KR20000060570A - Face and eye position detecting apparatus and method using T-type search region - Google Patents

Abstract

PURPOSE: An apparatus and method for detecting a position of a face and eyes are provided to reduce a face position detecting area and an estimation rule, thereby improving a detecting speed of a face position, and reducing an operation process. CONSTITUTION: The apparatus for detecting a position of a face and eyes using a T type searching area comprises a standard value adjusting portion(100) for adjusting a standard value as a standard for determining whether an edge signal exists according to a luminosity of an entire input image, an edge signal detecting portion(110) for detecting the edge signal with respect to a digital image including a face based on the standard value, and a face position detecting portion(120) for detecting the face using the T type searching area. A method for detecting a position of a face and eyes using a T type searching area comprises steps of adjusting the standard value according to the luminosity of the entire input image, detecting the edge signal in the input image, forming the edge image, and detecting the face position in the edge image using the searching area.

Description

Face and eye position detecting apparatus and method using T-type search region

The present invention relates to face detection, and more particularly, to an apparatus and method for detecting face and eye position in a two-dimensional digital image including a face.

In a system for receiving a human face image as a 2D digital image and outputting various responses, a technique for detecting a face and an eye position is a basic technology. For example, the basic technology in the configuration of a facial recognition system or a drowsiness recognition system in a car recently applied to a security system is a technique for detecting the position of a face. In the case of the drowsiness recognition system, there are a number of patents for a method of not using a face image, but it is cumbersome, such as attaching a separate additional device in actual use.

Thanks to the development of digital technology, various image processing and information extraction become easy through digital image processing, so that face image is accepted as digital image and processed and processed using dedicated hardware or computer. Facial expression is detected, eye detection or eye blink is detected. Digital technology is in the spotlight because it is more flexible than conventional data extraction by analog circuits and can acquire more diverse information. However, as the resolution of the image increases, more powerful computing power is required, and the increase in cost is inevitably a problem that occurs in digital processing technology. Therefore, in the digital image processing, obtaining a result with less computation amount is a very important problem.

Conventional techniques for finding the position of the face require many computations such as Hough transformation of curve components in neural networks or faces, correlation calculation by a plurality of templates, and most algorithms are developed in computers that are becoming more powerful. Increasingly, more complex algorithms are being developed. However, in practical applications, systems using face detection are not used only in powerful computers, and therefore, development of small and efficient detection algorithms using less computing power is required.

Face position detection and eye position detection are very important in many applications dealing with facial images. In particular, once the position of the eye is determined, the position of the other components of the face can be easily determined by an appropriate proportional relationship.

The technical problem to be solved by the present invention is to reduce the face position search area and the decision rule by using the T-type search area, thereby improving the face position detection speed, reducing the amount of computation, and thus operating in a low-cost processor, An apparatus and method for detecting a face position using a T-type search region are provided.

Another technical problem to be solved by the present invention is to reduce the face and eye position search area and the determination rule by using the T-type search area, thereby improving the face and eye position detection speed, reducing the calculation amount, and accordingly the low cost. The present invention provides a face and eye position detection apparatus using the T-type search region and a method thereof, which are also operable in a processor.

1 is a block diagram of a face and eye position detection apparatus according to the present invention.

2 is a flowchart for explaining a face and eye position detection method according to the present invention.

FIG. 3 is a flowchart according to a preferred embodiment of steps 200 and 210 of FIG. 2.

4A to 4B are flowcharts according to a preferred embodiment of step 220 of FIG. 2.

5 (a) to 5 (f) illustrate experimentally a process of detecting a face position using a T-type search area according to the present invention.

6 is a flowchart according to a preferred embodiment of step 230 of FIG.

In order to achieve the above object, the facial position detection apparatus according to the present invention in a two-dimensional digital image including a face, the reference value adjusting unit for adjusting the reference value as a reference for determining the presence or absence of the edge signal according to the brightness of the entire input image The edge signal detector detects an edge signal of a digital image based on a reference value, moves the position of the T-type search region in the edge image, and generates an edge image. And a face position detection unit for determining whether to satisfy the geometric conditions of and detecting a region where the total amount of the edge signal is maximum within the T-type search region satisfying the conditions as the face position.

In addition, in order to achieve the above object, the method for detecting a face position according to the present invention in a two-dimensional digital image including a face includes (a) a reference for determining the presence or absence of an edge signal according to the brightness of the entire image with respect to the input image. Adjusting a reference value, (b) detecting an edge signal in the input image based on the reference value, generating an edge image, and (c) detecting a face position in the edge image using a T-type search region; do.

