KR19980023002A - Talker tracking method and apparatus for video phone - Google Patents

Abstract

The present invention relates to a method and an apparatus for tracking a speaker in a video telephone using an infrared camera, which separates the background and the speaker according to the size and direction of the boundary pixel extracted from the input image and moves the camera through the central tracking. .

The present invention assigns a specific value to the corresponding boundary direction bit according to the boundary pixel and boundary line direction extracted from the input infrared image, extracts the dialogue image, estimates the movement through center point tracking, and moves the camera accordingly.

Therefore, the present invention tracks the dialog by separating the background and the dialog according to the register value by allocating the value of the register to be allocated to each input image by using the size and direction of the boundary pixel extracted from the infrared input image. The performance is improved.

Description

TARGET TRACKING METHODE AND DERICE FOR VIDEO PHONE

The present invention relates to a method and apparatus for tracking a speaker, which moves a camera by tracking a speaker.

SUMMARY OF THE INVENTION An object of the present invention is to provide a method and apparatus for tracking a speaker, which enables the talker to make a call while moving by automatically moving the camera by tracking the talker.

The present invention relates to a method and an apparatus for tracking a speaker in a video telephone using an infrared camera, which separates the background and the speaker according to the size and direction of the boundary pixel extracted from the input image and moves the camera through the central tracking. .

In general, there are two kinds of tracking methods for tracking an object, that is, a conversation, such as correlation tracking and center tracking.

The correlation tracking technique defines an area A of a size appropriate for the moving object of the previous frame, that is, the position of the dialoger as shown in Fig. 1A, and calculates the correlation between the defined area A and the search area within the current frame. Therefore, it is a method of estimating that an object has moved to the region A 'having the highest correlation.

That is, the correlation tracking technique defines a window area having a constant size, that is, a correlation area, including a moving object when a position of a moving object is given in a given n-th image, and for each position on the search area in the n + 1 th image. The correlation is calculated to regard the position of the region having the highest correlation as the position of the moving object in the n + 1 th image.

Here, the shape of the initial window when the correlation is calculated, that is, the area used when calculating the correlation between the current frame and the previous frame is mainly composed of a square window.

Therefore, the correlation tracking technique calculates the correlation directly from the image of the current frame input without performing the image segmentation process, so that the tracking performance is maintained even for a relatively complex image, but the computational amount is large.

That is, the correlation tracking technique generally uses contrast information of the image rather than the center point tracking technique, and thus, some background tracking performance can be expected even when image scattering is impossible due to background scattering. However, the correlation tracking technique has a disadvantage in that a large amount of calculation is required because the correlation must be calculated to estimate the movement of the moving object.

In addition, the center point tracking method is a method of tracking the center point B of the extracted moving object after separating the moving object from the background, as shown in FIG.

Therefore, when the image is relatively simple and the image segmentation is easy, the tracking stability is good and the tracking performance is good because the constraint on the speed of the traceable object is relatively small.

However, the center tracking technique uses a threshold to separate a moving object, that is, a dialog, from the background, and thus it is difficult to extract an accurate dialog from an image. That is, since the center tracking method binarizes the background and the object by using the threshold, it is difficult to extract the exact object from the image, and has a disadvantage of being affected by noise when there is noise in the input image.

An object of the present invention is to provide a speaker tracking method and apparatus for improving tracking performance by using an infrared image and a central tracking method for easily distinguishing a background from a face in a video telephone.

1A is a diagram for explaining a conventional correlation tracking technique.

1B is a diagram for explaining a conventional center point tracking technique.

2 is a flowchart of a method for tracking a speaker according to the present invention.

FIG. 3 is a diagram for describing a boundary pixel extraction step of FIG. 2. FIG.

4 is a view for explaining a dialog extraction step of FIG.

FIG. 5A is a diagram for explaining a boundary bit allocation step of FIG. 4; FIG.

FIG. 5B is a diagram for explaining a dialogue center point determination step of FIG. 4; FIG.

