JP4380308B2 - Face area detection method and apparatus, program, and computer-readable storage medium storing face area detection program - Google Patents

Description

  The present invention relates to a face area detection method and apparatus, a program, and a computer-readable storage medium storing the face area detection program, and in particular, a face area for detecting a face area whose number or size is unknown from an image. The present invention relates to a detection method and apparatus, a program, and a computer-readable storage medium storing a face area detection program.

  Conventionally, when the size of the face area is unknown, a method of performing processing while changing the resolution of the image is used as one of methods for detecting the face area from the image. For example, there is template matching often used for pattern detection (see, for example, Non-Patent Document 1).

  In template matching, as shown in FIG. 11, a template t (x, y) representing an object to be detected is overlapped at a point (i, j) in an image f (x, y), and t This is a method of measuring the degree of similarity between (x, y) and the partial pattern of the overlapping image, and using the value as the probability that the target exists at the point (i, j).

  In the case of face detection, using the face image as a template, the above operation is performed on all the points of the image to find the point where the similarity is the maximum, or the point where the similarity exceeds a certain threshold do it.

  As a formula for calculating the similarity m (u, v) at the point (u, v), for example, there is a formula like Formula 1.

上記の式１では、（ｕ，ｖ）はテンプレートの中央ではなく、左上端の位置を表している。但し、この方法では、テンプレートと同サイズの顔領域しか検出できないため、顔領域のサイズが未知の場合には、図１２のように画像の解像度を変化させながら、各解像度におけるテンプレートとの類似度を求める必要がある。

In Equation 1 above, (u, v) represents the position of the upper left corner, not the center of the template. However, in this method, only a face area of the same size as the template can be detected. Therefore, when the face area size is unknown, the degree of similarity with the template at each resolution is changed while changing the resolution of the image as shown in FIG. It is necessary to ask.

When a face area is detected from a moving image, the above procedure is applied for each frame to obtain the area.
Introduction to computer image processing pp. 149-152 Hideyuki Tamura Research Institute Publishing

  However, when the above-described conventional method is used, if the size of the face area is unknown, the face area is detected while sequentially changing the resolution of the image, so that the process takes a long time.

  In the case of a moving image, the number of images to be processed increases drastically, and there is a problem that the processing time increases significantly.

  As a solution to this problem, if the number of existing face areas is known, processing can be speeded up by stopping processing when all face areas are found, but there are multiple face areas. This method cannot be used when the number is unknown.

  The present invention has been made in view of the above points, and a face area detection method, apparatus, program, and face area capable of reducing a face area detection time from an image in which a plurality of face areas of different sizes exist. It is an object to provide a computer-readable storage medium storing a detection program.

  FIG. 1 is a diagram for explaining the principle of the present invention.

The present invention (Claim 1) is a face area detection method in an apparatus for detecting a face area from an image including a still image and a moving image.
An image reading process (step 1) for reading an input image and storing it in a storage means;
A converted image obtained by reading one input image from the storage unit and reducing the number of pixels on each side in the vertical and horizontal directions from the input image and converting the image from the lowest possible resolution to the resolution of the input image to the extent that a face can be identified. A resolution conversion step (step 2) in which a predetermined number is created by converting the input image without changing the aspect ratio of the input image and stored in the storage means;
A face area detection process (step 3) in which face area detection processing is performed using a converted image in which the resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution conversion process;
The resolution from the finding for the first time face region if the number of times of processing reaches a predetermined number of times when performing the face area detecting process by raising the order by the face area detection process, end determining process to terminate the process of the face image detection process (Step 4).

