JPH1172697A - Method and device for detecting observer's point position - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect the accurate position of an observer's point without being brought into contact by stably picking up the pupil of an observer by a simple device constitution using a single illuminator whose optical axis is also used as that of an image pickup unit at a visual point position detector by picking up a pupil used at a visual point follow-up type stereoscopic display device or the like. SOLUTION: This device possesses an infrared light illuminator 101 illuminating the face surface of an observer (person to be detected) 130, the image pickup unit 102 performing the image pickup of the face of the observer 130 illuminated and a picture processor applying a specified processing so as to detect the pupil position of the observer 130. Spatial differential calculation is applied to a picture stored in a picture memory 104 by a picture differential arithmetic part 105 at the picture processor 120. Then, the absolute value of the output picture of the picture differential arithmetic part 105 is calculated for each picture element by a picture absolute value arithmetic part 106, and a result is binarized for each picture element by a picture binarization arithmetic part 107, so that the observer's point position is detected from the coordinate positions of the bright part of the picture binarized by the observer's point position arithmetic part 108.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for detecting a viewpoint position of an observer, and more particularly, to a technique of always detecting the viewpoint position of the observer in a non-contact manner, and constantly giving the observer a position in accordance with the detected viewpoint position. The present invention relates to a viewpoint position detection method and a detection device suitable for use in a stereoscopic display device or the like that presents a correct stereoscopic image.

[0002]

2. Description of the Related Art In recent years, there has been an increasing demand for stereoscopic display, and various types of stereoscopic display devices have been studied. Recently, particularly, a lenticular screen has been used to display images having binocular parallax on the left and right eyes. A stereoscopic display device that realizes stereoscopic vision by spatially separating and presenting the images has attracted attention. This device has the advantage that no special glasses need to be worn. However, this device has a problem that an area where stereoscopic vision is possible is limited to about the distance between human eyes, and an observer is required to have an unnatural posture in which the observer always keeps his head fixed for observation. In order to solve such a problem, there is known a viewpoint-following stereoscopic display device that constantly detects a viewpoint position of an observer and always presents a correct stereoscopic image to the observer according to the detected viewpoint position. (See, for example, JP-A-1-107247,
No. 7123). In such a conventional viewpoint-following type stereoscopic display device, it is necessary to attach a magnetic sensor, a reflector, or the like to the observer in order to detect the viewpoint position, which gives the observer a sense of incongruity and special glasses. The advantage of the stereoscopic display device using the lenticular screen that it is not necessary to mount the lenticular screen is reduced. In addition, when used by a large number of unspecified observers, the use of a wearable viewpoint detection device causes a problem in hygiene.

As a viewpoint position detecting device capable of solving such a problem, for example, Japanese Patent Laid-Open Publication No. Hei 7-61256 discloses an image of an observer's face, and extracts only the pupil from the captured image. A pupil detection device that detects a position without contact has been proposed. In this pupil detection device, pupil detection is performed using the reflection characteristics of a human pupil.
The human pupil has a property that, when illuminated by an infrared light illuminating device having the optical axis coaxial with the imaging device, the reflectance of the pupil becomes larger than the reflectance of other portions of the face, and the optical axis of the pupil is not aligned with the optical axis. When illuminated by a coaxial infrared light illuminator, there is a property that the reflectance of the pupil becomes smaller than the reflectance of other parts of the face. Therefore, by taking the difference between the images obtained using the illumination devices having different arrangements, it becomes possible to extract only the human pupil.

FIG. 13 is a diagram conceptually showing a part of the technique described in Japanese Patent Application Laid-Open No. 7-61256. The pupil detection device includes an imaging unit 1 and a first illumination lamp 2 and a second illumination lamp 3 that illuminate an eyeball 17. The first illumination lamp 2 has its optical axis arranged coaxially with the imaging unit 1, and the second illumination lamp 3 has its optical axis arranged non-coaxially with the imaging unit 1. In this pupil detection device, first, only the first illumination lamp 2 is turned on by the illumination light emission control unit 6, and an image in which the pupil becomes bright is captured. The image picked up by the image pickup unit 1 is converted into a digital image by the A / D converter 8 and then stored in the image memory 9. Subsequently, only the second illumination lamp 3 whose optical axis is non-coaxial with the imaging unit 1 is turned on by the illumination light emission control unit 6, and an image in which the pupil becomes dark is captured. The image picked up by the image pickup unit 1 is converted into a digital image by the A / D converter 8 in the same manner as in the case of the first illumination lamp 2, and is stored in the image memory 9. Further, the pupil of the eyeball 17 is extracted by calculating the difference between the respective images illuminated by the first illumination lamp 2 and the second illumination lamp 3 held in the image memory 9 by the pupil extraction unit 10. .