6 is a block diagram of an apparatus for tracking a speaker according to the present invention;

* Explanation of symbols for the main parts of the drawings

300: camera, 400: motor unit, 410: motor, 420: motor driver, 500: controller, 600: image storage unit, 610: A / D converter, 620: memory, 700: boundary direction bit allocation unit, 710: boundary A pixel extractor, 720: boundary line extractor, 730: shifter, 740: boundary bit set unit, 800: dialog generator, 810: counter, 820: result image generator, 900: center point calculator

In order to achieve the above object, the method for tracking a speaker of a video telephone according to the present invention includes: a boundary extraction step of extracting boundary pixels and boundary lines from an infrared image signal input in a tracking mode; Allocating a specific value to the boundary direction bits corresponding to the extracted boundary line direction for all pixels located in the 90 degree direction of the boundary line direction of the extracted boundary pixel, and extracting and storing the dialogue image according to the number of assigned specific values Dialogue extracting step; A dialog image center point calculation step of calculating a center point of the extracted dialog image; And a motion estimation and a motor driving step of moving the camera by estimating the movement according to the center point of the calculated dialogue image.

In addition, the speaker tracking apparatus of the video telephone according to the present invention for achieving the above object comprises image storage means for A / D conversion and storing the infrared image input in the tracking mode; Extracts a boundary pixel and a boundary line direction from an infrared image stored in the image storage means, and generates a dialogue image according to the number of specific values assigned to the boundary direction bits of each pixel output from the corresponding boundary direction bit allocation means of each pixel. Talker generating means for storing in the video storing means; Center point calculation means for retrieving the talker image stored in the image storage means and calculating a center point of the talker; And control means for controlling the operation of the projection storage means and controlling the movement of the camera by estimating the movement of the dialoger according to the center point calculated by the center point calculation means.

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

The conversation tracking method of the video telephone according to the present invention includes the boundary extraction step 100, 101, 102, 103, the dialog extraction step 104, 105, 106, 107 and the center point calculation step 108 as shown in FIG. And motion estimation and motor drive step 109.

The boundary extracting step (100, 101, 102, 103) is a step of extracting the boundary pixel and the boundary line direction from the infrared image signal input in the tracking mode, A / D (Analog / Digital) Infrared image storage step (100, 101) of converting and storing, boundary pixel extraction step 102 of extracting a boundary pixel from the stored infrared image signal, and arc tangent values (atan) of the vertical and horizontal derivatives of the extracted boundary pixel (h, v)) is performed by the boundary direction extraction step 103 of extracting the boundary direction.

Here, the boundary pixel is determined by comparing the magnitudes of the differential values in the horizontal and vertical directions of the brightness values of the respective pixels in the set initial window with the set boundary determination threshold.

The dialogue extracting step (104, 105, 106, 107) assigns and assigns a specific value to the boundary direction bit corresponding to the extracted boundary line direction for all pixels located in the 90 degree direction of the boundary line direction of the extracted boundary pixel. Extracting and storing the dialogue image according to the number of specified values, and assigning a specific value to each pixel positioned in a 90 degree direction with respect to the boundary line direction of the extracted boundary pixel as a boundary direction bit according to the extracted boundary line direction. Dialogue bit allocation step 104, Dialogue center point determination step 105 of determining the dialog center line by counting the number of specific values assigned to the boundary direction bits, and Dialog image formed by a border pixel around the dialog center point And extracting and generating the dialogue image.

In the boundary direction bit allocation step, as illustrated in FIG. 4, the search pixel number setting step of setting the number of search pixels for allocating the boundary direction bits (200), and the extracted boundary according to the set number of pixels. A pixel selection step 201 of selecting a pixel located at 90 degrees with respect to the boundary line direction of the pixel, an allocation step 202 of allocating 8 bits as the boundary direction bits to the selected pixel, and the boundary bits of the allocated 8 bits Is performed by a boundary direction bit set step 203 of setting bits corresponding to the extracted boundary line direction to a specific value.

