color information and displaying.

However, it is necessary to use a color system converted according to the intended use in order to require excellent color discrimination information as in human vision and to use the color information accurately. .

As a representative example of the non-RGB system, a luminance signal,
There is a YIQ color system that consists of color difference signals. The YIQ color system is obtained by linearly converting the RGB color system as shown in Equation 1.

[0022]

[Equation 1] The Y-axis represents the luminance signal, the I-axis represents the color difference signal between red and blue-green, and the Q-axis represents the color difference signal.
The axis represents a color difference signal between yellow-green and purple, and the coefficient is selected so that the axis having a sensitive visual characteristic regarding color change is I and the axis having a dull axis is Q. In the present invention, a specific color difference signal is first set in order to find the position of the lips.

That is, in order to emphasize the lip region (R-G
Set a color difference signal such as -B / 2). The lip area is obtained as a difference by (RG), but some skin area is also included. Therefore, B / 2 was set to correct it, and it was reduced from (RG).

This color difference signal is further squared and the normalized signal is used for the next processing.

In the present invention, the "weighted histogram" obtained by summing the pixel values for each line in the Y (X) direction in the X (Y) direction is used to set the candidate regions for the lips, eyes, and eyebrows. To do.

FIG. 2 shows a conceptual diagram of the weighted histogram.

The histogram value becomes higher in a region where pixels of high brightness (or low brightness) are densely arranged. The constraint conditions regarding the peaks and valleys in this histogram are set for each part, and the rectangular candidate regions are set accordingly.

Hereinafter, a method for determining the lip candidate area using this weighted histogram will be described.

First, a weighted histogram in the Y direction is obtained. Here, it is considered that the point where the histogram has the peak value indicates the position of the lip in the Y direction. This histogram is smoothed by a low pass filter, and the width of the lip candidate region in the Y direction is determined from the positions of two valleys that sandwich the peak.

Next, of the widths in the Y direction, the widths in the X direction are determined by similarly using the weighted histogram for the X direction. Given the shape of the lips, there may be small valleys near the peaks. In that case, it is determined whether the value is appropriate from the size of the peak and the valley and the positional relationship.

In this way, the candidate area where the lips are likely to exist is set by the (R-G-B / 2) color difference signal and the weighted histogram.

When the candidate area is set, binarization processing is performed in this area, and feature points are extracted from the binary data.

Since the luminance change in the lip region is relatively monotonous, the binarization process of an algorithm suitable for it is performed. The operation is performed in the following steps. (1) A point at which the luminance change is maximum is searched for each line in the X direction, and threshold processing is performed for each line using the luminance value at that point. (2) The processing of (1) is also performed in the Y direction. (3) AND and EOR of binary data in X and Y directions
Ask for. (4) Using the zero crossing point of the difference histogram between AND and EOR as a threshold value, threshold processing is performed within the rectangular area. (5) The AND value obtained in (3) is ANDed with the binary image obtained in (4). (6) Areas other than the target area are determined by the size of the area, and the non-target area is removed.

FIG. 3 shows the binarization algorithm by the steps (1) to (6). However, since the binarization process for the lip region and the binarization process for the eyebrow region are performed by the same algorithm, FIG. 3 representatively shows the algorithm together with the image of the right eyebrow.

Next, a method of extracting feature points from the binary image thus obtained will be described.

As shown in FIG. 4, the characteristic points of the lips are A, B,
There are five points, C, D and E. These points correspond to points 21-25 in FIG.

As shown in FIG. 5, a histogram in the X-axis direction is taken for the binary image of the lips extracted in the candidate area. From the maximum value of the histogram, the point where the histogram value becomes 0 first in the positive or negative X direction is searched, and the X coordinates AX and BX of the left and right end points are searched.
To decide.

Next, as shown in FIG. 6, a strip area having widths on both sides around AX and BX is set, and Y is set in the area.
The histogram in the axial direction is taken, and the average value of the maximum value and the minimum value of the Y coordinate of the points where the histogram value exceeds 0 is defined as the Y coordinates AY and BY of the left and right end points, respectively.

