JPH11250267A - Method and device for detecting position of eye and record medium recording program for detecting position of eye - Google Patents

Abstract

PROBLEM TO BE SOLVED: To stably and properly detect the positions of both eyes in the face images of various persons photographed on various photographic conditions by selecting left and right eye candidate areas for generating a partial face image most similar to an average face image as a whole. SOLUTION: The partial face image is generated by a partial face image generating part 25. In this case, one of areas on the left side of a center line from side to side of a skin area is selected out of eye candidate areas and defined as the left eye candidate area, one of areas on the right side is selected and defined as the right eye candidate area. Next, the face section of the face image is segmented similarly to the generation of average face image based on the centroid positions of respective selected left and right eye candidate areas, and the partial face image is generated by expanding and reducing the size of the segmented image equally with the average face image. Next, a face image difference degree calculating part 26 calculates the degree of difference between the partial face image and the average face image. Then, the left and right eye candidate areas applying the minimum degree of difference are determined as left and right eye areas by left and right eye area determining parts 27.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an eye position detecting method, an eye position detecting device, and a recording medium on which an eye position detecting program is recorded. The present invention relates to an eye position detection method for automatically detecting an eye position from a face image of a person captured in an eye, and the like, an eye position detection device, and a recording medium storing an eye position detection program.

[0002]

2. Description of the Related Art For example, in a spectacle wearing simulation system for simulating how one's face looks when certain spectacles are worn, a technique for detecting an eye position from a human face image is required. That is, in the spectacle wearing simulation system, since the spectacle image is superimposed on the human face image, the superimposed position and the size of the spectacle image need to be determined based on the position of both eyes and the distance between the eyes in the face image. The following method (1) or (2) is known as a conventional method for detecting the position of an eye from a face image. (1) First, a skin color region is extracted from a face image, and the extracted region is regarded as a skin region. Next, a region different from the skin color is extracted from the skin region as an eye candidate region. A pair of two regions located above and symmetrically to the left and right with respect to the center of the skin region are regarded as a left eye candidate region and a right eye candidate region, and the center position of each of the two regions is defined as the position of both eyes. (2) Extract a skin-colored area from the face image, and extract a portion (edge) in which the luminance changes rapidly in the skin area,
From the extracted edges, the edge corresponding to the eye is selected from the positional relationship of the eyes, nose, mouth, etc. in the human face.

[0003]

However, the skin area in the face image is hardly uniform except for the eyes and the mouth, and is different from the skin color due to shading due to the unevenness of the face and uneven skin. There are many areas to be extracted. Therefore, the above (1),
As in the method (2), it is extremely difficult to correctly select a pair of regions corresponding to both eyes from such a large number of regions based only on the positional relationship of the eyes, nose, mouth, and the like.

For example, eyebrow parts other than eyes are often detected as areas that are significantly different from skin color. In such a case, the eyebrow area is regarded as an eye area, and an erroneous eye position is finally determined. The output of the position that has occurred has occurred with a relatively high probability.

SUMMARY OF THE INVENTION The present invention has been made to solve the problems of the prior art, and an object of the present invention is to provide an eye position detecting method and an eye position detecting apparatus which are used for photographing under various photographing conditions. Another object of the present invention is to provide a technology that enables stable detection of the positions of both eyes with respect to various human face images.

It is another object of the present invention to provide a recording medium on which an eye position detecting program for causing a computer to execute the eye position detecting method is recorded.

[0007] The above and other objects and novel features of the present invention will become apparent from the description of the present specification and the accompanying drawings.

[0008]

SUMMARY OF THE INVENTION Among the inventions disclosed in the present application, the outline of a representative one will be briefly described.
It is as follows.

That is, according to the present invention, first, an average face image generated by cutting out a face internal region from a typical face image of a person with reference to the position of the eyes is prepared in advance, and the input person's face image is prepared. The color face image is converted to a YIQ color expression to generate a YIQ face image, and the I component image of the YIQ face image is binarized by a threshold value determined based on the pixel value distribution to generate a binarized image. Generate a connected region located near the center of the face image and larger than a predetermined area as a skin region from the binarized image, and extract a plurality of connected regions that are not skin regions existing inside the outermost periphery of the skin region. Extracted as eye candidate area,
A pair of a left eye candidate region and a right eye candidate region is selected from the plurality of eye candidate regions, and a face internal region is selected from a color face image input based on the positions of the selected left eye candidate region and right eye candidate region. Is cut out to generate a partial face image, the degree of difference between the partial face image and the average face image is calculated, and the pair of the left face candidate area and the right eye candidate area in the eye candidate area is compared with the partial face image. A pair of a left-eye candidate region and a right-eye candidate region, which are generated and calculated for the degree of difference and have the smallest degree of difference, are determined as a left-eye region and a right-eye region, respectively, and are determined as eye positions.