The present invention (Claim 2) is a face area detection method in an apparatus for detecting a face area from a moving image.
An image reading process of reading an input moving image and storing it in a storage means;
A reference frame selection process of reading the first frame of the input moving image from the storage means and storing it in the storage means as a reference frame;
A face detection process for detecting a face area for a reference frame;
A comparison target frame selection process in which a frame subsequent to the reference frame is set as a comparison target frame;
A difference calculation process for obtaining a difference value between the reference frame and the comparison target frame;
If the difference value obtained in the difference value calculation process is equal to or smaller than a certain threshold value, the face area detected in the reference frame is determined as the face area in the comparison target frame, and the new comparison target frame is used as the next frame. Repeating the calculation process, and if the difference value is greater than a certain threshold value, performing a face area detection process and subsequent end processes using the comparison target frame as a new reference frame, and
The face detection process
A converted image obtained by reducing the number of pixels on each side in the vertical and horizontal directions from the reference frame image, and converting the converted image from the lowest possible resolution to the resolution of the reference frame image to the extent that the face can be identified. A resolution conversion process of creating a predetermined number of images by converting the image of the reference frame without changing the aspect ratio, and storing it in the storage means;
A face area detection process in which face area detection processing is performed using a converted image whose resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution conversion process;
In cases where processing count has reached the predetermined number of times when the finding for the first time the face region by the face region detection process by increasing the resolution in order perform the face area detecting process, end determination to end the processing of said pigment region detection process Process.

Further, according to the present invention (claim 3) , a difference value of each pixel value is obtained as a difference value in the difference value calculation process.

According to the present invention (Claim 4), in the difference value calculation process, the difference value is calculated using the difference of the color histogram of the entire image.

  FIG. 2 is a principle configuration diagram of the present invention.

The present invention (Claim 5) is a face area detection device for detecting a face area from an image including a still image and a moving image,
An image reading means 11 for reading an input image and storing it in the storage means 12;
A conversion image obtained by reading one input image from the storage unit 12 and reducing the number of pixels on each side in the vertical and horizontal directions from the input image, and converting from the lowest possible resolution to the resolution of the input image so that a face can be identified. A resolution conversion unit 13 for creating a predetermined number of images by converting the input image without changing the aspect ratio of the input image and storing the image in the storage unit 12;
A face area detection means 14 for performing face area detection processing using a converted image whose resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution conversion means 13;
In cases where processing count has reached the predetermined number of times when the finding for the first time the face region by the face region detection unit 14 to increase the resolution in order perform the face area detecting process, end determination to end the process of the face image detection process Means 16.

The present invention (Claim 6) is a face area detection device for detecting a face area from a moving image,
Image reading means for reading an input moving image and storing it in a storage means;
Reference frame selection means for reading out the first frame of the input moving image from the storage means and storing it in the storage means as a reference frame;
Face detection means for performing face area detection on the reference frame;
A comparison target frame selecting means to be a frame comparison target frame following the reference frame;
A difference value calculating means for obtaining a difference value between the reference frame and the comparison target frame;
If the difference value obtained by the difference value calculation means is equal to or less than a certain threshold value, a face area determination means for determining the face area detected in the reference frame as the face area in the comparison target frame;
A first repeating unit that repeats the processing of the comparison target frame selection unit, the difference value calculation unit, and the face region determination unit, with the new comparison target frame as the next frame after the face region determination unit;
A second repeater that repeats the process of the reference frame selector, the face detector, the difference value calculator, and the face area determiner if the difference value obtained by the difference value calculator is greater than a certain threshold value. ,
Face detection means
A converted image obtained by reducing the number of pixels on each side in the vertical and horizontal directions from the reference frame image, and converting the converted image from the lowest possible resolution to the resolution of the reference frame image to the extent that the face can be identified. A resolution conversion unit that creates a predetermined number of images by converting the image of the reference frame without changing the aspect ratio, and stores the resolution in the storage unit;
A face area detection means for performing face area detection processing using a converted image in which the resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution conversion means;
The first time if the number of times of processing reaches a predetermined number of times when the find a face region by increasing the resolution in order perform the face area detecting process, end determination to end the processing of said pigment region detection means by the face region detecting means Means.