[0005]

In this conventional pupil detection apparatus, an infrared light illuminating apparatus and an imaging apparatus having an optical axis coaxial with the imaging apparatus are used to extract a pupil portion from an image of the face of the subject. In order to obtain two types of illumination means of an infrared light illuminating device having an optical axis non-coaxial, and to separately obtain captured images illuminated by these two types of illumination means, time-division is required. It is necessary to provide a means for switching and lighting the lighting device, and there is a problem that the device configuration is complicated.

An object of the present invention is to extract an observer's pupil with a simple device configuration using only a single illumination device whose optical axis is coaxial with an imaging device, and to accurately detect a viewpoint position. It is an object of the present invention to provide a viewpoint position detecting method and a detecting device.

[0007]

A viewpoint position detecting method according to the present invention illuminates and captures an image of a face of a person to be detected, differentiates the face image value of the person obtained by the imaging, and calculates a differential value of the face image value. The absolute value of the calculation result is obtained, and the absolute value is compared with a threshold value, and a portion having a value larger than the threshold value is detected as the viewpoint position of the subject. in this case,
In addition to calculating the absolute value of the operation result of the differential operation, calculating the sum of the luminances of the operation target values, dividing the operation results by each corresponding value, comparing the division result with the threshold, and comparing the division result with the threshold. A location having a large value may be detected as the viewpoint position of the subject. Further, when differentiating the image, it is preferable that the interval between the calculation target values is made to match the pupil diameter.

Further, the viewpoint position detecting device of the present invention is obtained by an image pickup means for picking up an image of the face of a person to be detected, an illuminating means having an optical axis substantially coaxial with the image pickup means, and the image pickup means. Storage means for accumulating the face image of the subject, image differential operation means for performing a spatial differential operation on the image stored in the storage means, and an image for obtaining the absolute value of the operation result of the image differential operation means An absolute value calculating means, an image binarizing calculating means for binarizing the calculation result of the absolute value calculating means, and a means for detecting a viewpoint position of the subject from the calculation result of the image binarizing calculating means. Prepare. In this case, the normalization factor calculating means for calculating the sum of the luminances of the calculation target pixels in the image differentiation calculating means, and the calculation results obtained by the image absolute value calculating means and the normalizing factor calculating means are respectively associated with each other. It is also possible to provide an image dividing means for dividing for each pixel, and to binarize the operation result of this image dividing means. It is preferable that the image differentiating means has a configuration in which the distance between the pixels to be calculated is the pupil diameter.

In the method and apparatus for detecting a viewpoint position according to the present invention, the pupil image of the subject is emphasized from other parts of the image by differentiating the captured image of the face to obtain a pupil image signal having a high signal-to-noise ratio. Thus, the pupil of the observer can be stably extracted with a simple device configuration using only a single illumination device, and the non-contact and accurate viewpoint position detection can be performed.
In particular, by using the illumination unit and the imaging unit whose optical axis is substantially coaxial with the imaging unit, further, by performing a spatial image differentiation operation with the pupil diameter as the interval between calculation target pixels,
More accurate and accurate viewpoint position detection can be realized.

[0010]

Next, embodiments of the present invention will be described with reference to the drawings. (First Embodiment) FIG. 1 is a conceptual configuration diagram showing a first embodiment of a viewpoint position detecting device according to the present invention. Observer 130
Infrared light illuminating device 101 that illuminates the face of (detected person)
And the observer 13 illuminated by the infrared light illumination device 101
Imaging apparatus 102 for imaging the face of “0” and observer 13
An image processing apparatus 120 that performs a predetermined process for detecting a pupil position of 0 is provided. The image pickup device 102 is arranged so that the optical axis is coaxial with the infrared light illuminating device 101 by the half mirror 103, and can image the pupil of the observer 130 brightly. Here, the infrared light illumination device 10
A light emitting diode having a wavelength of 850 nm is used as one light source.