Further, the present invention is characterized in that the center of gravity of each of the determined left eye region and right eye region is determined to determine the position of the eye.

Further, the present invention is characterized in that the positions of the eyes are determined by obtaining the vertical and horizontal centers of the determined left eye region and right eye region.

Further, according to the present invention, a brightness pattern of light reflected by the eyeball when the eyeball is illuminated from the front is prepared in advance, and the brightness pattern and the left eye area or the left eye area are determined inside the determined left eye area and right eye area. The degree of difference from the luminance of the right eye area is calculated while moving the luminance pattern, and the center position of the luminance pattern when the degree of difference becomes minimum is set to the left eye or right eye position.

[0013]

Embodiments of the present invention will be described below in detail with reference to the drawings.

In all the drawings for describing the embodiments, those having the same functions are denoted by the same reference numerals, and their repeated description will be omitted.

FIG. 1 is a block diagram showing a schematic configuration of an eye position detecting device according to an embodiment of the present invention. In the figure, 10 is a control unit, 11 is a face image storage unit, 12 is an average face image storage unit, 13 is a luminance pattern storage unit, 14 is a face image storage unit 11, an average face image storage unit 12 and a luminance pattern storage unit 13 , An input unit, 21 a YIQ conversion unit, 22 a binarized image generation unit, 23 a skin region extraction unit, 24 an eye candidate region extraction unit, 25 a partial face image generation unit, 26 Denotes a face image dissimilarity calculation unit, 27 denotes a left eye / right eye area determination unit, 28 denotes a left eye / right eye difference calculation unit, and 29 denotes a left eye / right eye position determination unit.

FIG. 2 is a flowchart for explaining the eye position detecting method according to the present embodiment. Hereinafter, FIG.
An eye position detection method according to the present embodiment will be described with reference to FIG. First, the average face image input from the input unit 20 is stored in the average face image storage unit 12 in the storage unit 14 (Step 101). The average face image is, for example, a background, hair, clothing, and the like cut out from a typical face image of a person, as shown in FIG. This is an image of only a portion inside the face that is not included. That is, as shown in FIG. 3, when the distance between the eyes is 2L (or 距離 of the distance between the eyes is L) in a typical face image, a, b, and c are appropriately set in advance. It is a constant set to a value, which is a × L from the left and right eye positions in the left and right direction, b × L from the eye position in the upward direction, and c × L from the eye position in the downward direction. In this way, the range is determined and cut out to obtain an average face image. Here, for the sake of simplicity, the average face image is a monochrome image.

Next, the luminance pattern of the light reflected by the eyeball input from the input section 20 is stored in the luminance pattern storage section 13 in the storage means 14 (step 102). The luminance pattern of the light reflected by the glasses is, as shown in FIG.
This is an image in which a portion is cut out to a predetermined size around a spot-shaped bright spot generated when light is applied from the front to the eyes. Here, for the sake of simplicity, the luminance pattern is a monochrome luminance image, but may be a color image.

Next, a color face image input from the input unit 20 for detecting the position of the eyes is stored in the face image storage unit 11 in the storage unit 14 (step 103). Here, for the sake of simplicity, it is assumed that the color face image is represented by three primary colors (RGB) of red, green, and blue. However, a color image represented by another color expression method may be used. Good. Note that when capturing a face image, it is assumed that one face of the face is illuminated with one light source and captured.

Next, the YIQ conversion unit 21 converts the color face image into a YIQ color expression to generate a YIQ face image (step 104). The conversion from the RGB color expression to the YIQ color expression is well known, and can be performed, for example, by calculating the following equation (1).