Here, the boundary direction bit set step 203 sets a specific value to the first bit of the allocated 8-bit boundary direction bits when the extracted boundary line direction is (-1, -1) in the x direction and the y direction. When the extracted boundary line direction is (0, -1) in the x direction and the y direction, a specific value is set to the second bit among the allocated 8 bit boundary direction bits, and the extracted boundary line direction is the x direction and the y direction. If (1, -1), the specific value is set to the third bit among the allocated 8-bit boundary direction bits, and if the extracted boundary line direction is (1, 0) in the x direction and the y direction (1, 0), the allocated 8 bits A specific value is set in the fourth bit among the boundary direction bits of, and if the extracted boundary line direction is (1, 1) in the x direction and the y direction, the specific value is set in the fifth bit of the allocated 8 bit boundary direction bits. And the extracted boundary line direction is In the case of (0, 1) in the x direction and the y direction, a specific value is set to the sixth bit among the allocated 8 bit boundary direction bits, and the extracted boundary line direction is (-1, 1) in the x direction and the y direction. If a specific value is set to the seventh bit among the allocated 8-bit bits of the direction direction, and if the extracted boundary line direction is (-1, 0) in the x- and y-directions, This is done by setting a specific value to the eighth bit.

In addition, the number of search pixels is set to 1/2 or more of the number of pixels corresponding to the expected length of the talker's face.

On the other hand, the center point calculation step 108 is to calculate the center point of the extracted dialog image.

The motion estimation and motor driving step 109 is a step of moving the camera by estimating the motion according to the calculated center point of the dialog image.

The speaker tracking method of the video telephone according to the present invention thus made will be described in detail with reference to the accompanying drawings.

First, the boundary extraction step (100, 101, 102, 103) is performed to track the boundary of the input infrared image, which will be described in detail.

Since the call is initiated and there may be times when the user does not want to use the tracking function, the user first determines whether to use the tracking function (100). The distinction between tracking mode and non-tracking mode is simple with a switch. That is, if the switch is ON, the tracking function is performed. If the switch is OFF, the tracking function is not performed.

When the tracking mode is turned on to perform the tracking function, the infrared image storage step 101 of converting the infrared image input from the camera to A / D (Analog / digital) conversion is performed.

That is, the infrared image signal coming from the camera is stored as a digital image (I (x, y)) having a value within a specified range on a two-dimensional matrix through an A / D converter. The digital image is stored in a memory and used as an input image.

In the infrared image signal stored as described above, boundary pixels are extracted through differentiation. The boundary pixel extraction step 102 will be described with reference to FIG. 3 as follows.

The stored infrared image signal has a brightness value assigned to the two-dimensional matrix I (x, y).

The video signal thus made is differentiated in the horizontal and vertical directions as shown in Equations (1) and (2) below.

I (x-1, y-1) + I (x-1, y) + I (x-1, y + 1) -I (x + 1, y-1) -I (x + 1, y) -I (x + 1, y-1) + h

---- Formula (1)

I (x-1, y-1) + I (x, y-1) + I (x + 1, y-1) -I (x-1, y + 1) -I (x, y + 1) -I (x + 1, y + 1) = v ----- Equation (2)

That is, as shown in Fig. 3, the horizontal derivative is the brightness values A4, A5, A6 of the left pixel and the brightness values A1, A2, A3 of the right pixel with respect to the reference pixel (x, y) as the center. The vertical derivative is obtained by comparing the brightness values (A4, A7, A1) of the upper pixels and the brightness values (A6, A8, A3) of the lower pixels with respect to the reference pixel (x, y) as the center. This is done by saving a car.

The horizontal differential value (h) is calculated by performing horizontal differentiation as shown in Equation (1), and the vertical differential value (v) is calculated by performing vertical differentiation as shown in Equation (2) above. Calculate the size of h, v) ((h ² + v ² ) ^1/2 ). That is, the vertical and horizontal derivatives (h, v) are squared, respectively, and the square root is again used to calculate the magnitude ((h ² + v ² ) ^1/2 ) of the derivatives.

