JP4135702B2 - Intercom system - Google Patents

Description

  The present invention relates to an interphone system including a door phone slave unit and an interphone master unit, and more particularly to an interphone system having a function of displaying an image captured by a camera of the door phone slave unit on the interphone master unit.

  An intercom system having a function of displaying an image captured by a camera of a door phone slave unit with an interphone master has been provided. In this type of system, when the subject (visitor) captured by the camera is backlit, the brightness of the entire captured image, that is, the level of the video signal is increased to prevent the subject from being blacked out. The interphone master unit has a function (so-called backlight correction function). For example, as shown in FIG. 13, in an image captured by a camera of a door phone slave unit, a sky with relatively high brightness is located in the upper part in the vertical direction, and a ground with relatively low brightness in the lower part in the vertical direction. It is considered that the face of the visitor who is the subject often overlaps the sky. Therefore, the backlight correction is performed by detecting the luminance of the lower position corresponding to the ground in the captured image and increasing the level of the video signal until the detected luminance is equal to or higher than the optimum value (see FIG. 13B). .

On the other hand, the sampling value of the video signal for one screen is compared with a predetermined value for each of white and black, and the number of sampling values exceeding the predetermined value for white (count value) and the number of sampling values falling below the predetermined value for black ( An interphone system is also proposed in which it is determined whether or not the screen is backlit by comparing the count value), and the level of the video signal is increased when it is determined that the screen is backlit (see Patent Document 1).
Japanese Utility Model Publication No. 7-24757

  However, the backlight is corrected so that the luminance at the bottom of the screen becomes the optimum value, or the backlight is corrected based on the size relationship between the white area and the black area in one screen as in the conventional example described in Patent Document 1. When the backlight correction is performed by determining whether or not to perform the backlight correction, the position of the visitor's face that should be preferentially adjusted to the optimum brightness in the screen is not taken into consideration, so the visitor's face in the backlight corrected image is not taken into consideration. May be too bright or too dark to capture the face of the visitor with appropriate brightness.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an intercom system capable of always capturing and displaying a visitor's face with appropriate brightness.

In order to achieve the above object, the invention of claim 1 includes a door phone slave installed at an entrance outside a dwelling unit, and an interphone master set installed inside the dwelling unit and performing a call between the door phone slaves. The door phone slave unit includes an imaging unit that images a visitor and a call unit that performs a call between the interphone master unit, and the interphone master unit includes a call unit that performs a call between the door phone slave unit, Display means for displaying an image picked up by the image pickup means of the door phone slave unit, and performing image processing on the image picked up by the image pickup means on one of the door phone slave unit or the interphone master unit Image processing means for detecting the position of the face of the visitor, and adjustment means for adjusting the brightness of the image captured by the imaging means so that the position of the face detected by the image processing means has an appropriate brightness. , The picture The processing means includes a face detection template image storage means for storing a face detection template image in advance, a captured image captured by the imaging means of the door phone slave unit, and a face detection stored in the face detection template image storage means. Concentration gradient direction extraction means for extracting density gradient direction images of the template image, density gradient direction image storage means for storing each density gradient direction image extracted by the density gradient direction extraction means, and density gradient direction image storage means Information on the distance between the reference point on the density gradient direction image of the stored face detection template image and each coordinate point to be referenced, and the angle at which the line connecting the reference point and the coordinate point intersects the horizontal axis passing through the coordinate point The shape feature extracting means for extracting the information and the distance and angle information extracted by the shape feature extracting means are classified and described for each value of the density gradient direction of the face detection template image. Shape feature storage means, density gradient direction values at each coordinate point to be referred to in the density gradient direction image of the captured image stored in the density gradient direction image storage means, and shape features stored in the shape feature storage means Based on the voting means for voting the reference point candidate point in the density gradient direction image of the captured image, and the face detecting means for detecting the position of the face based on the voting result obtained by the voting means. Features.

According to the present invention, the image processing unit included in one of the door phone slave unit or the interphone master unit performs image processing on the image captured by the imaging unit of the door phone slave unit, whereby the face of the visitor in the image is displayed. The position of the face of the visitor's face is detected because the adjustment means adjusts the brightness of the image so that the position of the face detected by the image processing means from the image picked up by the image pickup means becomes an appropriate brightness. Regardless of the position on the screen, the face of the visitor can always be captured and displayed with appropriate brightness. Moreover, even if the size of the body and the face detection template image may be the same, the values of the density gradient directions are different, so that votes are not easily concentrated on the position of the body, and the body is not erroneously detected. Further, by using the density gradient direction image, it is possible to realize an image processing means that is resistant to illumination fluctuations.