[0020]

Y = 0.299R + 0.587G + 0.114B I = 0.596R−0.274G−0.322B Q = 0.221R−0.522G + 0.311B (1) Next, a binarized image The generation unit 22 sets the value of each pixel of the I component image of the YIQ face image to “1” when the value of the pixel is larger than a certain threshold, and sets the value of the pixel to “0” when the value is smaller than the threshold. Is binarized (step 105). 2
The threshold for converting into a value can be determined based on the distribution of values of each pixel of the I component image. There are various ways to determine this threshold. For example, a histogram is generated in which the horizontal axis is the pixel value and the vertical axis is the number of pixels having that value, and the histogram is classified into two classes by a certain threshold value. Method to select the threshold value so that the variance between classes when splitting into classes is maximized (Otsu: “Automatic threshold value selection method based on discrimination and least square criterion”, IEICE Transactions,
vol. J63-D, no. 4, PP. 349-35
6, 1980).

Next, the skin region extraction unit 23 extracts a connected region which is a set of adjacent pixels having a value of “1” from the binarized image, and also calculates the center of gravity and area of the connected region. Is determined, and a connected region having the largest area is selected from the connected regions whose center of gravity is within a predetermined range indicating the center of the image, and is set as a skin region (step 106). The barycentric position of the region is defined as the average of the x coordinate values of the pixels included in the region, the x coordinate of the barycentric position, and the average of the y coordinate values of the pixels included in the region as the y coordinate of the barycentric position. It is a position to do.

Next, the eye candidate area extracting section 24 connects the I-component image with a pixel having a value of "0" inside a closed curve passing through the outermost periphery of the skin area, that is, the skin area. Regions that are not skin regions existing inside the outermost periphery (that is, holes in the skin region) are extracted, and their connected regions are set as eye candidate regions (step 107). FIG. 5 shows an example of the skin region 31 and the eye candidate region 32.

Next, a partial face image generation section 25 generates a partial face image. That is, one area on the left side of the left and right center lines of the skin area (31 shown in FIG. 5) is selected from the eye candidate area (32 shown in FIG. 5) as a left eye candidate area, and one area on the right side is selected. Select to make a right eye candidate area (Step 10
8). Next, based on the barycentric positions of the left eye candidate area and the right eye candidate area selected in step 108, the face portion of the face image is cut out in the same manner as the generation of the average face image, and the size of the cut out image is equal to the average face image. A partial face image is generated by enlarging or reducing so as to be the same (step 109).

FIG. 6 shows how a partial face image is cut out when the eyebrow part is a left eye candidate area and a right eye candidate area. As shown in FIG. 6, when the distance between the left and right eyebrows, which is the left eye candidate area and the right eye candidate area, is 2X (or 1/2 of the distance between the left and right eyebrows is X), L shown in FIG. Instead of X, a predetermined range is cut out from the input color face image 41 in the same manner as in FIG.
In addition, in order to simplify the following description, the partial face image 42
Is a black and white image obtained by cutting out the Y component image of the YIQ face image.

Next, the face image difference calculating section 26 calculates the difference between the partial face image (42 in FIG. 6) and the average face image (step 110). The calculation of the degree of difference is performed by, for example, setting the luminance (value) of the pixel at the position (i, j) of the partial face image and the average face image to P (i, j) and A (i, j), respectively.
Is calculated by the following equation (2).

[0026]

(Equation 2)

In the equation (2), N and M are the horizontal or vertical size (number of pixels) of the partial face image or the average face image, respectively. Further, as shown in the above equation (2), P
Not the square root of the sum of squares of the difference between (i, j) and A (i, j), but the sum of the absolute value of the difference between P (i, j) and A (i, j) There is also a method of calculating the cross-correlation coefficient between the partial face image 42 and the average face image, and a method of determining the reciprocal of the obtained cross-correlation coefficient as the degree of difference.

Next, when the difference obtained by the current calculation is smaller than the minimum difference until now, the current difference and the left-eye candidate area and the right-eye candidate area given the difference are recorded. (Step 111). Then, the processing from step 108 to step 111 is performed for all possible selections of the pair of the left-eye candidate area and the right-eye candidate area selected from the eye candidate areas, and the left-eye / right-eye area determination unit 27 determines the minimum difference. The left-eye candidate area and the right-eye candidate area that give the degree are determined as the left-eye area and the right-eye area (step 113).