The calculated magnitude (h ² + v ² ) ^1/2 is compared with a preset boundary determination threshold and the reference pixel (x, y) becomes a boundary pixel when it is larger than the boundary determination threshold.

This differentiation process is performed for all image signals of one frame.

After the boundary pixels are extracted as described above, the boundary line direction extraction step 103 of extracting the boundary line direction using the arc tantent values atan (h, v) of the vertical and horizontal differential values of the extracted boundary pixels is performed.

That is, the direction of the boundary line is extracted by calculating the arc tangent (atan (h, v)) values of the vertical and horizontal differential values h and v of the extracted boundary pixels. At this time, the calculated arc tangent value (atan (h, v)) may have a value from -90 degrees to 90 degrees.

Dialogue extraction step 104, 105, 106 for extracting and storing a dialogue image by assigning a specific value to a corresponding boundary direction bit according to the extracted boundary pixel and boundary line direction using the extracted boundary pixel and boundary line direction. In order to perform 107, the boundary direction bit allocation step 104 and the dialogue center point determination step 105 are performed, which will be described with reference to FIGS. 4, 5A, and 5B.

First, a search pixel number setting step 200 of setting the number of search pixels for allocating the boundary direction bits is performed.

In this case, the number of the search pixels is set to 1/2 or more of the number of pixels corresponding to the expected length of the face of the talker. This is because the center point of the face of the talker must be extracted to extract the talker's movement.

After setting the number of search pixels, a pixel to which the boundary direction bits are to be set is selected (201). That is, the first adjacent pixel is selected among pixels located 90 degrees with respect to the boundary line direction of the extracted boundary pixel, that is, pixels 90 degrees to the left with respect to the boundary line direction (201). 8 bits are allocated to the selected pixel as the boundary direction bits (202).

That is, a register of '00000000' is allocated to each pixel of the input image, and a memory having a size of 1 byte for every pixel of the input image I (x, y) is used as a register.

On the other hand, the boundary direction has a value ranging from -90 degrees to 90 degrees, so it is divided into eight at 45 degree intervals. That is, as shown in Fig. 5A, the boundary line direction is (-1, -1), (0, -1), (1, -1), (1, 0), ( It divides into 1, 1), (0, 1), (-1, 1), and (-1, 0) so that the boundary direction can be divided into 8 types.

For this purpose, the horizontal and vertical differential values (h, v) of each boundary pixel and the corresponding (x-direction, y-direction) for calculating the boundary line direction are stored in a lookup table, so that the vertical and horizontal differential values (h, v) of the boundary pixels are stored. ), The boundary direction can be extracted easily.

After the boundary line direction is calculated in the (x direction, y direction) as described above, a specific value is set in the boundary direction bit allocated to each pixel (203), which will be described with reference to FIG. 5A.

That is, in the case of the pixel X1, since the boundary line direction is (-1, 1), '000000' 'is set as the boundary direction bit by assigning' 1 'to the seventh bit. In the case of the pixel X2, since the boundary line direction is (0, 1), '1' is assigned to the sixth bit and '00000100' is set as the boundary direction bit (203). In the pixel X3, since the boundary line direction is (1, 1), '1' is assigned to the fifth bit and '00001000' is set as the boundary direction bit. Also, in the case of the pixel X4, since the boundary line direction is (1, 0), '000' is set as the boundary direction bit by assigning '1' to the fourth bit. Also, in the case of the pixel X5, since the boundary line direction is (1, -1), '00100000' is set as the boundary direction bit by assigning '1' to the third bit. In the case of the pixel X6, since the boundary line direction is (0, -1), '1' is assigned to the second bit and '01000000' is set as the boundary direction bit. In the pixel X7, since the boundary line direction is (-1, -1), '1' is assigned to the first bit and '100000000' is set as the boundary direction bit.