  According to a second aspect of the present invention, in the first aspect of the invention, the image processing means and the adjusting means are provided in a door phone sub unit.

According to a third aspect of the present invention, in the first aspect of the present invention, the face detecting means changes the size of the face detection template image, and based on the change of the vote value corresponding to the size change, A face size detecting means for extracting the size is provided.
According to the present invention, a face can be detected even if the size of the face is not constant.

According to a fourth aspect of the present invention, in the first aspect of the present invention, the face detection means changes the distance information in the shape feature, and based on the change of the vote value according to the change of the distance information, the size of the face is determined. It is characterized by comprising face size detecting means for detecting the height.

  According to the present invention, a face can be detected even if the size of the face is not constant.

According to a fifth aspect of the present invention, in the invention according to any one of the first to fourth aspects, the face detection means changes the rotation angle of the face detection template image, and the voting value corresponding to the change in the rotation angle is changed. A face rotation angle extracting means for extracting a face rotation angle based on the change is provided.

  According to the present invention, a face can be detected even if the rotation angle of the face is not constant.

According to a sixth aspect of the present invention, in the invention according to any one of the first to fourth aspects, the face detection means changes angle information in the shape feature and is based on a change in vote value in accordance with the change in the angle information. And a face rotation angle extracting means for extracting the face rotation angle.

  According to the present invention, a face can be detected even if the rotation angle of the face is not constant.

According to the first aspect of the present invention, the image processing means provided in one of the doorphone slave unit or the interphone master unit performs image processing on the image picked up by the image pickup means of the door phone slave unit, whereby the visitor in the image Because the adjustment means adjusts the brightness of the image so that the position of the face detected by the image processing means from the images picked up by the image pickup means is adjusted to an appropriate brightness from the image picked up by the image pickup means. The face of the visitor's face can always be captured and displayed at an appropriate brightness no matter where the face is located on the screen , and the size of the body and face detection template image may match. However, since the values of the density gradient directions are different from each other, voting is difficult to concentrate on the position of the body, so that the body is not erroneously detected, and image processing means that is resistant to fluctuations in illumination by using the density gradient direction image. Realized in There is an effect that that.

(Embodiment 1)
1 and 2 show a system configuration diagram and a block diagram of an intercom system according to the present embodiment, respectively. Here, the door phone slave unit 20 is connected to the interphone master unit (hereinafter, abbreviated as “master unit”) 1 by two pairs of signal lines L1 including a two-wire communication line and a two-wire control line.

  The door phone sub unit 20 receives a voltage applied to the call line by the base unit 1 to create an internal operating power source, and performs a two-to-four line conversion between the internal call path and the call line; A voice input / output unit 22 having a microphone and a speaker, a voice processing unit 23 for amplifying a voice signal inputted from the microphone, amplifying a voice signal sent from the master unit 1 through a call line, and the like, A camera unit 24 that has such an individual image sensor and an optical system such as a lens and images a visitor, and a microcomputer as main components, and image processing (described later) on a video signal output from the camera unit 24 ) And frequency-modulating the video signal after image processing, and a switch that is turned on when the visitor presses the call button 28 (see FIG. 1). A display element (light-emitting diode) is supplied with power via the control unit 26 during a call with the operation unit 26 that sends a call signal by changing the line voltage of the call line when it is turned on. ) Is lit. Here, the video signal (hereinafter referred to as “FM video signal”) frequency-modulated to the audio signal (baseband signal) output from the audio processing unit 23 is multiplexed by the power receiving unit 21 and is transmitted through the communication line to the master unit. Sent to 1.

  The main unit 1 applies a DC voltage to the telephone line to supply power to the intercom 20 and feeds the multiplexed audio signal and FM video signal, and a microcomputer. Main unit control unit 3 as a component, audio input / output unit 4 having a microphone and a speaker, an audio signal input from audio input / output unit 4 and an audio signal from doorphone slave unit 20 separated by power supply unit 2 , And a voice processing unit having a function of a voice switch for switching a call direction (base unit 1 → door phone slave unit 20 or master unit 1 ← door phone slave unit 20) by attenuating a voice signal having a lower signal level 5, a display unit 6 composed of a liquid crystal display (LCD), and a video processing unit 7 that demodulates the FM video signal sent from the intercom 20 and outputs the video signal to the display unit 6 to display the video. There are a call switch that is turned on by operation of the call button 11 (see FIG. 1) and a monitor switch that is turned on by operation of the monitor button 12 (see FIG. 1). A base unit operation unit 8 that outputs to the control unit 3 and a power source unit 9 that creates an operating power source for each unit from a commercial power source are provided.