Further, the difference value calculation means of the present invention (claim 7) obtains a difference value of each pixel value as the difference value.

Further, the difference value calculation means of the present invention (claim 8) calculates the difference value using the difference of the color histogram of the entire image.

The present invention (Claim 9) is a face area detection program for causing a computer to function as each means constituting the face area detection apparatus according to any one of Claims 5 to 8 .

The present invention (Claim 10) is a computer-readable storage medium storing the face area detection program according to Claim 9 .

  As described above, in the present invention, attention is paid to the shooting characteristics of a still image or a moving image. The feature on photography is that, in general, when photographing a person, there are many cases where the face is well photographed.

  Also, there are many cases where a photograph is taken so that a person who is paying attention in the image or a person having some meaning is photographed largely.

  For this reason, a face that is large is highly important, and a small face may be ignored in comparison. For example, when making a list of main characters.

  Therefore, in the present invention, face area detection processing is performed in the order of high resolution from low resolution images, and if the processing result is equal to or smaller than a predetermined threshold, it is determined that the face area is detected, The processing is terminated when processing is performed on an image having a higher resolution than the resolution of the image by a predetermined ratio.

  Further, when detecting a face area, there is a case where an accurate position is not necessary for the purpose of checking whether or not a person is present in the image. For example, it is a case where a specific person appearance scene is searched for in a moving image.

  Therefore, as another method, in the moving image, the first frame is used as a reference frame, and the difference between the pixel values at the same position is obtained between the reference frame and the core frame following the reference frame. If the sum of the values is equal to or less than a certain threshold value, there is almost no change between frames, so the face area detected in the reference frame is determined as the face area in that frame.

  If the sum of the absolute values of the differences is larger than a certain threshold value, the face area is obtained and set as a new reference frame. By repeating this, the face area is determined for the entire moving image.

  In other words, the detection process of frames with a small difference value between frames is skipped, and the frame interval for performing the detection process is adaptively adjusted.

  In the above invention, it is also possible to use the difference of the color histogram of the entire image instead of the difference of each pixel value, or to obtain the square sum of the difference instead of the absolute value sum of the differences.

  As described above, according to the present invention, it is possible to shorten the processing time by omitting the detection process of a small face area.

  Further, the processing time can be shortened by assuming that a face area exists in the same position as the previous reference frame, except for the face area detection process between the reference frames from the reference frame among the face areas existing in the moving image.

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

[First Embodiment]
FIG. 3 shows the configuration of the face area detection apparatus according to the first embodiment of the present invention.

  The face area detection device 10 shown in FIG. 1 includes an image reading unit 11 that reads an input image, a memory 12 that temporarily stores the read image, and reads the input image from the memory 12, converts the resolution, and stores it in the memory 12. The resolution conversion unit 13 that performs conversion, the face region detection unit 14 that performs face region detection processing from the low-resolution image to the high-resolution image, the template 15 that is referred to by the face region detection unit 14, and the end of the processing It is comprised from the process end determination part 16 which performs determination. Although the template 15 is included in the configuration shown in the figure, it is because template matching is used in face detection processing as an example of the present embodiment, and is not necessary for processing other than template matching. .

  FIG. 4 is a flowchart of face area detection processing according to the first embodiment of the present invention, in which an input image reading step (step 301), an input image resolution conversion step (step 302), and a face area detection step (step 303). , A repeat determination step (step 304) is shown.

  Hereinafter, the face area detection operation will be described in detail with reference to FIG.

  First, in step 301, the input to the image reading unit 11 is a digitized still image or an arbitrary single frame of a moving image. The image resolution, storage format, content, and recording medium are not particularly limited, but the read frame is temporarily stored in the memory 12 as an example in FIG.

  In step 301, an image to be subjected to face area detection processing is selected and stored in the memory 12 from a recording medium external to the apparatus. The selection of the image may be freely made at the operator's discretion.

  The resolution of the original image obtained here is (P × Q).