When the boundary direction bits are set in this way, in the case of the pixel X0, the boundary direction is eight, and thus, '1111111' is set as the boundary direction bits by allocating '1' to all eight bits. In particular, in the case of the pixel X0, all boundary directions, that is, (-1, 1), (0, 1), (1, 1), (1, 0), (1, -1), (0,- Since 1), (-1, -1) are all in the boundary line direction, '1111111' is set as the boundary direction bit.

After the boundary direction bits are set in the selected pixel as described above (203), the number of specific values, that is, '1', is counted among the set boundary direction bits (204). In this case, when the counting value is greater than '6', it is determined as the center point, which is the center part of the conversation face (205, 206).

For example, in each pixel X1, X2, X3, X4, X5, X6, X7, and X8 of FIG. 5A, since only one bit of the boundary direction bit is set to '1', it is not a center point of the face of the talker. However, in the pixel X0 of Fig. 5A, since all eight boundary bits are set to '1', it becomes the center point of the dialogue.

In addition, as shown in FIG. 5B, in the case of the pixel M, seven bits among eight direction bits may be set to '1', thereby forming a center point of the dialogue.

That is, as shown in FIG. 5B, since the boundary bit is set to '11111011' in the pixel M, the number of '1's is greater than' 6 'to' 7 'and the pixel M is determined as the center point of the dialogue. (206).

After retrieving the boundary direction bits of the selected pixel as described above, it is searched whether the number of selected pixels is equal to the number of search pixels (207).

That is, since the pixels located in the 90-degree direction with respect to the boundary line direction from the corresponding boundary pixels must be searched for the number of search pixels set, the boundary direction bits must be set, and thus the above-mentioned selection pixel number search step 207 must be performed.

For example, if the length of the face of the talker is 20 pixels, the number of search pixels is 10 or more. Therefore, if the number of search pixels is set to '10', the boundary direction bits should be set by searching up to ten pixels located in the 90 degree direction from the boundary pixel.

If the boundary bits of the ten pixels are not retrieved as described above, pixels adjacent to the pixel selected in the opposite direction to the direction in which the boundary pixels are located are selected (208), and the boundary bit allocation step 202 and the boundary bit set step ( Continue to step 204).

On the other hand, if all of the boundary direction bits of the ten pixels are searched, the neighboring boundary pixels are selected in the direction of the boundary line in step 209. In this case, it is searched whether the selected boundary pixel is the boundary pixel in which the boundary direction bit search described above is completed (210). That is, if the search for the selected word is already completed or not, the search proceeds to the pixel selection step 201.

In the case of the boundary pixel that is already selected and the search is completed, that is, when the search for all the boundary pixels forming one closed curve is completed, the plane formed by the boundary pixel surrounded by the center of the dialogue center determined above is determined as the dialogue image. (211). The speaker image determined as described above is extracted from the infrared image of one frame, generated, and stored, in operation 107.

Then, the center point calculation step 108 is performed to calculate the center point of the extracted and stored speaker image, and then the motion estimation and motor driving step 109 is performed to move the motion according to the calculated center point position of the speaker image. Move the camera by estimation.

As illustrated in FIG. 6, the apparatus for tracking a speaker of a video telephone according to the present invention includes a motor unit 400, a controller 500, an image storage unit 600, a boundary bit allocator 700, and a speaker generator 800. ), And a center point calculator 900.

The image storage unit 600 A / D converts and stores the infrared image input in the tracking mode, and A / D converts the infrared image signal input in the tracking mode under the control of the control unit 500. A memory 620 for storing an infrared image signal output from the A / D converter 610 and a speaker image signal output from the speaker generator 800 under the control of the converter 610 and the controller 500. It consists of.

Here, the memory 620 preferably uses a VRAM form to store an input image signal.