Further, as shown in FIG. 7, the X coordinate of the upper and lower end points is the average of AX and BX as CX, and the Y coordinate is AY and BY. A strip area is set, a histogram in the Y-axis direction is taken in that area, and the maximum and minimum values of the Y coordinate of points where the histogram value exceeds 0 are set as the Y coordinates CY and DY of the upper and lower end points, respectively.

The point E shown in FIG. 4 which is the center of the lips is the point where the value of (R + G + B) becomes the minimum, by searching between the points C and D for each of the R, G and B images.

With such an algorithm, five points, which are characteristic points of the lips, are extracted.

Once the feature points of the lips have been extracted, the following 4
Two candidate areas can be set. (1) Candidate region of lower jaw point X coordinate: X _{mouth_lower} ± β Y coordinate: Y _{mouth_lower to} Y _{mouth_lower} + (Y
_{mouth_lower} −Y _{mouth_upper} ) × 3 (2) Candidate area for the lower end point of nose X coordinate: X _{mouth_upper} ± β Y coordinate: Y _chin − (Y _chin −Y _{mouth_center} ) × 0.7
_{~ Y mouth_upper} -5 (3) As a candidate region of the lateral end point of the nose, X coordinate: X _{nose_top} ~ X _{mouth_right} , X _{nose_top} ~ X
_{mouth_left} Y coordinate: Y _{nose_top} ± β (where X _{nose_top} = X
_{mouth_upper} , Y _{nose_top} = (Y _chin −Y _nose ) × 1.
05) (4) X coordinate: X _{mouth_right} ~ (X _{mouth_right} -X
_{mouth_left} ) × 1.2 Y coordinate: Y _{mouth_right} ± β, Y _{mouth_left} ± β where β is (0.25 / 100) N to (1.25 / 1)
00) is an integer between N and N is the number of pixels from the top of the head to the lower end of the jaw.

Edge processing (luminance change) is used to extract feature points from these candidate areas.

The image used is a Y-axis image, and each feature point is determined by using the edge size and positional relationship. Conditions are set for each part. For example, in the case of extracting face contour points as shown in FIG. 8, the edge closer to the lips is selected.

Next, a method for setting the eye candidate area will be described.

First, as preprocessing, a wide area as shown in FIG. 9 obtained by the extracted feature points of the lips, the face contour points corresponding to the height of the lips, the lower end point of the nose and the lower end point of the chin is set to the Y-axis ( Brightness signal). In order to clarify the difference between the black eye (low brightness) included in the eye area and the white eye and skin (high brightness), the Y-axis value is squared within the area shown in FIG. Normalize with. Also, an image in which the brightness is inverted so that the histogram responds to the low brightness value is used.

Next, a weighted histogram in the Y direction is obtained. After low-pass filtering, the histogram is searched in the -Y direction, and the position of the first large peak observed is Y
It is considered to be the existence position of the direction. Depending on the image, the shadow on the side of the nose may be observed as the first peak, so a judgment is made based on the size and positional relationship of the peak, and an appropriate value is derived. Then, a width of ± 20β pixels is added to the obtained Y coordinate, and this is set as the width of the candidate area in the Y direction.

Next, in the width in the Y direction, a weighted histogram is taken in the X direction. A histogram is searched from the X coordinate of the contour point corresponding to the height of the lips to the inside of the face, and the width in the X direction is determined from the size of the peak and the positional relationship.

If the eye candidate area is set, then 2
Feature points of the eye are extracted by the binarization process. The binarization algorithm is as follows.

Looking at the brightness values in the rectangular area set in the Y-axis image, it can be seen that there are many areas to be obtained in areas of lower brightness than the skin area, that is, in the valleys of brightness. Therefore, as shown in FIG. 10, the brightness value is searched for each line, and the next point for one pixel from the maximum point and the next maximum point 1
A binary image is generated in the rectangular area by setting "1" between the point before the pixel and the other area.