In order to achieve the above object, the face and eye position detection apparatus according to the present invention in a two-dimensional digital image including a face adjusts a reference value, which is a reference for determining the presence or absence of an edge signal, according to the brightness of the entire input image. A reference value adjusting unit, an edge signal detection unit that detects an edge signal for a digital image based on the reference value, generates an edge image, and moves the position of the T-type search area in the edge image, and the edge signal in the T-type search area has a face shape. A face position detector which determines whether the predetermined geometric conditions are satisfied and detects a region where the total amount of the edge signal is maximum in the T-shaped search region satisfying the conditions as a face position and a T-shaped search region at the detected face position. On the eye to detect the position of the eye using the polarity of the edge signal in the horizontal direction in the T-shaped search region. And a detection unit.

In addition, in order to achieve the above another object, the face and eye position detection method according to the present invention in a two-dimensional digital image including a face includes (a) determining the presence or absence of an edge signal according to the brightness of the entire image with respect to the input image. Adjusting a reference value as a reference, (b) detecting an edge signal from an input image based on the reference value, generating an edge image, and (c) detecting a face position from an edge image using a T-type search region And (d) positioning the T-type search region at the detected face position, and detecting the position of the eye using the polarity of the edge signal in the horizontal direction in the T-type search region.

Hereinafter, a face and eye position detection apparatus and a method thereof according to the present invention will be described with reference to the accompanying drawings.

1 is a block diagram of a face and eye position detection apparatus according to the present invention. The face position detecting apparatus includes a reference value adjusting unit 100, an edge signal detecting unit 110, and a face position detecting unit 120, and the face and eye position detecting apparatus further includes an eye position detecting unit 130.

Apparatus and method thereof in accordance with the present invention are generally a video camcorder or a closed circuit T.V. When a face image photographed by an imaging device such as a digital image is received, a face and an eye position are detected in a two-dimensional digital image using a conventional microprocessor, a digital signal processor, or a computer, and the numerical data is represented as numerical data. . In particular, the apparatus and method according to the present invention detect an edge signal from a two-dimensional digital image or pixels sampled at a low resolution for the image, and convert the edge signal generated in the face from the edge image into a T shape. The facial position is detected by applying a predetermined geometrical relationship in the search area consisting of the?, And then the eye position is detected using the polarity of the edge signal in the horizontal direction through appropriate processing.

Referring to FIG. 1, the reference value adjusting unit 100 inputs a digital image including a face obtained through a video camcorder or various camera inputs through an input terminal IN. The reference value adjusting unit 100 appropriately adjusts a reference value for obtaining an edge image according to the brightness of the entire image with respect to the input digital image. Here, the reference value is a threshold value as a reference for determining the presence or absence of the edge signal.

The edge signal detector 110 detects an edge signal for the digital image based on the reference value adjusted by the reference value adjuster 100, and generates an edge image as a result. The difference between the upper and lower image pixel values is obtained from the digital image, and if the difference is greater than or equal to the reference value, it is determined as the horizontal edge. In addition, the difference between the left and right video pixel values is determined, and if the difference is greater than or equal to the reference value, it is determined as the vertical edge.

The face position detector 120 detects a face position in an edge image by using a T-type search region. The T-shaped search region is a T-shaped search region that is matched to an area including two eyes, between two eyes, a nose, authentication, and lips when accurately positioned at a face position to be detected. Specifically, it is determined whether the edge signal within the T-shaped search region satisfies predetermined geometric conditions of the face shape by moving the position of the T-shaped search region in the edge image. An area where the total amount of the edge signal is maximum in the T-type search area that satisfies these conditions is detected as the face position. Here, the predetermined geometric conditions are conditions that are a reference for determining whether the amount of edge signals characteristically represented by, for example, the face shape is appropriate.

The eye position detector 130 detects the eye position based on the face position detected by the face position detector 120. The T-type search region is positioned at the detected face position, and the eye position is detected using the polarity of the edge signal in the horizontal direction in the T-type search region. In the T-type search region, an area in which edge signals having a positive polarity are concentrated in the form of a horizontal line can be detected. In this case, the region means the upper eyelid line region. The upper eyelid line area can be detected simply by eye position.

2 is a flowchart for explaining a face and eye position detection method according to the present invention. The operation of the apparatus shown in FIG. 1 will be described in detail with reference to FIG. 2.