The boundary direction bit allocation unit 700 extracts a boundary pixel and a boundary line direction from an infrared image stored in the memory 620 of the image storage unit 600, and allocates a specific value to a corresponding boundary direction bit of each pixel. The boundary pixel extracting unit 710 extracts a boundary signal from an infrared image signal stored in the memory 620 of the image storage unit 600, and a boundary line according to the boundary pixel extracted by the boundary pixel extracting unit 710. Shifter for shifting the pixels located in the 90-degree direction with respect to the boundary line direction of the boundary line extracted from the boundary line direction extraction unit 720, the boundary pixel extraction unit 720 and the boundary direction extraction unit 720 to extract the direction 730 and a boundary direction bit set unit 740 that sets a value corresponding to the extracted boundary line direction to each of the pixels output from the shifter 730 as boundary direction bits. It is composed.

The dialog generator 800 generates the speaker image according to the number of specific values allocated to the boundary direction bits of each pixel output from the boundary direction bit set unit 740 of the boundary direction bit allocation unit 700 and stores the image. The number of specific values allocated to the boundary direction bits of each pixel output from the boundary direction bit set unit 740 of the boundary direction bit allocation unit 700 by storing in the memory 620 of the unit 600. When the counter 810 and the counter 810 have a count value greater than or equal to a predetermined value, the corresponding pixels are determined as the center point of the talker to generate a talker image, and store the same in the memory 620 of the image storage unit 600. The result is an image generator 820.

The center point calculator 900 searches for the talker image stored in the memory 620 of the image storage unit 600, calculates a center point of the talker, and outputs the center point of the talker to the controller 500.

The controller 500 controls the operations of the A / D converter 610 and the memory 620 of the image storage unit 600 and estimates the movement of the dialoger according to the center point calculated by the center point calculator 900. The motor unit 400 is controlled to move.

The motor unit 400 includes a motor 410 for moving the camera 300 and a motor driver 420 for driving the motor 410 under the control of the controller 500.

The operation of the speaker tracking apparatus of the video telephone according to the present invention configured as described above will be described.

First, the A / D converter 610 is turned on under the control of the control unit 500 to A / D convert the NTSC signal from the camera 300 and output it as a digital signal.

That is, when the tracking mode is turned on to start a phone call, the A / D converter 610 converts the video signal received from the camera 300 and the A / D converted digital video signal into the memory 620. Will be stored in.

The image signal stored in the memory 620 is input to the boundary pixel extracting unit 710, and the magnitudes of the horizontal and vertical differential values h and v of each pixel are calculated by the above equations (1) and (2). The light weight pixel is extracted by comparing with the set threshold pixel discrimination threshold.

As described above, the boundary pixel extracted by the boundary pixel extracting unit 710 is input, and the boundary line direction extracting unit 720 extracts the direction of the boundary line.

That is, the boundary line direction extractor 720 extracts arc tangent values atan (h, v) of vertical and horizontal derivatives of the boundary pixels extracted by the boundary pixel extractor 710 in the boundary line direction.

In this case, the extracted boundary line direction may have a value ranging from 90 degrees to -90 degrees as described above. In other words, the arc tangent values are (-1, -1), (0, -1), (1, -1), (1, 0), (1, 1), It is divided into (0, 1), (-1, -1), and (-1, 0) so that the boundary direction can be divided into eight types.

The arc tangent, which is a real mathematical function, is stored in the form of a lookup table in ROM in the speaker tracking device and uses the results for the corresponding horizontal and vertical differential values (h, v).

These eight boundary line directions are represented by one bit in the eight-bit register in the boundary direction bit set unit 740. That is, as described above, the 8-bit boundary direction bits are set to '10000000', '01000000', '00100000', '00010000', '00001000', '00000100', '00000010', and '00000001' respectively according to the boundary line direction. Express as

Thus, the operation of setting the boundary direction bits will be described.

First, as illustrated in FIG. 5A, the pixel output from the boundary pixel extracting unit 9710 is shifted to select each pixel located in a 90 degree direction in the boundary line direction of the boundary pixel extracted from the boundary pixel extracting unit 710. It outputs to the boundary direction bit set unit 740 while shifting at 730.

As described above, the boundary direction bit set unit 740 sets the bit corresponding to the boundary line direction output from the boundary line direction extraction unit 720 among the eight boundary direction bits of the input pixel to '1' and outputs the bit.