The image pickup device 102 uses a solid-state image pickup device, and has a built-in wavelength filter for blocking light having a wavelength of 760 nm or less so as not to be affected by an external illumination device for visible light such as a room light. It is. The imaging device 102 outputs a monochrome / analog video signal of the NTSC standard, and has a horizontal frequency and a vertical frequency of 1 respectively.
5.75 kHz and 60 Hz. This imaging device 102
Has an imaging distance of 2 m, and the lens magnification is adjusted such that the resolution of the image of the observer 130 is 1 mm at this imaging distance.

The image processing device 120 is an image pickup device 102
Memory 1 for storing the captured image of the observer 130 by the camera
04, an image differential operation unit 105 for performing a spatial differential operation on the image stored in the image memory 104, an image absolute value operation unit 106 for obtaining an absolute value of the result of the image differential operation unit 105, and an image absolute value operation Binarization operation unit 10 for binarizing the result of unit 106 and extracting the pupil of observer 130
7. A viewpoint position calculation unit 108 for detecting the viewpoint position of the observer 130 from the result of the image binarization calculation unit.

In this configuration, the image memory 10
Reference numeral 4 denotes an analog video signal output from the imaging device
The image is converted into a digital image having a resolution of 512 × 512 pixels by a D converter, and a captured image of the face of the observer 130 is stored for each field, that is, 60 times per second. Then, the image stored in the image memory 104 is stored in the image differential operation unit 10.
5, a spatial image differential operation is performed. Thereafter, an absolute value is calculated for each pixel by the image absolute value calculation unit 106, and a binarization process is performed for each pixel by the image binarization calculation unit 107 to extract a pupil. After that, the viewpoint position of the observer is detected by the viewpoint position calculation unit 108 from the coordinate position of the extracted pupil. Next, various calculations and viewpoint position calculation in the image processing device 120 are sequentially performed by a computer in synchronization with the vertical synchronization signal of the imaging device 104, and the viewpoint position of the observer 130 is output from the image processing device 120 60 times per second. .

Next, details of the viewpoint position detecting operation in the viewpoint position detecting device having the above configuration will be described. First,
The operation principle will be described. FIG. 2 is a diagram illustrating an image captured by the imaging device of the present invention. FIG. 2A is an image of the observer obtained by the imaging device 102 in FIG. 1, and FIG. 2B is a horizontal luminance distribution of an image between line segments AB in FIG. 2A. As shown in FIG.
1 is imaged brighter than the face portion 202.

FIG. 3 is a diagram for explaining the principle of the image processing device in the viewpoint position detecting device according to the present invention. The image obtained by the imaging device 102 is recorded in the image memory 104 as an image I (x, y). Here, x and y represent the horizontal and vertical coordinate positions of the captured image. FIG.
(A) is a horizontal luminance distribution of the observer's image recorded in the image memory 104, and corresponds to the one shown in FIG. As described above, the image differentiation operation unit 105 performs the image differentiation operation on the image I (x, y) recorded in the image memory 104. As shown in FIG. 3A, the image differentiation operation is performed by calculating the difference I (x + dx / 2, y) -I (x-dx / 2, y) between pixels separated by the number of pixels dx corresponding to the pupil diameter 303. ). By performing this image differentiation operation, FIG.
As shown in (b), the pupil image 30
Only one can be emphasized. After this image differentiation operation,
By performing the image absolute value calculation by the image absolute value calculation unit 106, the component having a negative value in FIG. 3A is converted into a positive value as shown in FIG. Further, after the image absolute value processing, the image binarization processing unit 107 performs processing shown in FIG.
By performing the image binarization operation of the threshold value 304 on the image of, only the pupil image can be extracted as shown in FIG.

FIG. 4 is a diagram for explaining the operation of the viewpoint position calculation unit 108 in the viewpoint position detection device according to the present invention. In order to apply the viewpoint position detecting device of the present invention to a viewpoint-following stereoscopic display device, it is necessary to detect a viewpoint position 403 corresponding to the midpoint of the pupil portion 402 of both eyes in the observer image 401. In the present embodiment, the pupil extraction image 4
The viewpoint position 403 is detected by calculating the midpoint between the horizontal minimum coordinates 406 and the horizontal maximum coordinates 407 of the pupil portion 405 extracted in 04.

An example will be described in which the face of the observer is actually photographed using the above viewpoint position detecting device, and the viewpoint position of the observer is detected. FIG. 5A is a diagram illustrating a luminance distribution of an image of the face of the observer. In the image shown in FIG. 5A, a pupil image 501 and a face image 502 are respectively captured.
A differential operation was performed on the image shown in FIG. 5A by calculating the difference between pixels separated by 5 pixels corresponding to the diameter of the pupil image 501. FIG. 5B shows the result of performing the image absolute value calculation after the image differentiation calculation. In FIG. 5B, the luminance ratio between the pupil image 501 and the face image 502 is 5: 1, and these can be separated by the image binarization operation. Here, when the threshold value 503 is set to “25” in FIG. 5B and the binarization process is performed, only the pupil image 501 can be completely extracted as shown in FIG. 5C.