First, a reference value, which is a reference for determining the presence or absence of an edge signal, is adjusted based on the brightness of the entire image with respect to the input image (step 200). Next, an edge signal is detected from the input image based on the reference value, and as a result, an edge image is generated (step 210).

FIG. 3 is a flowchart according to a preferred embodiment of steps 200 and 210 of FIG. 2.

In detail, when a two-dimensional digital image including a face is input (operation 300), a reference value is adjusted before generating an edge image (operation 310). In the method of adjusting the reference value, first, a change range of all pixel values is set for the input image (step 312). If the same reference value is applied to all input images without adjusting a reference value set for an image having appropriate brightness, the edge signal becomes small when the input image is totally dark or totally bright. In consideration of the fact that the edge signal is strengthened under the appropriate illumination, a change range of all pixel values is first obtained to obtain an edge image that is resistant to changes in illumination. After operation 312, the relative position on the change range of all pixel values of the input image is determined (operation 314). If the relative position is close to the minimum or maximum value of the change range of all the pixel values, the reference value is decreased, and if the relative position is close to the intermediate value, the reference value is increased (step 316).

Next, based on the reference value adjusted in step 310, for the input image, if the difference between the upper and lower image pixel values is greater than or equal to the reference value, it is determined as the horizontal edge, and if the difference between the left and right image pixel values is greater than or equal to the reference value, it is determined as the vertical edge. (Step 320). Accordingly, it is possible to minimize a face position determination error that may occur due to the change of the edge image according to the change in illumination of the image.

Referring back to FIG. 2, after operation 210, a face position is detected from an edge image using a T-type search region (operation 220).

4 is a flowchart according to a preferred embodiment of step 220 of FIG.

In detail, first, a face search position is determined (operation 400). The face is searched by moving the T-shaped search area with respect to the entire original edge image or the edge image sampled at low resolution. At this time, the initial position of the T-type search area uses the information when there is information on a previous image where the face position was previously detected. That is, the search range for the face search position is reduced by moving the position of the T-shaped search area around the previously detected face position. On the other hand, when the first face position is detected, the face search position is determined by moving the T-type search area over all positions of the edge image sampled at the edge image or the low resolution.

If the face search position determined in operation 400 is not the last search position of the edge image, the process proceeds directly to the next step.

If it is not the last search position, it is determined whether the edge signal in the T-shaped search region satisfies predetermined geometric conditions. The following determination steps are merely preferred embodiments, and the determination order and rules may be modified. If the condition is not satisfied at every determination step, the process proceeds to operation 400 again. First, it is determined whether the amount of edge signals is less than or equal to the reference value 1 by applying a condition that the amount of edge signals is scarce to the nostril region of the face in the T-type search region (step 404). If so, then it is determined whether the edge signal amount is less than or equal to the reference value 2 by applying the condition that the amount of edge signals is thin in the area between both the nostril region and the upper lip region (step 406).

If so, then it is determined whether the amount of edge signals generated in both eye areas is greater than or equal to the reference value 3 and the reference value 4, respectively (step 408). If so, then it is determined whether the difference between the amount of edge signals generated in both eye regions is equal to or less than the reference value 5 (step 410). It is determined whether the amount of edge signals generated in both nostril regions is greater than or equal to the reference value 6 and the reference value 7, respectively (step 412). If so, then it is determined whether the difference in the amount of edge signals respectively generated in both nostril regions is equal to or less than the reference value 8 (step 414). If so, then it is determined whether the amount of edge signals greater than or equal to the reference values 9, 10 and 11 are simultaneously generated in both the eye region, both the nostril region and the lip region (step 416).

A face search position that satisfies at least one or more conditions may be set as a face candidate position. In addition, in order to detect the face position more accurately, the following conditions can be further applied. Although not included in the T-type search region, it is determined whether the amount of edge signal at this time is equal to or less than the reference value 12 by applying the condition that the amount of edge signals is thin in both ball regions (step 418). If so, it is determined whether the average value of the brightness of each ball area is equal to or less than the reference value 13 and the reference value 14, respectively (step 420). If so, it is determined whether the difference between the average brightness of the upper and lower regions of the right cheek is less than or equal to the reference value 15 and the difference between the average brightness of the upper and lower regions of the left cheek is less than or equal to the reference value of 16 (step 422).