The 8-bit boundary direction bits R (x, y) output from the boundary direction vis set unit 740 are input to the counter 810, and the number of '1' is counted.

As a result, when the number of '1' counted by the counter 810 is greater than '6', in the image generating unit 820, when seven or more lines that form 90 degrees with respect to the boundary line are collected, the corresponding pixel is the center point of the face of the speaker. The resultant image of the boundary pixels surrounded by the center point of the talker is extracted and the face image of the talker is generated.

As a result, the image of the speaker face generated by the image generator 820 is stored in the memory 620, and the center point is calculated by the center point calculator 900 as shown in Equations 3 and 4 below.

xcenter = 1 / N ∑ I (i, j) X i ---- Equation (3)

y center point = 1 / N ∑ I (i, j) X j ---- Equation (4)

Here, N is the number of pixels constituting the resultant image, that is, the image of the dialogue face.

The calculated center point is input to the controller 500, and the controller 500 uses the same to estimate the movement, that is, the movement of the talker, and to control the motor driver 420 according to the estimated movement. The motor 410 moves the infrared camera 300 by the motor driver 420 to track the talker.

Meanwhile, when an infrared image is input from the infrared camera 300, the controller 500 may turn on the A / D converter 610 and the memory 620 to A / D convert and store an image signal from the infrared camera 300. Make sure

After storing in the memory 620 as described above, the A / D converter 610 is turned off so that the video signal from the camera 300 cannot be received again.

After the resultant image is generated by the resultant image generator 820, the A / D converter 610 and the memory 620 are turned on to receive the next image signal from the camera 300.

This operation is to prevent the current input image from changing until the resultant image is extracted from the input image signal and the center point is calculated.

As described above, the method and apparatus for tracking a dialog of a video telephone according to the present invention utilizes the size and direction of a boundary pixel extracted from an infrared input image to assign a register value to each input image. Tracking performance is improved by separating the background and the dialog according to the value.

Claims (11)