  Next, basic operations of the master unit 1 and the door phone slave unit 20 will be described.

  First, when a visitor presses the call button 28 of the door phone cordless handset 20, the operation unit 26 transmits a call signal from the power receiving unit 21 to the base unit 1 through the telephone line, and the video processing unit 25 and the camera unit 24 further. The video of the visitor that is activated and imaged by the camera unit 24 is transmitted from the power receiving unit 21 to the base unit 1 via the telephone line. In the base unit 1, the base unit control unit 3 that has detected the calling signal controls the voice processing unit 5 to ring the ringing tone from the speakers of the voice input / output unit 4 and also transmits the FM transmitted from the doorphone slave unit 20. The video signal is demodulated by the video processing unit 7 and the video of the visitor is displayed on the display unit 6. When the housekeeper presses the call button 11 of the base unit 1, an operation signal is output from the base unit operation unit 8, and the base unit control unit 3 that receives this operation signal is connected to the voice processing unit 5 and the voice input / output unit 4. A call path is formed between the call lines, and a call with the door phone slave unit 20 becomes possible. When the monitor button 12 is pressed and an operation signal is output from the base unit operation unit 8, the base unit control unit 3 that has received this operation signal is sent from the power supply unit 2 to the door phone sub unit 20 through the call line. Since the camera unit 24 and the video processing unit 25 of the door phone slave unit 20 are activated and the video imaged by the camera unit 24 is sent to the base unit 1, the camera unit 24 of the door phone slave unit 20 is displayed on the display unit 6. The image captured with is displayed.

  Next, image processing and backlight correction processing in the video processing unit 25 of the door phone slave unit 20 that is the gist of the present invention will be described.

  When the camera unit 24 and the video processing unit 25 are activated by pressing the call button 28 as described above, the video processing unit 25 detects the position of the visitor's face in the image captured by the camera unit 24 ( (Hereinafter referred to as “face position detection processing”), the brightness (luminance level) of the image of the detected face position is detected, and if the detected value is below a predetermined threshold value, the entire image Backlight correction processing is performed to increase the brightness.

  First, the face position detection process will be described. FIG. 3 is a block diagram showing a configuration for performing the face position detection process of the video processing unit 25. An image input unit 251 that captures a captured image from the camera unit 24 by A / D conversion, and a face detection template in advance. The stored face detection template image storage means 252 and the face detection template image A and image input means 251 stored in the face detection template image storage means 252 as shown in FIG. The density gradient direction extracting means 253 for extracting the density gradient direction images A ′ and B ′ (see FIGS. 4B and 4A) of the captured image B as shown in FIG. 5, and the extracted density gradient The density gradient direction image storage unit 254 that stores the direction images A ′ and B ′, and the density gradient direction image A ′ of the face detection template image A stored in the density gradient direction image storage unit 254 The distance L between the reference point O and each reference coordinate point (hereinafter referred to as reference point) R, and the line connecting the reference point O and each reference point R intersects the horizontal axis (X axis) passing through the reference point R Shape feature extraction means 255 for extracting information on the angle α to be performed, shape feature storage means 256 for separating and storing the extracted information for each value θ in the density gradient direction of the face detection template image A, and density gradient Based on the density gradient direction value θ at each reference point R ′ of the density gradient direction image B ′ of the captured image B stored in the direction image storage unit 254 and the shape feature information stored in the shape feature storage unit 256. The voting means 257 for voting the reference point candidate points in the density gradient direction image B ′, and the face detecting means 258 for detecting the position of the face based on the voting result obtained by the voting means 257. The

The operation of each means will be described below.
First, the density gradient direction extracting means 253 is used for face detection, which is captured by the camera unit 24 and stored in the captured image B and face detection template image storage means 252 shown in FIG. 5A input by the image input means 251. The density gradient direction images B ′ and A ′ are extracted from the template image A. FIG. 4B shows the density gradient direction image A ′ of the face detection template image A. FIG. 6A shows a density gradient direction image B ′ of the captured image B, respectively. When extracting these density gradient direction images A ′ and B ′, as shown in FIG. 5, a 3 × 3 x-direction Sobel filter 12x and a y-direction Sobel filter 12y are applied to all pixels included in the input grayscale image. The differential value dx in the horizontal axis (X-axis) direction and the differential value dy in the vertical axis (Y-axis) direction in each image are obtained from (Formula 1) and (Formula 2).

dx = (c + 2f + i) − (a + 2d + g) (Formula 1)
dy = (g + 2h + i) − (a + 2b + c) (Formula 2)
(A to h indicate the values a to d and f to h of the target pixel e and eight pixels in the vicinity of the target pixel e as shown in FIG. 7.) At this time, the differential direction value θ Is represented by the following (formula 3).