FIG. 6 is a diagram showing the result of detecting the observer's viewpoint position using the viewpoint position detecting device of the present invention. The horizontal minimum coordinate 601 and the horizontal maximum coordinate 602 of the pupil extraction image are respectively detected 60 times per second by the viewpoint position detecting device of the present invention, and the viewpoint position 603 is stably changed from these middle points to 6 points per second.
0 detections were possible.

An example in which the viewpoint position detecting device of the present invention is applied to a viewpoint-following three-dimensional display device will be described. FIG. 7 is a diagram showing the overall configuration of a viewpoint-following stereoscopic display device using the viewpoint position detection device of the present invention. In the present viewpoint tracking type stereoscopic display device, an observer 706 observes a stereoscopic image through a lenticular screen 705. In FIG. 7, the viewpoint position detection device shown in FIG. An optical system control unit 702 is connected to the viewpoint position detecting device 701, and a viewpoint tracking optical system mechanism unit 703 in the liquid crystal projector 704 is controlled according to the viewpoint position of the observer 706. When the viewpoint position detecting device according to the present embodiment described above is applied to a viewpoint-following type stereoscopic display device, it is possible to accurately detect the viewpoint position in a non-contact manner, and to provide a stereoscopic display device with less burden on an observer. Was completed.

(Second Embodiment) FIG. 8 is a view showing a second embodiment of the viewpoint position detecting device of the present invention. This second
In the embodiment, the image memory 104, the image differential operation unit 105, and the image absolute value operation unit 10 in the configuration of the image processing device 120 (see FIG. 1) in the first embodiment described above.
6. The image binarization calculation unit 107 and the viewpoint position calculation unit 108 are the same, but in addition to this configuration, the normalization factor calculation unit 8
01, an image division operation unit 802 is provided. The normalization factor calculation unit 801 calculates the sum of the luminances of the calculation target pixels in the image differentiation calculation unit 105. The image division calculation unit 802 calculates the calculation result of the image absolute value calculation unit 106 by the normalization factor calculation unit 801. The calculation result is divided for each pixel.
Since the configuration of the parts other than the image processing apparatus 120 is the same as that of the first embodiment, the same reference numerals as those in FIG. 1 are used and the description thereof is omitted.

The principle of operation of the second embodiment will be described in detail. 9 and 10 are diagrams for explaining the principle of the image processing device in the viewpoint position detecting device according to the present invention. An image of the observer obtained by the imaging device 102 is obtained as shown in FIG. 2 similarly to the first embodiment of the present invention, and this image is recorded in the image memory 104 as an image I (x, y). You. Here, x and y represent the horizontal and vertical coordinate positions of the captured image. FIG. 9A shows a horizontal luminance distribution of an observer image recorded in the image memory 104, and corresponds to the one shown in FIG. An image differentiation operation is performed by the image differentiation operation unit 105 on the image I (x, y) recorded in the image memory 104. As shown in FIG. 9A, the image differentiation operation is performed by calculating the difference I (x + dx / 2, y) -I between pixels separated by the number of pixels dx corresponding to the pupil diameter 903.
(X-dx / 2, y). By performing this image differentiation operation, only the pupil image 901 can be enhanced with respect to the face image 902 as shown in FIG. 9B. Thereafter, the image absolute value calculation unit 106 performs the image absolute value calculation, so that the component having a negative value in FIG. 9A is positive as shown in the image 901 after the absolute value processing in FIG. Is converted to the value of

Further, the normalization factor of the pixel to be differentiated is obtained by performing the operation I (x + dx / 2, y) + I (x-dx / 2, y) by the normalization factor operation unit 801. Normalization factor calculator 8
The normalization factor calculated by 01 is shown as a normalization factor 905 in FIG. Here, the image shown by the absolute value processing image 904 in FIG. 9C obtained by the image absolute value calculation unit 106 is divided by the normalization factor 905 obtained by the normalization factor calculation unit 801 into an image division operation unit 802. Performs the image division operation to perform the normalization process. With this processing, a constant value can be always maintained even if the luminance of the pupil image changes. FIG. 9A shows the image after the normalization processing.
After the normalization processing, the binarization processing unit 107 performs the processing shown in FIG.
By performing the binarization operation of the threshold value 906 on the image of, a pupil image can be extracted as shown in FIG.