It has been empirically known that the binarization of the I component image of the YIQ face image can stably extract a skin region. By extracting a non-skin color portion in the skin region, the eyes and eyebrows can be extracted. Region can be stably extracted. Therefore, as in the eye position detection method of the present embodiment, the left eye candidate area and the right eye candidate area are appropriately selected from these eye candidate areas, and cut out based on the positions of the left eye candidate area and the right eye candidate area. The partial face image (42 shown in FIG. 6) is compared with an average face image generated from a typical human face image to generate a partial face image that is overall most similar to the average face image. By selecting the right and left eye candidate areas, the left and right eye areas can be reliably selected. In the eye position detection method according to the present embodiment, only the left eye candidate region and the right eye candidate region are detected and the mouth is not detected. Therefore, uncertain mouth detection (because the lip color is close to skin color, (It becomes difficult to extract the mouth region from the skin region), so that a stable result can be obtained.

Next, the left-eye / right-eye difference calculation unit 28 calculates the difference between the luminance pattern stored in the luminance pattern storage unit 13 and the Y component image of the face image. That is, the center of the luminance pattern of the reflected light from the eyeball stored in the luminance pattern storage unit 13 is placed at an appropriate place in the left eye area and the right eye area determined by the left eye / right eye area determination unit 27 (step 114) At that position, the difference between the luminance pattern and the Y component image of the face image is calculated within the range of the luminance pattern (step 115).

The calculation of the degree of difference is performed while moving the center of the luminance pattern in the left-eye area or the right-eye area (steps 116 and 117), and the luminance pattern giving the smallest degree of difference in each of the left-eye area and the right-eye area. And the left eye / right eye position determination unit 29
Then, the center position of the luminance pattern giving the minimum difference is determined as the position of the left eye or the position of the right eye (step 11).
8). Each of the above steps is executed based on an instruction from the control unit 10. As described above, in the eye position detection method according to the present embodiment, the luminance pattern of the light reflected by the eyeball when the eyeball is illuminated from the front is detected from the left eye area and the right eye area. Can be obtained. In the present embodiment, as a final eye position, a spot-shaped bright spot generated by reflection of illumination light on the eyeball is detected, and the position is determined as the eye position (steps 114 to 118). However, the position of the center of gravity of the left eye region and the right eye region obtained in step 113 can be used as the eye position, and the position of the left, right, up, and down centers of the left eye region and the right eye region can be used as the eye position. . Further, in the above-described embodiment, for example, processing for reducing minute noise in an input image, processing for correcting color and brightness, and extraction of a skin area or an eye candidate area ignore a minute connected area. If necessary, a well-known image process such as a smoothing process may be added.

In the eye position detecting method according to the present embodiment, the control unit 10, the input unit 20, the YIC conversion unit 2
1, binarized image generation unit 22, skin region extraction unit 23, eye region extraction unit 24, partial face image generation unit 25, face image difference calculation unit 26, left eye / right eye region determination unit 27, left eye / right eye difference The case has been described in which the calculation unit 28 and the left-eye / right-eye determination unit 29 are each configured by dedicated hardware, but the present invention is not limited to this.
It is also possible to execute by software processing of a computer. In this case, the eye position detection method of the present embodiment is performed by the central processing unit (CPU) 210 shown in FIG. 7 executing the eye position detection program read into the main memory 202. In this case, the YIC conversion section 21, the binarized image generation section 22, the skin area extraction section 23, the eye area extraction section 24, the partial face image generation section 25, the face image difference degree calculation section 26, the left eye / right eye area determination section 27, a left eye / right eye dissimilarity calculation unit 28, and a left eye / right eye determination unit 29 include a YIC conversion unit 211, a binarized image generation unit 212, a skin region extraction unit 213, an eye region extraction unit 214 shown in FIG. The central processing unit (C) includes a face image generation unit 215, a face image difference calculation unit 216, a left eye / right eye region determination unit 217, a left eye / right eye difference calculation unit 218, and a left eye / right eye determination unit 219.
PU) 210. In addition,
FIG. 7 is a block diagram illustrating a schematic configuration of a computer that executes the eye position detection program according to the present embodiment.
In FIG. 1, reference numeral 201 denotes an image input device for inputting an image, 202 denotes a main memory, 203 denotes an input device such as a keyboard and a mouse, 204 denotes an auxiliary storage device, 205 denotes a display device, 206 denotes a bus line, and 210 denotes a center. Processing unit (CP
U). The eye position detection program includes, for example,
Main memory 20 provided by a recording medium such as a CD-ROM
2 or, for example, an auxiliary storage device 2
04.