In order to further improve accuracy, the following constraint conditions are introduced.・ Ignore if the difference between the maximum value and the minimum value that is currently being focused is 10 or less.・ Ignore if there is only one pixel between the maximum point that is currently focused and the next maximum point.

Since the binary image obtained at this stage contains many unnecessary regions in addition to the target region, the connected regions are searched and the target region or the non-target region is searched according to their areas. Then, the non-target area is removed.

Labeling is used for division into a target area and a non-target area within the rectangular area (for example, division into a binarized area of the eye and a binarized area of the eyebrow within the rectangular area of the eye).

2 in the rectangular area of the eye as shown in FIG.
In order to extract only the eye area from the value area, the connected areas in the eight directions are searched for and labeled as shown in FIG. 11 (b). Then, of these labeled regions, the one having the largest area is set as the target region (FIG. 11).
(C)).

Conventionally, a method of generating a binary image by threshold processing or the like has been often used for extracting the eye region. However, the region obtained by the threshold processing did not always match the true shape due to the influence of shadows and the like. However, according to the method of the present invention, it becomes possible to determine the eye feature points with high accuracy.

The characteristic points of the eyes are A, B, and
There are four points, C and D.

As the method for extracting the feature points from the binary image, the algorithm already described in the lip feature point extraction method is used.

Once the eye feature points have been extracted, it is possible to set the head vertex candidate area by utilizing this. That is, the candidate region of the head vertex X coordinate: ((X _{eye_right} −X _{eye_left} ) / 2) ± β Y coordinate: 0 to Y _eye To extract feature points from this candidate region, edge processing (luminance change) is performed. To use.

FIG. 13 shows a summary of feature points required for setting candidate regions other than lips, eyes, and eyebrows.

Next, a method of extracting eyebrow feature points will be described.

Similarly, a weighted histogram is used to set the eyebrow candidate area.

As described above, the position of the eye in the Y direction is given as the position of the first peak in the histogram. Therefore, it can be said that the position of the peak that appears next to the peak is the position of the eyebrow in the Y direction. The obtained Y coordinate −20β pixel position is the upper limit of the rectangular area, and the lower limit is the Y coordinate of the eye upper end point already obtained at this stage.

Next, the width in the X direction is determined from the positional relationship of the faces.

The binarization algorithm for the eyebrow region uses the same algorithm as for the lips. As preprocessing for binarization, in the case of the eyebrow area, the brightness of one line at the left and right ends of the rectangular area set on the Y-axis image is compared to easily and linearly correct the shadow component.

As shown in FIG. 14, the feature points of the eyebrows are A, B,
There are four points, C and D. Four feature points are extracted in the same way as done for the lips.

By using the feature points of the face image thus obtained, the wire frame model can be automatically generated. All the grid points N for constructing the wireframe model are interpolated and generated by the ratio of the 32 feature points obtained so far and the standard wireframe model.

However, here, N points are not directly generated from 32 characteristic points, but N / 5 to N / 3 points which include more face information of an individual such as an eyeball, a nostril, and a chin contour point are first generated. Is generated, and the final grid point N is generated from this.

FIG. 15 shows a flowchart of the feature point automatic extraction method described so far.

[0069]

As described above, according to the present invention,
It is possible to automatically and quickly perform feature point extraction of a face image, which has conventionally required a lot of time and labor.

Further, according to the present invention, a feature point group suitable for various applications such as facial expression synthesis using facial feature points, facial expression transformation, facial expression recognition, personal identification, and average value synthesis is automatically extracted. You can

In addition, the conventional method has a problem in that the extraction of feature points significantly reduces the detection accuracy depending on the part because the algorithm is fixed even if the detected part on the face is different. In the present invention, since the optimum algorithm is used according to the part, the detection accuracy can be constantly increased.

By using the feature points of the face image obtained by the present invention, the following processing and analysis can be easily performed using a computer or the like. (1) Facial expression transformation A facial expression can be easily transformed by constructing a wireframe model using feature points and changing it.

[Brief description of drawings]

FIG. 1 is a diagram showing feature points of a face.

FIG. 2 is a conceptual diagram of a weighted histogram.