The observer's face was photographed by using the above viewpoint position detecting device, and the observer's viewpoint position was detected. FIG.
(A) is a figure which shows the brightness distribution of the image which image | photographed the observer's face. The pupil image 100 is included in the image shown in FIG.
1 and the face image 1002 are respectively imaged. FIG.
An image differentiation operation was performed on the image shown in (a) by calculating the difference between pixels separated by 5 pixels corresponding to the diameter of the pupil image 1001. Furthermore, the normalization factor was calculated by taking the sum of the pixels separated by 5 pixels. FIG. 11B shows the result of performing the image absolute value calculation after the image differentiation calculation. FIG. 11C shows a graph comparing the result of FIG. 11B with the normalization factor. In FIG. 11B, there is a difference in luminance between the right eye pupil image 1003 and the left eye pupil image 1004. Therefore, when the normalization processing is performed by dividing the differential absolute value processed image 1005 in FIG. 11C by the normalization factor 1006, the right eye pupil image 1003 and the left eye pupil image are obtained as shown in FIG. The difference between the luminance of the pupil image and that of the face image can be easily separated by binarization processing. Here, in FIG.
When the image binarization operation is performed by setting 7 to 0.25, FIG.
As shown in (b), only the pupil image 1001 could be completely extracted.

As a result of detecting the viewpoint position by the viewpoint position detecting device of the present invention, the horizontal minimum coordinate and the horizontal maximum coordinate of the pupil extraction image are detected 60 times per second as in the first embodiment, and the stable point is obtained from these middle points. The viewpoint position could be detected 60 times per second. Further, the overall configuration of the viewpoint-following stereoscopic display device using the viewpoint position detection device of the present embodiment is the same as that of the first embodiment. The viewpoint position detecting device of the present embodiment is indicated by reference numeral 701 in FIG. When the viewpoint position detecting device according to the present embodiment described above is applied to a viewpoint-following type stereoscopic display device, it is possible to accurately detect the viewpoint position in a non-contact manner, and to provide a stereoscopic display device with less burden on an observer. Was completed.

In the first and second embodiments, the position of the illuminating device may not be exactly coaxial, but may be arranged near the optical axis as long as the pupil of the observer can be imaged brightly. Also, in order to remove noise due to reflected light from specular objects such as glasses, a polarizing plate or 1 /
A four-wave plate or the like may be inserted. Further, the imaging device is N
In addition to the TSC standard, various standards such as the PAL standard and the SECAM standard may be used, and those having higher horizontal and vertical frequencies may be used. The image memory may have various resolutions. Further, an image differentiation operation unit, an image absolute value operation unit, an image division operation unit, an image 2
The binarization processing in the binarization operation unit is performed by a computer,
A dedicated digital logic circuit or an analog operation circuit using an operational amplifier may be used. The viewpoint position detection in the viewpoint position detection unit may use a dedicated digital logic circuit in addition to the computer.

On the other hand, the above-mentioned I (x + dx / 2, y)-
In addition to I (x-dx / 2, y), I (x, y) -I (x-dx, y)
Or I (x, y) -I (x + dx, y) or the like, or I (x, y) -I (x-dx, y) + I (x
, y) -I (x + dx, y). In addition, the image differentiation operation may be performed together with a vertical operation such as I (x, y + dx / 2) -I (x, y-dx / 2). The image differentiation operation in the vertical direction is performed in the same manner as in the horizontal direction.
In addition to (x, y + dx / 2) -I (x, y-dx / 2), I (x, y)-
I (x, y-dx) or I (x, y) -I (x, y + dx) may be used, or I (x, y) -I (x, yd) obtained by superimposing them.
x) + I (x, y) -I (x, y + dx) may be used. Further, when the image differentiation operation is performed only in the horizontal direction, the pupil image may be extracted for each scanning line using the image memory as a line memory. Further, the viewpoint detection in the viewpoint position detection unit may use the image gravity center calculation.