θ = tan ⁻¹ (dy / dx) (Equation 3)
Here, the density gradient direction images A ′ and B ′ are images having the density gradient direction value θ as a pixel value, and this image is stored in the density gradient direction image means 254.

  The shape feature extraction unit 255 first calculates the center of gravity in the density gradient direction image B of the face detection template image A stored in the density gradient direction image storage unit 254 and sets it as the reference point O. The distance L between the reference point O and the reference point R shown in FIG. 4C and the angle α between the line connecting the reference point O and the reference point R and the horizontal axis β passing through the reference point R are calculated. The shape feature storage unit 256 stores the feature information of the face detection template image A for each angle between the density gradient direction γ and the line β, that is, for each density gradient direction value θ. As shown in FIG. 4A, the face detection template image A used in the present embodiment is created in a form including an average human face characteristic part (eyes, eyebrows, nose, mouth, etc.). Is adopted. The shape feature of the face detection template image A may be stored in the shape feature storage unit 256 in advance.

  Table 1 shows the shape features stored in the shape feature storage means 256, and is in a table format in which the angle α and the distance L are entered for each density gradient direction value θ of each reference point.

The voting unit 257 stores the density gradient direction value θ of the density gradient direction image B ′ of the captured image B shown in FIG. 6A stored in the density gradient direction image storage unit 254 and the shape feature storage unit 256. Using the shape feature of the face detection template image A, a reference point candidate point is obtained from the following (Expression 4) and (Expression 5), and the vote is voted for that point. Then, the face detecting means 258 detects the peak of the vote value from the voting result shown in FIG. 8A, determines the reference point O ′ (o′x, o′y), and determines the position of the face. FIG. 8B shows the relationship with the voting value of the segment range shown in FIG.

ox ′ = L × cos (α) + r′x (Formula 4)
oy ′ = L × cos (α) + r′y (Formula 5)
Thus, according to the face position detection process of the present embodiment, even if the size of the body and the face detection template image may match, the values of the respective density gradient directions are different, so the vote of the body position Are less likely to concentrate, the body is not erroneously detected, and an image processing means (video processing unit 25) that is resistant to variations in illumination can be realized by using a density gradient direction image.

  By the way, in the face position detection process described above, instead of the face detection template image A stored in the face detection template image storage means 252, as shown in FIG. A template image created in this way may be used. That is, since the template image of the average contour of the face is used, the eyes, the eyebrows, the nose, the mouth, and the like that indicate the features of the face are not correctly captured as in backlight, as shown in FIG. 9C. The position of the face in the image B can be detected correctly. FIG. 9B shows the density gradient direction image A ′ extracted by the density gradient direction extraction means 253 from the face detection template image A shown in FIG.

  After detecting the face position as described above, the video processing unit 25 detects the brightness (luminance level) of the range S including the face position as shown in FIG. If the detected value is equal to or smaller than the threshold value, backlight correction processing is performed to increase the brightness (luminance level) of the entire image until the detected value exceeds the threshold value. As a result, the face of the visitor H can be displayed with appropriate brightness in the backlight-corrected image as shown in FIG.

  Thus, according to the present embodiment, the image processing unit 25, which is an image processing unit included in the door phone slave unit 20, performs image processing on the image captured by the camera unit 24 of the door phone slave unit 20. The video processing unit 25 as an adjusting unit detects the position of the face of the visitor in the image, and adjusts so that the position of the face detected by the video processing unit 25 from the images captured by the camera unit 24 has an appropriate brightness. Since the brightness of the visitor is adjusted, the visitor's face can always be captured and displayed at an appropriate brightness regardless of the position of the visitor's face on the screen.

(Embodiment 2)
The present embodiment is characterized by face position detection processing in the video processing unit 25 of the door phone slave unit 20, and other configurations and operations are the same as those in the first embodiment, and thus illustration and description thereof are omitted.