FIG. 3 is a diagram showing a binarization algorithm for lips and eyebrows.

FIG. 4 is a diagram showing feature points of a lip.

FIG. 5 is a diagram of a histogram in the X-axis direction.

FIG. 6 is a diagram of a histogram in the Y-axis direction.

FIG. 7 is a diagram illustrating a method of obtaining X coordinate values and Y coordinate values of upper and lower end points.

FIG. 8 is a diagram illustrating determination of a feature point based on edge strength.

FIG. 9 is a diagram showing an area for setting an eye candidate area.

FIG. 10 is a diagram illustrating a binarization algorithm of an eye area.

11A and 11B are diagrams illustrating extraction of a target region by labeling, FIG. 11A is a diagram showing a binary region in a rectangular region of an eye, FIG. 11B is a labeled diagram, and FIG. 11C is a target region. It is a figure.

FIG. 12 is a diagram showing eye feature points.

FIG. 13 is a diagram showing feature points required for setting candidate regions other than lips, eyes, and eyebrows.

FIG. 14 is a diagram showing feature points of eyebrows.

FIG. 15 is a flowchart of a feature point automatic extraction method according to the present invention.

[Explanation of symbols]

 1 head vertex 2 right eyebrow left end point 3 right eyebrow right end point 4 right eyebrow upper end point 5 right eyebrow lower end point 6 left eyebrow left end point 7 left eyebrow right end point 8 left eyebrow upper end point 9 left eyebrow lower end point 10 right eye left end point 11 right eye right end point Point 12 Right eye upper end point 13 Right eye lower end point 14 Left eye left end point 15 Left eye right end point 16 Left eye upper end point 17 Left eye lower end point 18 Nose left end point 19 Nose right end point 20 Nose lower end point 21 Lip left end point 22 Lip right end point 23 Lip upper end point 24 Lip center point 25 Lip lower edge point 26 Nose left side face contour point 27 Nose right side face contour point 28 Lip left side face contour point 29 Lip right side face contour point 30 Jaw left side face contour point 31 Jaw right side face contour point 32 Jaw lower end point

Claims (16)