In operation 424, a face search position that satisfies the above conditions may be set as a face candidate position. After operation 424, the process proceeds to operation 400 again to determine another search position and repeats the above-described operation. If it is determined in step 402 that the search is finished, an area in which the total amount of the edge signal is maximized is searched among the set face candidate positions (step 426). The detected region is determined as the face position (operation 428). In FIG. 4, reference values 1 to 16 may be determined through experiments.

5 (a) to 5 (f) illustrate experimentally a process of detecting a face position using a T-type search area according to the present invention.

5 (a) shows an image including a face photographed by a CCD camera installed in an automobile dashboard. Such an image is converted into an edge image as shown in FIG. 5B through the edge signal detector 110 and step 210 in the present invention. As shown in (c) and (d) of FIG. 5, the face position is detected by moving the T-type search area in the edge image. FIG. 5C illustrates a case in which the T-type search region exactly matches the face position to be searched. In this situation, the position where the maximum edge signal satisfying the conditions shown in FIG. 4 is detected in the T-type search region. Meanwhile, FIG. 5 (d) shows a case where the T-type search region is located at a position other than a face, which is excluded from the face candidate position because the conditions shown in FIG. 4 are not satisfied.

Even when the T-type search area is normally matched to the face position, a large face may be seen due to the distance between the face and the camera (FIG. 5 (e)) and a small face may be seen (FIG. 5 (f)). have. In the example automotive environment, the edge signals of the major components of the face are included in the T-shaped search area for most driver positions. Since the proportion of the face in the image changes depending on the distance, it may not be easy to apply when the distance is extreme. However, since many systems operate within an appropriate distance range rather than a system corresponding to extreme distance change, designing a T-type search area considering this in actual application aspect may affect the size of the face as shown in FIGS. 5 (e) and 5 (f). It can be applied regardless.

Referring back to FIG. 2, the eye position is detected from the face position detected in operation 220 (operation 230). The T-type search area is positioned at the detected face position, and the eye position is detected using the polarity of the edge signal in the horizontal direction in the T-type search area.

6 is a flowchart according to a preferred embodiment of step 230 of FIG.

Specifically, first, the edge signal for the eye area is detected based on the position determined as the face position (operation 600). The eye position may be determined using the eye position, and an appropriate region of the eye portion in the T-type search region is determined as the eye search position (operation 602).

If the eye search position determined in step 602 is not the last search position, the process proceeds to the next step. If the eye search position is the last search position, the process proceeds to step 614 to be described later.

If it is not the last search position, for the eye search position, edge signals whose difference between the values of the pixels located downward from the values of the pixels located above the vertical line are detected to be equal to or greater than a positive reference value. Here, the reference value is a reference value adjusted in step 200. In operation 606, it is determined whether a region in which the detected edge signals having a positive polarity are concentrated in the form of a horizontal line is detected. Since the detected area corresponds to the upper eyelid line area in most cases, this area can be simply determined as the eye position.

In addition, in order to detect the eye position more accurately, the following conditions can be further applied. The following determination steps are merely preferred embodiments, and the determination order and rules may be modified. First, it is determined whether the edge signal at this time is equal to or less than the reference value 1 by applying a condition that the horizontal and vertical edge signals are scarce in the black and white areas in the area under the detected eyelid line with respect to the eye search position (step 608). ). If so, in the area under the detected eyelid line, it is determined whether the vertical edge signals on both sides of the pupil are greater than or equal to the reference value 2 (step 610). An area satisfying the above conditions is set as an eye candidate position (step 612).

After operation 612, the flow returns to operation 602 to repeat the above-described operation over the entire eye region of the eye search position. If it is determined in step 602 that the search has been completed, an area in which the total amount of the edge signal is maximized is searched among the set eye candidate positions (step 614). The searched area is determined as an eye position (step 616). In FIG. 6, the reference value 1 and the reference value 2 may be determined through experiments.

As described above, the apparatus for detecting a face position using the T-type search region and the method thereof according to the present invention can configure a system that is resistant to a change in illumination state by adjusting a reference value and using an edge signal, and the T-type search region. By reducing the face location search area and the decision rule, the speed of face detection can be improved, and the amount of computation can be reduced.