[0027]

As described above, according to the present invention, the image of the subject obtained by the imaging means is subjected to an image differentiation operation, whereby the pupil image of the subject is emphasized from other parts of the image.
By obtaining a pupil image signal with a high signal-to-noise ratio, the pupil of the observer can be stably extracted with a simple device configuration using only a single illumination device whose optical axis is coaxial with the imaging device. Accurate viewpoint position detection can be performed. in this case,
By using a spatial image differential calculating means in which the distance between the calculation target pixels is set to the pupil diameter, position detection can be realized with higher precision and accuracy.

[Brief description of the drawings]

FIG. 1 is a conceptual configuration diagram showing a configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a diagram showing an image picked up by the image pickup device of the present invention.

FIG. 3 is a diagram for explaining the principle of an image processing device in the viewpoint position detecting device of the present invention.

FIG. 4 is a diagram for explaining the operation of a viewpoint position calculation unit in the viewpoint position detection device of the present invention.

FIG. 5 is a diagram illustrating a luminance distribution of an image obtained by photographing the face of an observer.

FIG. 6 is a diagram showing a result of detecting a viewpoint position of an observer using the viewpoint position detection device of the present invention.

FIG. 7 is a diagram showing an overall configuration of a viewpoint-following stereoscopic display device using the viewpoint position detection device of the present invention.

FIG. 8 is a diagram showing a second embodiment of the present invention.

FIG. 9 is a first diagram illustrating the principle of an image processing device in a viewpoint position detecting device according to the present invention;

FIG. 10 is a second diagram illustrating the principle of the image processing device in the viewpoint position detecting device according to the present invention;

FIG. 11 is a first diagram illustrating a luminance distribution of an image obtained by photographing the face of the observer.

FIG. 12 is a second diagram illustrating a luminance distribution of an image of the face of the observer.

FIG. 13 is a conceptual diagram showing a part of an example of a conventional viewpoint position detecting device.

[Explanation of symbols]

 Reference Signs List 101 infrared illumination device 102 imaging device 103 half mirror 104 image memory 105 image differentiation operation unit 106 image absolute value operation unit 107 image binarization operation unit 108 viewpoint position calculation unit 120 image processing device 701 viewpoint detection device 702 optical system control unit 703 Viewpoint tracking optical system mechanism 704 Liquid crystal projector 705 Lenticular screen 706 Observer 801 Normalization factor calculator 802 Image division calculator

Claims (6)

[Claims] 1. An image of a face of a person to be detected is illuminated and imaged, a face image value of the person to be detected obtained by the imaging is differentiated, and an absolute value of the calculation result is obtained. A viewpoint position detection method, wherein a position having a value greater than the threshold value is detected as a viewpoint position of the subject, as compared with a threshold value. 2. A method for illuminating and imaging a face of a person to be detected, performing a differential operation on the face image of the person to be detected obtained by the imaging, and obtaining an absolute value of the operation result.
Calculate the sum of the luminance of the calculation target values, divide the calculation results by each corresponding value, compare the division result with a threshold, and find a location of a value larger than the threshold of the subject. A viewpoint position detection method characterized by detecting as a viewpoint position. 3. The viewpoint position detecting method according to claim 1, wherein, when differentiating the image, an interval between calculation target values is made to match a pupil diameter. 4. An image pickup means for picking up an image of a face of a person to be detected,
Illumination means having an optical axis substantially coaxial with the imaging means;
Storage means for accumulating the face image of the subject obtained by the imaging means; image differential operation means for performing a spatial differential operation on the image stored in the storage means; and operation of the image differential operation means Image absolute value calculation means for obtaining an absolute value of the result; image binarization calculation means for binarizing the calculation result of the absolute value calculation means; A viewpoint position detecting device, comprising: means for detecting a viewpoint position. 5. An image pickup means for picking up an image of a face of a person to be detected,
Illumination means having an optical axis substantially coaxial with the imaging means;
Storage means for accumulating the face image of the subject obtained by the imaging means; image differential operation means for performing a spatial differential operation on the image stored in the storage means; and operation of the image differential operation means Image absolute value calculating means for calculating the absolute value of the result, normalizing factor calculating means for calculating the sum of the luminances of the calculation target pixels in the image differential calculating means, and the image absolute value calculating means and the normalizing factor calculating means. Image division means for dividing the obtained operation results for each corresponding pixel, and calculating the operation result of the image division means by 2
A viewpoint position detecting apparatus, comprising: an image binarization operation unit for converting into a value; and a unit for detecting a viewpoint position of the subject from the operation result of the binarization operation unit. 6. The viewpoint position detecting device according to claim 4, wherein said image differentiation calculating means sets a distance between calculation target pixels to a pupil diameter.