  Next, in step 302, the resolution converter 13 converts the original image (P × Q) read into the memory 12 in step 301 into a low resolution image (K × L).

  In the first loop, the image is converted to an image with as low a resolution as possible so that the face can be identified (for example, about 16 × 16).

  From the next loop, the image is converted to a higher resolution image than the image used in the previous loop. In the conversion, the aspect ratio of the image is not changed.

  Here, an example of converting the original image of FIG. 5 into the low-resolution image shown in FIG. 6 will be described. However, each pixel is assumed to correspond to each grid point.

  First, the pixel a ′ in the low-resolution image in FIG. 6 corresponds to the point a ′ at the same position in the original image in FIG. 5. Here, the periphery of the point a 'in FIG. 5 is as shown in FIG.

  At this time, assuming that the pixel value at each point is f (x, y), the pixel value f at the point a 'is expressed by Equation 2.

f = f (u, v) (1-α) (1-β) + f (u + 1, v) α (1-β)
+ F (u, v + 1) (1-α) β + f (u + 1, v + 1) αβ (2)
Similarly, the pixel values of other lattice points in the low-resolution image of FIG. 6 can be obtained from the pixel values of four points around the point at the same position in the original image.

  Next, in step 303, the face area detection unit 14 performs face detection processing on the low resolution obtained in step 302.

  The face detection processing method includes a method using template matching, a method using a neural network, a method using a support vector machine, and the like. Here, a method using template matching will be described.

  First, a face image (M × N) serving as the template 15 is prepared on a storage medium (memory).

  In this template, face images of a plurality of persons are cut into a size of (M × N), and an average value of each pixel of the plurality of face images is set as a value of a corresponding pixel of the template. As shown in FIG. 8, the center of the template t (x, y) overlaps with the point (i, j) in the low resolution image g (x, y) obtained in step 302, and t (x, y ) And the similarity m (i, j) between the overlapping partial pattern of the image is calculated by Equation 3.

同様に、低解像度画像ｇ（ｘ，ｙ）中の全ての点について類似度を算出し、その類似度ｍ（ｘ，ｙ）が予め定めた閾値以下であれば、点（ｘ，ｙ）に顔領域が存在すると見做す。また、全ての点の類似度が予め定めた閾値より大きければ何もしない。

Similarly, the similarity is calculated for all points in the low-resolution image g (x, y), and if the similarity m (x, y) is less than or equal to a predetermined threshold, the point (x, y) is calculated. It is assumed that a face area exists. If the similarity of all points is greater than a predetermined threshold value, nothing is done.

  By this step 303, the coordinates of the face area can be obtained.

  In step 304, it is determined whether or not to repeat the face detection process. If the following conditions are not satisfied, steps 302 to 304 are repeated while gradually increasing the resolution of the converted image used in the previous loop. In this case as well, the aspect ratio of the image is not changed.

  In step 303, when the loop processing is repeated a predetermined number of times from the loop in which the face area is detected for the first time, or when the converted image whose resolution is gradually increased for each loop matches the original image, the loop is terminated.

  According to the present embodiment, it is possible to shorten the processing time by omitting the detection process of a small area among the face areas existing in the image.

[Second Embodiment]
FIG. 9 shows a configuration of a face area detection apparatus according to the second embodiment of the present invention.

  The face area detection apparatus 20 shown in FIG. 1 includes an image reading unit 21 that reads an input image, a memory 22 that stores the read image, and the first frame stored in the memory 22 is selected as a reference frame and stored in the memory 22. The reference frame selection unit 23 and the reference frame stored in the memory 22 and the frame following the reference frame are set as comparison target frames, a difference value between the reference frame and the comparison target frame is obtained, and the difference value is equal to or less than a certain threshold value. If there is a face area detection unit 24 that determines a face area detected in the reference frame as a face area in the comparison target frame, a comparison target frame selection unit that acquires the next comparison target frame from the memory 22 and passes it to the face area detection unit 24 26, a template 25 referred to by the face area detection unit 24, and a process end determination unit that determines the end of the process of each unit 7 consists of.