Claims (10)

An apparatus for detecting a face position in a two-dimensional digital image including a face, A reference value adjusting unit which adjusts a reference value which is a reference for determining the presence or absence of an edge signal according to the brightness of the entire input image; An edge signal detector for detecting an edge signal of the digital image based on the reference value and generating an edge image; And Moving the position of the T-type search region in the edge image and determining whether the edge signal in the T-type search region satisfies predetermined geometric conditions of the face shape, and determines an edge signal in the T-type search region that satisfies the conditions. And a face position detection unit for detecting an area where the total amount of the maximum is as a face position. An apparatus for detecting face and eye position in a two-dimensional digital image including a face, A reference value adjusting unit which adjusts a reference value which is a reference for determining the presence or absence of an edge signal according to the brightness of the entire input image; An edge signal detector for detecting an edge signal of the digital image based on the reference value and generating an edge image; Moving the position of the T-type search region in the edge image and determining whether the edge signal in the T-type search region satisfies predetermined geometric conditions of the face shape, and determines an edge signal in the T-type search region that satisfies the conditions. A face position detector for detecting a region where the total amount of the maximum is as a face position; And And an eye position detector for locating the T-shaped search area at the detected face position and detecting the position of the eye using polarities of horizontal edge signals in the T-shaped search area. Position detection device. A face position detection method in a two-dimensional digital image including a face, (a) adjusting a reference value, which is a reference for determining the presence or absence of an edge signal, according to the brightness of the entire image with respect to the input image; (b) detecting an edge signal from the input image based on the reference value and generating an edge image; And and (c) detecting a face position in the edge image by using a T-type search region. The method of claim 3, wherein step (c) comprises: (c1) determining a face search position by moving the position of the T-type search area in the edge image; (c2) For the face search position, the condition that the edge signal amount should be less than or equal to the reference value 1 in the nostril region of the face, and the condition that the edge signal amount should be less than or equal to the reference value 2 in both the nostril area and the upper lip area; The condition that the amount of edge signal generated in the eye area should be more than the reference value 3 and the reference value 4 respectively, The condition that the difference between the amount of edge signal generated in both eye areas should be less than the reference value 5, The amount of edge signal generated in each nostril area The reference value 6 and the reference value 7 respectively, the difference between the amount of the edge signal generated in each nostril region must be equal to or less than the reference value 8 and different reference values simultaneously in both eye areas, both nostril areas and the lip area 9 At least, using a condition that an edge signal amount greater than or equal to 10 and 11 must be generated. Setting a face search position that satisfies the conditions as a face candidate position; And (c3) detecting a face position from the face candidate positions obtained by repeating the steps (c1) and (c2), in which the total amount of the edge signal is maximum in the T-type search region. Face position detection method. The method of claim 4, wherein step (c2) comprises: Condition that the amount of edge signal is sparse in both ball areas, that the average brightness of each ball area should be 12 or less, and that the difference between the average brightness of the upper and lower areas of the right ball should be 13 or less, and Further applying the condition that the difference between the average values of the brightnesses of the upper left and lower areas of the left cheek is equal to or less than the reference value 14 to set the face search position that satisfies the above conditions as the face candidate position. . The method of claim 4 or 5, wherein the step (c1), And when there is information on a previous image in which a face position has been previously detected, reducing the search range for the face search position by moving the position of the T-type search region using the information. The method of claim 3, wherein step (b) comprises: And sampling the input image in low resolution and detecting an edge signal from the sampled pixels. A method for detecting a face and an eye position in a two-dimensional digital image including a face, (a) adjusting a reference value, which is a reference for determining the presence or absence of an edge signal, according to the brightness of the entire image with respect to the input image; (b) detecting an edge signal from the input image based on the reference value and generating an edge image; (c) detecting a face position in the edge image using a T-type search region; And (d) placing the T-shaped search region at the detected face position, and detecting a position of the eye by using a polarity of an edge signal in a horizontal direction in the T-shaped search region; Eye position detection method. The method of claim 8, wherein step (d) (d1) determining an eye search position using the T-type search region at the detected face position; (d2) detecting edge signals with respect to the eye search position, wherein the difference between the values of the pixels located above the vertical line is greater than the reference value having a positive value; And (d3) a face and eye position detection method, comprising detecting as an eye position an area where detected edge signals having a positive polarity are concentrated in the form of a horizontal line. The method of claim 9, wherein after step (d3), (d4) Applying the condition that the horizontal and vertical edge signals should be less than or equal to the reference value 1 and the condition that the vertical edge signals on both sides of the pupil should be greater than or equal to the reference value 2, both inside the black and the inside of the white, eye candidates are selected. Setting to a location; And (d5) detecting eye regions from the eye candidate positions obtained by repeating step (d4) over the entire eye region of the eye search position, in which the total amount of edge signals is maximum; Face and eye position detection method.