As described above, the invention made by the present inventor is:
Although a specific description has been given based on the above-described embodiment, the present invention is not limited to the above-described embodiment, and it is needless to say that various modifications can be made without departing from the gist of the invention.

[0034]

The effects obtained by typical ones of the inventions disclosed in the present application will be briefly described as follows.

(1) According to the present invention, the left eye candidate area and the right eye candidate area are appropriately selected from the eye candidate areas extracted as non-skin color parts in the skin area, and the positions of the left eye candidate area and the right eye candidate area are selected. Left eye candidate area that generates a partial face image that is overall most similar to the average face image by comparing the partial face image cut out based on the average face image generated from the typical face image of a person Since the left eye region and the right eye region are selected, the left eye region and the right eye region can be reliably selected.

(2) According to the present invention, since the luminance pattern of the light reflected by the eyeball when the eyeball is illuminated from the front is detected from the left eye area and the right eye area, the center position of the eye can be accurately determined. It is possible to obtain.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an eye position detection device according to an embodiment of the present invention.

FIG. 2 is a flowchart for explaining an eye position detection method according to the present embodiment.

FIG. 3 is a diagram for explaining the average face image shown in FIG. 1;

FIG. 4 is a diagram for explaining a luminance pattern of light reflected by the eyeball shown in FIG. 1;

FIG. 5 is a diagram showing an example of a skin region and an eye candidate region in the present embodiment.

FIG. 6 shows how a partial face image is cut out when the eyebrow portion is a left eye candidate region and a right eye candidate region in the present embodiment.

FIG. 7 is a block diagram illustrating a schematic configuration of a computer that executes an eye position detection program according to the present embodiment.

[Explanation of symbols]

10 control section, 11 face image storage section, 12 average face image storage section, 13 luminance pattern storage section, 14 storage means,
20 input unit, 21, 211 ... YIC conversion unit, 22, 2
12 ... binarized image generation unit, 23, 213 ... skin region extraction unit, 24, 214 ... eye region extraction unit, 25, 215 ... partial face image generation unit, 26, 216 ... face image dissimilarity calculation unit, 2
7, 217: left-eye / right-eye area determination unit, 28, 218: left-eye / right-eye difference calculation unit, 29, 219: left-eye / right-eye determination unit, 31: skin area, 32: eye candidate area, 41: color face image, 42 ... partial face image, 201 ... image input device, 20
2 main memory, 203 input device such as keyboard / mouse, 204 auxiliary storage device, 205 display device, 206
... bus line, 210 ... central processing unit (CPU).

 ──────────────────────────────────────────────────の Continuing on the front page (72) Inventor Satoshi Suzuki Nippon Telegraph and Telephone Corporation, 3-19-2 Nishishinjuku, Shinjuku-ku, Tokyo

Claims (9)