FIG. 10 is a flowchart of face area detection processing according to the second embodiment of the present invention. The flowchart shown in FIG. 10 includes an input moving image diagram reading step (step 801), a reference frame selection step (step 802), and a face region. A detection processing step (step 803), a comparison target frame selection step (step 804), a repetition determination step (step 805), an inter-frame difference value calculation step (step 806), and a processing method determination step (step 80 7 ) are shown.

  The face area detection operation will be described in detail below with reference to FIG.

  First, in step 801, it is assumed that the image reading unit 21 receives a digitized moving image. Here, the resolution of the moving image, the storage format, the contents, and the recording medium are not particularly limited. The read moving image subject to face area detection processing is stored in the memory 22 from the recording medium.

  In step 802, the reference frame selection unit 23 selects a reference frame from the moving image obtained in step 801 and stores it in the memory 22. In the first loop, the first frame of the moving image is read from the memory 22 and selected as a reference frame.

  In step 803, the face detection processing unit 24 performs face detection processing on the reference frame obtained in step 802. The face detection processing method includes a method using template matching, a method using a neural network, a method using a support vector machine, and the like. Here, a method using template matching will be described.

  First, a face image (M × N) serving as a template is prepared on a storage medium (memory).

  The template 25 cuts face images of a plurality of persons into a size of (M × N), and uses the average value of each pixel of the plurality of face images as the value of the corresponding pixel of the template.

  As shown in FIG. 8, the face area detection unit 24 makes the center of the template t (x, y) overlap the point (i, j) in the reference frame image g (x, y) obtained in step 803. Then, the similarity m (i, j) between t (x, y) and the partial pattern of the overlapping image is calculated by the above-described equation 3.

  Similarly, the similarity is calculated for all points in the reference frame image g (x, y), and if the similarity m (x, y) is equal to or less than a predetermined threshold, the point (x, y) is calculated. It is assumed that a face area exists.

  If there are a plurality of points whose similarity is equal to or less than the threshold value, it is considered that a face region exists at all of the plurality of points. Nothing is done if the similarity of all points is greater than a predetermined threshold.

  In step 803, the coordinates of the face area can be obtained.

  In step 804, the comparison target frame selection unit 26 selects a frame to be compared from the memory 22. The comparison target frame is the next frame of the previous comparison target frame. For the first time, the frame next to the reference frame is set as a comparison target frame.

  In step 805, the process end determination unit 27 determines whether to repeat the process. In step 804, if the next frame of the comparison target frame does not exist, that is, if the processing is performed up to the last frame of the moving image, the loop is terminated. In other cases, loop processing is continued.

  In step 806, the face area detection unit 24 obtains a difference value between the reference frame and the comparison target frame. The difference value may be the sum of the difference values for each color in the color histogram of the entire image or the difference square sum of the corresponding pixel values at the same position. An example using the sum of absolute differences will be described.

  The difference value d between the reference frame image f (x, y) and the comparison target frame g (x, y) is calculated by Equation 4.

ステップ８０７では、顔領域検出部２４において、ステップ８０６で得られた差分値ｄに基づき、処理の分岐判定を行う。ステップ８０６で得られた差分値ｄが予め定めた閾値以下であれば、基準フレームと比較対象フレーム間で画像の変化が少ないため、基準フレームと同様の位置に顔領域が存在するとみなし、比較対象フレームの次フレームを新たな比較対象フレームとし、ステップ８０４〜８０７を繰り返す。

In step 807, the face area detection unit 24 determines whether to branch the process based on the difference value d obtained in step 806. If the difference value d obtained in step 806 is less than or equal to a predetermined threshold value, there is little change in the image between the reference frame and the comparison target frame, so it is considered that a face region exists at the same position as the reference frame, and the comparison target Steps 804 to 807 are repeated with the next frame as a new frame to be compared.