[Claims] 1. A feature point of a face image is a head vertex, a right eyebrow left end point, a right eyebrow right end point, a right eyebrow upper end point, a right eyebrow lower end point, a left eyebrow left end point, a left eyebrow right end point, a left eyebrow upper end point, Left eyebrow lower end point, right eye left end point, right eye right end point, right eye upper end point, right eye lower end point, left eye left end point, left eye right end point, left eye upper end point, left eye lower end point, nose left end point,
Right nose, bottom nose, left lip, right lip, top lip,
Use lip center point, lip bottom point, nose left profile point, nose right profile point, lip left profile point, lip right profile point, chin left profile point, chin right profile point, chin bottom point Automatic feature extraction method for face images. 2. The method for automatically extracting feature points of a face image according to claim 1, wherein the number of pixels N from the apex of the head to the lower end of the jaw is 256 or more. 3. When setting candidate regions for lips, eyes, and eyebrows, Y
An automatic feature point extraction method for a face image, which uses a weighted histogram obtained by summing the number of pixels per line in the (X) direction in the X (Y) direction. 4. A binarization process is performed in the candidate region in the order of lips, eyes, and eyebrows, feature points are extracted from the extracted target region, and other parts are feature points by edge processing in the candidate region. A method for automatically extracting feature points of a face image, which is characterized by determining. 5. When automatically setting a lip candidate region, (R-
A method for automatically extracting a feature point of a face image, which uses a color difference signal of GB / 2) and a weighted histogram. 6. When deciding the eye and eyebrow candidate regions, the region obtained by the extracted feature points of the lips, the face contour points corresponding to the height of the lips, the lower end point of the nose and the lower end point of the chin are set on the Y-axis. (Brightness signal) is set, the Y-axis value is squared in the area, the value is normalized from 0 to 255, and an image in which the brightness is inverted is used, and the feature point automatic extraction of the face image is used. Method. 7. In the binarization in the eye candidate region, 1
The brightness value is searched for each line, and "1" is set between the point next to the maximum point by one pixel and the point one pixel before the next maximum point, and "0" is set in the other areas. A binary image is generated in the rectangular area, and if the difference between the maximum value of interest and the minimum value of 10 or less is ignored as a constraint condition, it is ignored, and between the maximum point of interest and the next maximum point. If there is only one pixel, it is ignored. Further, the connected regions are searched, the target region or the non-target region is judged by their areas, and the non-target region is removed. Feature point automatic extraction method. 8. As a lips and eyebrows binarization means, a point where the luminance change is maximum is searched for each line in the X direction, and the threshold value processing is performed for each line by the luminance value at that point, The process is also performed in the Y direction, AND and EOR of binary data in the X and Y directions are obtained, the zero crossing point of the difference histogram between AND and EOR is used as the threshold value, and the threshold value process is performed in the rectangular area. A feature point of a face image, characterized in that the AND value and the binary image after the threshold processing are ANDed to determine a region other than the target region based on the size of the area and remove the non-target region. Automatic extraction method. 9. As eye candidate regions, X coordinate: X _PEAK ± 2α, Y coordinate: Y _PEAK ± α (where α = 0.1N to 0.11).
A method for automatically extracting a feature point of a face image, wherein an integer value between N and N is the number of pixels from the top of the head to the lower end of the jaw. 10. As the candidate region of the lips, X coordinate: X _PEAK ± 2α, Y coordinate: Y _PEAK ± α (where α = 0.1N to 0.11).
A method for automatically extracting a feature point of a face image, wherein an integer value between N and N is the number of pixels from the top of the head to the lower end of the jaw. 11. As eyebrow candidate regions, X coordinate: X _PEAK ± 2α, Y coordinate: Y _PEAK −α (where α = 0.1N to 0.11).
A method for automatically extracting a feature point of a face image, wherein an integer value between N and N is the number of pixels from the top of the head to the lower end of the jaw. 12. As a candidate region for the lower jaw point, X coordinate: X _{mouth_lower} ± β, Y coordinate: Y _{mouth_lower to} Y _{mouth_lower} + (Y
_{mouth_lower} −Y _{mouth_upper} ) × 3 (where β =
Automatic extraction of feature points of a face image using an integer value between (0.25 / 100) N and (1.25 / 100) N, where N is the number of pixels from the head vertex to the lower jaw point Method. 13. A nose lower end point candidate region has: X coordinate: X _{mouth_upper} ± β, Y coordinate: Y _chin − (Y _chin −Y _{mouth_center} ) × 0.7
_~ Y _{mouth_upper} −5β (where β = (0.25 / 1
00) N to (1.25 / 100) N, where N is the number of pixels from the apex of the head to the lower end of the jaw). 14. As a candidate region of a lateral nose end point, X coordinates: X _{nose_top to} X _{mouth_right} , X _{nose_top to} X.
_{mouth_left} , Y coordinate: Y _{nose_top} ± β (where X _{nose_top} = X
_{mouth_upper} , Y _{nose_top} = (Y _chin −Y _nose ) × 1.
05, β = (0.25 / 100) N to (1.25 / 10)
0) An integer value between N, N is the number of pixels from the apex of the head to the lower end of the jaw, and is used to automatically extract the feature points of the face image. 15. A face contour point candidate area is defined by: X coordinate: X _{mouth_right} to (X _{mouth_right} −X
_{mouth_left} ) × 1.2, Y coordinate: Y _{mouth_right} ± β, Y _{mouth_left} ± β (where β = (0.25 / 100) N to (1.25 / 10)
0) An integer value between N, N is the number of pixels from the apex of the head to the lower end of the jaw, and is used to automatically extract the feature points of the face image. 16. As a candidate region of a head vertex, X coordinate: ((X _{eye_right} −X _{eye_left} ) / 2) ± β, Y coordinate: 0 to Y _eye (where β = (0.25 / 10
0) An integer value between N and (1.25 / 100) N, where N is the number of pixels from the apex of the head to the lower end of the jaw.