Since the basic configuration of the video processing unit 25 is the same as that of the first embodiment shown in FIG. 3, reference is made to FIG. The video processing unit 25 according to this embodiment is characterized in that a face size detecting unit 258a is provided in the face detecting unit 258 as shown in FIG.
The face size detecting means 258a performs voting by the voting means 257 while changing the size of the face detection template image, and as shown in FIG. Therefore, it is means for detecting the size that maximizes the value as the face size.

As described above, this embodiment has an advantage that the face can be detected even if the size of the face is not constant.
(Embodiment 3)
The present embodiment is characterized by face position detection processing in the video processing unit 25 of the door phone slave unit 20, and other configurations and operations are the same as those in the first embodiment, and thus illustration and description thereof are omitted.
In the second embodiment, the face size is detected by changing the size of the face detection template image. However, the present embodiment detects the face size by changing the distance information in the shape feature described above. There is a feature. Since the basic configuration of the video processing unit 25 is the same as that of the first embodiment shown in FIG. 3, refer to FIG. 3, and in the face detection means 258, the face size detection means 258 a ( 11 (a)) is provided.

  Thus, the face size detection means 258a (FIG. 11A) of the present embodiment changes the distance information by changing the magnification s with respect to the distance L in the shape feature, and the voting means 257 causes the density gradient direction image storage means. The density gradient direction value θ of the density gradient direction image B ′ of the captured image B shown in FIG. 6A stored in the H.254 and the shape feature of the face detection template image A stored in the shape feature storage unit 256. Is used to obtain a reference point candidate point according to the following (formula 6) and (formula 7), and vote for that point in the voting space. Then, as shown in FIG. 11B, since the peak of the vote value is obtained from the vote result for each face size, the size with the maximum value is detected as the face size.

ox ′ = s × L × cos (α) + r′x (Formula 6)
oy ′ = s × L × cos (α) + r′y (Expression 7)
s: magnification of distance (Embodiment 4)
The present embodiment is characterized by face position detection processing in the video processing unit 25 of the door phone slave unit 20, and other configurations and operations are the same as those in the first embodiment, and thus illustration and description thereof are omitted. The basic configuration of the video processing unit 25 is the same as that of the first embodiment shown in FIG.

  Then, as shown in FIG. 12A, the video processing unit 25 in the present embodiment includes a face size detection unit 258a in the face detection unit 258 (any of the face size detection units 258a in the second and third embodiments). And a feature that the face rotation angle output means 258b is provided.

  On the other hand, the face rotation angle extraction means 258b is for corresponding to the picked-up image B captured with the face tilted as shown in FIG. 12B, and the voting means while rotating the face detection template image A. By voting by 257, as shown in FIG. 12C, a peak of the voting value is obtained from the voting result for each rotation angle. Therefore, the rotation angle that maximizes the value is extracted as the rotation angle of the face. Means.

This embodiment has an advantage that the face can be detected even if the rotation angle of the face is not constant.
(Embodiment 5)
Although the face rotation angle extraction means 258b in the fourth embodiment rotates the face detection template image A, the face rotation angle detection means 258b in the present embodiment converts angle information in the shape feature described above. There is a feature in the point. Since the basic configuration of the image processing apparatus according to the present embodiment is the same as that of the first embodiment shown in FIG. 1, refer to FIG. 3, and in the face detection unit 258 as in the fourth embodiment, for example, the embodiment 3 is provided with a face rotation angle output means 258b in addition to the face size detection means 258a used in FIG. 3 (see FIG. 12A).
Thus, the face rotation angle extraction unit 258b of the present embodiment changes the angle information in the shape feature described above, and the density of the density gradient direction image stored in the density gradient direction image storage unit 254 by the voting unit 257. Gradient direction value θ (when rotation angle is φ and density gradient direction of captured image B is θ ′, θ = θ′−φ) and face detection template image stored in shape feature storage means 7 Using the shape feature of A, a reference point candidate point is obtained from (Equation 8) and (Equation 9) below, and a vote is given to that point in the voting space. Then, as shown in FIG. 12C, since the peak of the vote value is obtained from the vote result for each rotation angle, the rotation angle at which the maximum value is obtained is detected as the face rotation angle.

ox ′ = L × cos (α + φ) + r′x (Expression 8)
oy ′ = L × cos (α + φ) + r′y (Equation 9)
φ: Rotation angle (change the angle information by changing the value of this rotation angle)
In the first to fifth embodiments, the image processing unit that performs the face position detection process and the adjustment unit that performs the backlight correction process are configured by the video processing unit 25 of the door phone slave unit 20. The image processing unit and the adjustment unit may be configured as described above.