  If the difference value d obtained in step 806 is larger than a predetermined threshold value, it is necessary to update the face area position because the image changes greatly between the reference frame and the comparison target frame.

  Therefore, the face area detection unit 24 passes the comparison target frame to the reference frame selection unit 23 for updating the face area position, and the reference frame selection unit 23 sets the comparison target frame as a new reference frame. Select and repeat steps 802-807.

  According to the present embodiment, the processing time is eliminated by omitting the face area detection process from the reference frame to the reference frame among the face areas existing in the moving image and assuming that the face area exists at the same position as the previous reference frame. Can be shortened.

  Further, the operations shown in the flowcharts of FIG. 4 in the first embodiment and FIG. 10 in the second embodiment are installed in a computer used as a face area detection device and executed by a control means such as a CPU. Or distributed via a network.

  In addition, the constructed program is stored in a hard disk device connected to a computer used as a face area detection device, a portable storage medium such as a flexible disk, a CD-ROM, and the like. It is also possible to install.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

  The present invention can be applied to a system having a function of detecting a face area in an image processing technique, a face authentication system, a security system, or the like.

It is a figure for demonstrating the principle of this invention. It is a principle block diagram of this invention. It is a block diagram of the face area detection apparatus in the 1st Embodiment of this invention. It is a flowchart of the face area detection process in the 1st Embodiment of this invention. It is an example of the high resolution image in the 1st Embodiment of this invention. It is an example of the low-resolution image in the 1st Embodiment of this invention. It is a figure which shows the resolution conversion calculation image in the 1st Embodiment of this invention. It is an example of the template matching in the 1st Embodiment of this invention. It is a block diagram of the face area detection apparatus in the 2nd Embodiment of this invention. It is a flowchart of the face area detection process in the 2nd Embodiment of this invention. It is a schematic diagram of conventional template matching. It is a figure which shows the conventional image resolution conversion.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Face area detection apparatus 11 Image reading means, Image reading part 12 Memory | storage means, Memory 13 Resolution conversion means, Resolution conversion part 14 Face area detection means, Face area detection part 15 Template 16 End determination means, Process end determination part 20 Face area Detection device 21 Image reading unit 22 Memory 23 Reference frame selection unit 24 Face area detection unit 25 Template 26 Comparison target frame selection unit 27 Processing end determination unit

Claims (10)