[Claims] 1. An eye position detecting device for detecting an eye position from a color face image of a person, wherein first means for converting an input color face image of the person into a YIQ color expression to generate a YIQ face image. Second means for binarizing the I component image of the YIQ face image generated by the first means with a threshold value determined based on the pixel value distribution to generate a binarized image; A third means for extracting, as a skin area, a connected area located near the center of the face image and larger than a predetermined area from the binary image generated by the second means; Means for extracting, as eye candidate areas, a plurality of connected areas that are not skin areas existing inside the outermost periphery of the skin area, and selecting a left eye candidate from the plurality of eye candidate areas extracted by the fourth means. Fifth means for selecting a pair of a region and a right eye candidate region; For all pairs of the left-eye candidate area and the right-eye candidate area selected by the fifth means, based on the positions of the left-eye candidate area and the right-eye candidate area, a face internal area is obtained from the input color face image. And generating a partial face image by cutting out a partial face image generated by the sixth means, and extracting a face internal area from a typical face image of a person based on the position of the eyes. Calculating the degree of difference from the average face image generated by
Means, and an eighth means for determining a pair of a left-eye candidate area and a right-eye candidate area having the smallest degree of difference calculated by the seventh means as a left-eye area and a right-eye area. Eye position detector. 2. The position of the center of gravity of the left eye region and the right eye region determined by the eighth means, or the center of the left and right and top and bottom of each of the left eye region and the right eye region determined by the eighth means The eye position detecting apparatus according to claim 1, further comprising ninth means for determining the position of the eye as the eye position. 3. A difference between a luminance pattern of light reflected by an eyeball when the eyeball is illuminated from the front and a luminance of a left-eye area or a right-eye area in the left-eye area and the right-eye area determined by the eighth means. Ninth means for calculating the luminance pattern while moving the luminance pattern, and determining the center position of the luminance pattern when the degree of difference is the smallest as the position of the left eye or the right eye. Item 3. An eye position detecting device according to Item 1. 4. An eye position detecting method for detecting an eye position from a color face image of a person, wherein an average face generated by cutting out a face internal region from a typical face image of the person based on the position of the eyes is provided. A first step of preparing an image in advance, a second step of converting an input color face image of a person into a YIQ color representation to generate a YIQ face image, and a YIQ face generated in the second step. A third step of binarizing the I component image of the image with a threshold value determined based on the pixel value distribution to generate a binarized image; and a binarization generated in the third step. A fourth step of extracting, as a skin area, a connected area located near the center of the face image from the image and larger than a predetermined area; and a skin area existing inside the outermost periphery of the skin area extracted in the fourth step. Multiple connected areas that are not A fifth step of extracting as a candidate eye area; a sixth step of selecting a pair of a candidate left eye area and a candidate right eye area from the plurality of candidate eye areas extracted in the fifth step; A seventh step of generating a partial face image by cutting out a face internal area from the input color face image based on the positions of the left eye candidate area and the right eye candidate area selected in the step; An eighth step of calculating a degree of difference between the partial face image generated in the step and the average face image prepared in the first step; and an eye candidate area extracted in the fifth step. The above-mentioned sixth to eighth steps are executed for all pairs of the left-eye candidate area and the right-eye candidate area, and the pair of the left-eye candidate area and the right-eye candidate area with the smallest degree of difference is defined as the left-eye area and the right-eye area. Decide Position detecting method eyes and having a ninth step. 5. The position of the center of gravity of the left eye region and the right eye region determined in the ninth step, or the center of the left and right and top and bottom of each of the left eye region and the right eye region determined in the ninth step 5. The eye position detecting method according to claim 4, further comprising a tenth step of determining the position of the eye as the eye position. 6. A tenth step in which a brightness pattern of light reflected by the eyeball when the eyeball is illuminated from the front is prepared in advance, and inside the left eye area and the right eye area determined in the ninth step, Calculating the difference between the luminance pattern prepared in advance in step 10 and the luminance of the left eye area or the right eye area while moving the luminance pattern;
The eye position detecting method according to claim 4, further comprising: an eleventh step of determining a center position of the luminance pattern when the degree of difference becomes minimum as a left eye or right eye position. 7. A recording medium recording an eye position detection program for detecting an eye position from a color face image of a person by a computer, wherein the eye position detection program is input to the computer. Convert a human color face image to YIQ color representation
A Q face image is generated, and I of the generated YIQ face image is generated.
The component image is binarized by a threshold value determined based on the pixel value distribution to generate a binarized image. The binarized image is located near the center of the face image from the generated binarized image and has a predetermined area. A large connected region is extracted as a skin region, and a plurality of connected regions that are not skin regions existing inside the outermost periphery of the extracted skin region are extracted as eye candidate regions. A pair of a left-eye candidate region and a right-eye candidate region is selected from all of the selected pairs of the left-eye candidate region and the right-eye candidate region with respect to the position of the left-eye candidate region and the right-eye candidate region. The face internal region is cut out from the color face image to generate a partial face image, and the face internal region is cut out from the generated partial face image and a typical face image of a person based on the position of the eyes to generate the partial face image. Average face image and The dissimilarity is calculated, the calculated allowed the degree of difference becomes minimum eye candidate region and the right-eye candidate region recording medium for recording a position detecting program eye, characterized in pairs to be determined as the left eye region and the right eye region of. 8. The eye position detection program further causes the computer to determine the position of the center of gravity of the determined left-eye area and right-eye area, or the left and right of the determined left-eye area and right-eye area. The recording medium according to claim 7, wherein the position of the upper and lower centers is determined as the position of the eye. 9. The eye position detection program further causes the computer to determine, within the determined left-eye area and right-eye area, a luminance pattern of light reflected by the eyeball when the eyeball is illuminated from the front and a left-eye area or The difference between the luminance of the right eye region and the luminance of the right eye region is calculated while moving the luminance pattern, and the center position of the luminance pattern when the difference is minimum is determined as the position of the left eye or the right eye. A recording medium recording the eye position detection program according to claim 7.