1 is a system configuration diagram of Embodiment 1. FIG. It is a block diagram same as the above. It is a circuit block diagram of the image | video process part of the door phone sub unit in the same as the above. It is a figure for operation | movement description of a video processing part in the same as the above, Comprising: (a) is an image figure of the template image for face detection, (b) is an image figure of the gradient direction image of the template image for face detection, (c) is a face detection It is a shape feature explanatory view of a template image. FIG. 6 is a captured image diagram for explaining the operation of the video processing unit in the first to fifth embodiments. It is a figure for operation | movement description of a video processing part in the same as the above, Comprising: (a) is an image figure of the gradation gradient direction image of a captured image, (b) is a figure characteristic explanatory drawing of a captured image. It is explanatory drawing of the Sobel filter used for the same as the above. It is explanatory drawing of the vote result of a reference point candidate point same as the above. It is a figure for operation | movement description of the image | video process part in the same as the above, Comprising: (a) is an image figure of the template image for face detection, (b) is an image figure of the gradation gradient direction image of the template image for face detection, (c) is a detection object. It is an image figure of the captured image which imaged the face of. It is a figure for operation | movement description of a video processing part in the same as the above, Comprising: (a) is an image figure of the captured image before backlight correction | amendment, (b) is an image figure of the captured image after backlight correction | amendment. (A) is a block diagram of the face detection means used in Embodiments 2 and 3, and (b) is an explanatory diagram of the relationship between the face size and the vote value in the face size detection of the face detection means of the above. (A) is the block diagram of the face detection means used for Embodiment 4, 5, (b) is the image figure of the picked-up image which imaged the face of the detection object same as the above, (c) is the face rotation angle in the face rotation detection same as the above. It is relationship explanatory drawing with a vote value. It is explanatory drawing of the backlight correction in a prior art example.

Explanation of symbols

1 Interphone main unit 20 Door phone slave unit 24 Camera unit 25 Video processing unit

Claims (6)

It is composed of a doorphone slave unit installed at the entrance outside the dwelling unit and an interphone master unit installed inside the dwelling unit to communicate with the doorphone slave unit. And a communication means for making a call with the interphone master unit. The interphone master unit displays the image captured by the call means for making a call with the door phone slave unit and the image pickup means of the door phone slave unit. Image processing means for detecting the position of the face of the visitor in the image by performing image processing on the image picked up by the image pickup means on one of the door phone slave unit or interphone master unit, and an image and an adjustment means for adjusting the brightness of an image captured by the imaging means so that the position of the detected face in the processing means is proper brightness, said image processing means, the template image for advance face detection Respectively stored face detection template image storage means, a captured image taken by the image pickup means of the door phone slave unit, and a density gradient direction image of the face detection template image stored in the face detection template image storage means are respectively extracted. Density gradient direction extraction means, density gradient direction image storage means for storing each density gradient direction image extracted by the density gradient direction extraction means, and density gradient of the face detection template image stored in the density gradient direction image storage means A shape feature extracting means for extracting information on the distance between the reference point on the direction image and each coordinate point to be referred to and the angle at which the line connecting the reference point and the coordinate point intersects the horizontal axis passing through the coordinate point; Shape feature storage means for classifying and storing the distance and angle information extracted by the extraction means for each value of the density gradient direction of the face detection template image, and density gradient direction image recording In the density gradient direction image of the captured image, based on the value of the density gradient direction at each coordinate point to be referred to in the density gradient direction image of the captured image stored in the means and the shape feature stored in the shape feature storage means. An intercom system comprising: voting means for performing voting processing on reference point candidate points; and face detection means for detecting a face position based on a voting result obtained by the voting means .   2. The intercom system according to claim 1, wherein the image processing means and the adjusting means are provided in a door phone slave unit. The said face detecting means changes the size of the face detection template image, that based on the change in the voting value corresponding to the magnitude of the change, and a face size detecting means for extracting the size of the face The intercom system according to claim 1 . The face detection means includes face size detection means for changing the distance information in the shape feature and extracting the face size based on a change in vote value according to the change in the distance information. The intercom system according to claim 1, wherein: The face detection means includes face rotation angle extraction means for changing the rotation angle of the face detection template image and extracting the face rotation angle based on a change in the vote value corresponding to the change in the rotation angle. The intercom system according to any one of claims 1 to 4, wherein the intercom system is provided. The face detection means includes face rotation angle extraction means that changes angle information in the shape feature and extracts a face rotation angle based on a change in vote value in accordance with the change in the angle information . intercom system according to any one of claims 1 to 4, characterized in.