In a face area detection method in an apparatus for detecting a face area from an image including a still image and a moving image,
An image reading process of reading an input image and storing it in a storage means;
A conversion image obtained by reading out one input image from the storage means and reducing the number of pixels on each side in the vertical and horizontal directions from the input image, and converting from the lowest possible resolution to the resolution of the input image so that a face can be identified. A resolution conversion process of creating a predetermined number of images by converting the input images without changing the aspect ratio of the input images, and storing the images in a storage unit;
A face area detection process in which face area detection processing is performed using a converted image in which the resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution conversion process;
In cases where processing count has reached the predetermined number of times when performing the face area detecting process by increasing the resolution in order from finding a first face region by the face region detecting step, the end to end process of the face image detection process Judgment process,
A face area detection method characterized by: In a face area detection method in an apparatus for detecting a face area from a moving image,
An image reading process of reading an input moving image and storing it in a storage means;
A reference frame selection process of reading the first frame of the input moving image from the storage means and storing it in the storage means as a reference frame;
A face detection process for performing face area detection on the reference frame;
A comparison target frame selection process in which a frame subsequent to the reference frame is a comparison target frame.
A difference calculation process for obtaining a difference value between the reference frame and the comparison target frame;
If the difference value obtained in the difference value calculation process is equal to or smaller than a certain threshold value, the face area detected in the reference frame is determined as the face area in the comparison target frame, and a new comparison target frame is set as the next frame. The difference calculation process is repeated, and if the difference value is larger than a certain threshold value, an end process of repeating the face area detection process and the subsequent steps using the comparison target frame as a new reference frame is performed.
The face detection process includes:
A converted image obtained by reducing the number of pixels on each side in the vertical and horizontal directions from the image of the reference frame, and a converted image from the lowest possible resolution to the resolution of the image of the reference frame to the extent that a face can be identified. A resolution conversion process in which a predetermined number is created by converting the image of the reference frame without changing the aspect ratio of the image, and stored in the storage means;
A face area detection process in which face area detection processing is performed using a converted image in which the resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution conversion process;
In cases where processing count has reached the predetermined number of times when performing the face area detecting process by increasing the resolution in order from finding a first face region by the face region detecting step, the end to end the processing of said pigment region detection process Judgment process,
A face area detection method characterized by: In the difference value calculation process,
The face area detection method according to claim 2, wherein a difference value of each pixel value is obtained as the difference value. In the difference value calculation process,
The face area detection method according to claim 2, wherein the difference value is calculated using a difference of a color histogram of the entire image. A face area detection device for detecting a face area from an image including a still image and a moving image,
Image reading means for reading an input image and storing it in a storage means;
A conversion image obtained by reading out one input image from the storage means and reducing the number of pixels on each side in the vertical and horizontal directions from the input image, and converting from the lowest possible resolution to the resolution of the input image so that a face can be identified. A resolution conversion unit that creates a predetermined number of images by converting the input image without changing the aspect ratio of the input image, and stores the image in a storage unit;
A face area detecting means for performing a face area detecting process using a converted image whose resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution converting means;
After finding the first face area if the number of times of processing reaches a predetermined number of times when performing the face area detecting process by increasing the resolution in the order by the face region detecting means, end determining unit for terminating said pigment image detection means And a face area detecting device. A face area detection device for detecting a face area from a moving image,
Image reading means for reading an input moving image and storing it in a storage means;
Reference frame selection means for reading the first frame of the input moving image from the storage means and storing it in the storage means as a reference frame;
Face detection means for performing face area detection for the reference frame;
Comparison target frame selection means to be a frame comparison target frame following the reference frame;
Difference value calculating means for obtaining a difference value between the reference frame and the comparison target frame;
If the difference value obtained by the difference value calculation means is equal to or less than a certain threshold value, a face area determination means for determining a face area detected in the reference frame as a face area in the comparison target frame;
A first repeating unit that repeats the processing of the comparison target frame selecting unit, the difference value calculating unit, and the face region determining unit with a new comparison target frame as the next frame after the face region determining unit;
If the difference value obtained by the difference value calculation means is larger than a certain threshold value, a second repetition means for repeating the processes of the reference frame selection means, the face detection means, the difference value calculation means, and the face area determination means And having
The face detection means includes
A converted image obtained by reducing the number of pixels on each side in the vertical and horizontal directions from the image of the reference frame, and a converted image from the lowest possible resolution to the resolution of the image of the reference frame to the extent that a face can be identified. A resolution conversion unit that creates a predetermined number of images by converting the image of the reference frame without changing the aspect ratio of the image, and stores the number in a storage unit;
A face area detecting means for performing a face area detecting process using a converted image whose resolution is increased in order from the converted image having the lowest resolution among the converted images converted by the resolution converting means;
After finding the first face area if the number of times of processing reaches a predetermined number of times when performing the face area detecting process by increasing the resolution in the order by the face region detecting means, the end to end process of said pigment region detection means A determination means;
A face area detecting apparatus characterized by comprising: The difference value calculating means includes:
The face area detection device according to claim 6, wherein a difference value of each pixel value is obtained as the difference value. The difference value calculating means includes:
The face area detection apparatus according to claim 6, wherein the difference value is calculated using a difference of a color histogram of the entire image.   A face area detection program for causing a computer to function as each means constituting the face area detection device according to claim 5.   A computer-readable storage medium storing the face area detection program according to claim 9.

Images