A boundary extraction step (100, 101, 102, 103) for extracting the boundary line direction of the boundary pixel from the infrared image signal input in the tracking mode; Allocating a specific value to the boundary direction bits corresponding to the extracted boundary line direction for all pixels located in the 90 degree direction of the boundary line direction of the extracted boundary pixel, and extracting and storing the dialogue image according to the number of assigned specific values Dialog extraction step 104, 105, 106, 107; A center point calculation step (108) of calculating a center point of the extracted dialog image; And a motion estimation and motor driving step (109) of moving the camera by estimating the motion according to the calculated center point of the speaker image. The method of claim 1, wherein the boundary extraction step (100, 101, 102, 103) is an infrared image storage step (100, 101) for converting and storing the infrared image signal input in the tracking mode A / D (Analog / Digital) ; A boundary pixel extraction step (102) of extracting a boundary pixel from the stored infrared image signal; And a boundary line extraction step (103) of extracting a boundary line direction with arc tangent values (atan (h, v)) of vertical and horizontal differential values of the extracted boundary pixels. Way. The method of claim 2, wherein the boundary pixel is determined by comparing magnitudes of differential values in horizontal and vertical directions of brightness values of each pixel in the set initial window with a threshold for determining a boundary line. . The method of claim 1, wherein the dialogue extracting step (104, 105, 106, 107) bounds a specific value according to the extracted boundary line direction to each pixel located in a 90 degree direction with respect to the boundary line direction of the extracted boundary pixel A boundary direction bit allocation step of allocating 104 direction bits; A dialogue center point determining step (105) of determining a dialogue center point by counting the number of specific values assigned to the boundary direction bits; And a dialog image generation step (106) of extracting and generating a dialog image of a border pixel surrounded by the center point of the dialog center. 5. The method of claim 4, wherein the boundary bit allocation step comprises: a search pixel number setting step of setting a number of search pixels for allocating the boundary direction bits; A pixel selecting step (201) of selecting a pixel located at 90 degrees with respect to a boundary line direction of the extracted boundary pixel according to the set number of pixels; An allocating step (202) of allocating 8 bits as the boundary direction bits to the selected pixel; And a boundary direction bit set step (203) of setting the bits corresponding to the extracted boundary line direction among the allocated 8-bit boundary direction bits to a specific value. 6. The method of claim 5, wherein the boundary direction bit set step (203) is specified to the first bit of the allocated 8-bit boundary direction bits when the extracted boundary line direction is (-1, -1) in the x and y directions. Set a value, and when the extracted boundary line direction is (0, -1) in the x direction and the y direction, set a specific value to the second bit among the allocated 8 bit boundary direction bits, and the extracted boundary line direction is x In the case of (1, -1) in the direction and y direction When a specific value is set in the third bit among the allocated 8 bit boundary direction bits, and the extracted boundary line direction is (1, 0) in the x direction and the y direction A specific value is set in the fourth bit among the allocated 8 bits of the boundary direction, and when the extracted boundary line direction is (1, 1) in the x direction and the y direction, the fifth bit among the allocated 8 bit boundary direction bits is set. Set a specific value, and extract the boundary If the direction is (0, 1) in the x direction and the y direction, a specific value is set in the sixth bit among the allocated 8 bit boundary direction bits, and the extracted boundary direction is (-1, In the case of 1), a specific value is set in the seventh bit among the allocated 8 bit boundary direction bits, and when the extracted boundary line direction is (-1, 0) in the x direction and the y direction, the allocated 8 bit boundary direction The method of claim 1, wherein the method is performed by setting a specific value to the eighth bit among the bits. 6. The method of claim 5, wherein the number of search pixels is set to 1/2 or more of the number of pixels corresponding to the expected length of the talker's face. Image storage means 600 for A / D converting and storing the infrared image input in the tracking mode; Boundary direction bit allocation means (700) for extracting a boundary pixel and a boundary line direction from an infrared image stored in the image storing means (600) to assign and set a specific value to a corresponding boundary direction bit of each pixel; Talker generation means (800) for generating a talker image according to the number of specific values assigned to the boundary direction bits of each pixel output from the boundary direction bit allocation means (700) and storing it in the image storage means (600); Center point calculation means (900) for retrieving the center point of the dialog by searching the dialog image stored in the image storage means (600); And a control means 500 for controlling the operation of the image storing means 600 and controlling the movement of the camera by estimating the movement of the talker according to the center point calculated by the center point calculating means 900. Talker tracking device for video telephone. The apparatus of claim 8, wherein the image storing means (600) comprises: an A / D converter (610) for A / D converting an infrared image signal input in a tracking mode under the control of the control means (500); And a memory 620 for storing the infrared image signal output from the A / D converter 610 and the speaker image signal output from the speaker generating means 800 under the control of the control means 500. Talker tracking device of the video telephone, characterized in that. The method of claim 8, wherein the boundary direction bit allocation means (700) comprises: a boundary pixel extracting unit (710) for extracting boundary pixels from an infrared image signal stored in the image storing means (600); A boundary line direction extracting unit 720 for extracting a boundary line direction according to the boundary pixels extracted by the boundary pixel extracting unit 710; A shifter 730 for shifting and outputting pixels positioned in a 90 degree direction with respect to the boundary line direction of the boundary pixel extracted by the boundary pixel extracting unit 710 and the boundary line direction extracting unit 720; And a boundary direction bit set unit 740 for setting a corresponding value according to the extracted boundary line direction to each of the pixels output from the shifter 730 as boundary direction bits. Device. 9. The apparatus of claim 8, wherein the dialog generating means (800) comprises: a counter (810) for counting the number of specific values assigned to the boundary direction bits of each pixel output from the boundary direction bit allocation means (700); And a result image generator 820 for determining a corresponding pixel as a center point of the talker when the counting value of the counter 810 is equal to or greater than a predetermined value to generate a talker image and storing the talker image in the image storage means 600. Talker tracking device of Mars telephone, characterized in that configured.