JPH10162118A - Device and method for image processing - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect a specified image from one sheet of image data speedily with a simple constitution without using any background information. SOLUTION: Image data from an image data supply part 10 is divided by a band division part 20 into fHL including a horizontal high-frequency component and a vertical low-frequency component, fLH including a vertical high-frequency component and a horizontal low-frequency component, and fLL consisting of a low-frequency component. An edge detection part 30 detects a part, where an edge is present in a T shape by using fLH and fHL. Further, a symmetry axis detection part 40 calculates the degree of symmetry as to the part detected by the edge detection part 30 and a matching part 50 matches the part which is thus limited against a face standard pattern to detect the position of the face.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and method for automatically detecting the position of a predetermined image such as a human face from image data. More specifically, the present invention provides:
Automatically detects the position of a predetermined image such as a human face from image data, and tracks the human face, specifies the image to be transmitted and compressed,
Alternatively, the present invention relates to an image processing method and apparatus applicable to a surveillance camera, a video conference system, a videophone, a video camera, a video data editing apparatus, and the like with functions such as focusing and automatic skin color adjustment.

[0002]

2. Description of the Related Art Conventionally, in an automatic face detection method and apparatus,
The following method is used. That is, a difference from a known background is obtained, and a face is automatically detected assuming that a face exists at a position where the difference value is large.A difference between frames or fields of a television image is obtained, and a face exists at a portion where the difference is large. Automatically detects faces as a means to detect them, Automatically detects faces by searching for skin-colored areas from color video signals, Similar patterns based on some knowledge about faces (eg, contour models and statistical features) To automatically detect faces by searching for images, assuming a plain background, calculating the projection values of the image data in the horizontal and vertical directions, and automatically detecting faces based on these projection values. General.

[0003]

However, the above-mentioned conventional methods have the following problems. In other words, in the method, since the background information is known, the application range is limited to those installed in a predetermined place such as a video conference or a surveillance camera. Expensive image memory is required. In addition, it is difficult to correctly detect when a moving object other than a person is included in the screen or the main body of the image input device (for example, a camera) moves. Change, which imposes severe conditions on the lighting.
It is necessary to search for a region similar to the face pattern in a wide range of the image by, for example, a correlation operation. For this reason, the method of requiring a large amount of calculation and requiring a low-speed or large-scale device has a problem that the application is limited when the background is plain and the versatility is poor.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has been made to enable a simple configuration and high speed detection of a specific image from one piece of image data without using background information. It is an object of the present invention to provide an image processing method and an image processing method.

It is another object of the present invention to provide an image processing method and apparatus which are not easily affected by changes in illumination conditions and can perform stable image detection.

Another object of the present invention is to enable a simple area detection and edge processing to quickly remove a plain area or a low-contrast area where image information is scarce from a target area for face search by a simple edge detection and edge processing. It is an object of the present invention to provide an image processing method and an apparatus capable of detecting a complicated face.

Another object of the present invention is to extract a symmetrical area from a complicated background, thereby greatly reducing the burden of the subsequent face search and enabling high-speed face detection. An image processing method and apparatus are provided.

[0008]

The image processing method of the present invention for achieving the above object has the following steps. That is, an image processing method for automatically detecting a predetermined image from image data, comprising: a dividing step of dividing the image data into bands; and searching for the predetermined image based on the image data band-divided by the dividing step. A limiting step of limiting an area to be performed; and a detecting step of detecting a position of the predetermined image in the area limited by the limiting step.

According to another aspect of the present invention, there is provided an image processing apparatus for automatically detecting a predetermined image from image data, comprising: a dividing unit for dividing the image data into bands; A limiting unit that limits an area in which the predetermined image is to be searched based on the image data band-divided by the unit; and a detecting unit that detects a position of the predetermined image in the area limited by the limiting unit.

[0010]

According to the embodiment described below, a band dividing means divides image data to be processed into a high frequency component and a low frequency component in horizontal and vertical directions. And
The face area limiting means limits an area (face area) where a face may exist by using the image data divided by the band dividing means. Here, in the band dividing means, processing for reducing the resolution of the image or the sampling plate is performed. Then, the face area limiting means limits the face area by directly using the output signal of the band dividing means. For this reason, a fast and simple face detection method and apparatus can be provided.

Further, the edge is limited to T
A portion distributed in a character shape is extracted. For this reason, a low-contrast region having a small number of edges and a contour region of an object on a plain background having no T-shaped edge distribution are removed from the face region. Therefore, face detection can be easily performed from an image having a relatively flat background. Also, for the edge detection, the high-frequency component in the horizontal direction and the high-frequency component in the vertical direction of the output of the band dividing means are directly used, so that extremely high-speed processing can be performed.

Further, according to the face area limiting means, an area having a high degree of left / right symmetry is extracted. Even in an image having a complicated background, by extracting a unique region having left / right symmetry, the region where the face exists can be effectively limited. Also, since the output of the band dividing means is directly used for the symmetric search, extremely efficient processing is possible.

Hereinafter, preferred embodiments of the present invention will be described more specifically with reference to the accompanying drawings.

(First Embodiment) FIG. 1 is a block diagram showing a schematic configuration of an image processing apparatus according to the present embodiment. In FIG. 1, reference numeral 101 denotes a CPU, which implements various processes of the image processing apparatus by executing a control program stored in a ROM 102 or the like. A ROM 102 stores a control program executed by the CPU 101 and various data. A RAM 103 provides a work area when the CPU 101 executes various controls. Reference numeral 104 denotes an external storage device, for example, a hard disk, a floppy disk, or the like. In the external storage device 104,
Application programs, image data, and the like can be stored. An input unit 105 includes a keyboard, a mouse, and the like. 106 is a display, a CPU
Various displays are performed under the control of 101.

Reference numeral 107 denotes a camera interface.
The video data captured by the camera 108 is converted into digital data and output to the bus 109. Reference numeral 108 denotes a camera, which includes an image sensor such as a CCD. Reference numeral 109 denotes a bus that connects the above-described components. In the above description, the CPU 101 executes the program stored in the ROM 102. However, the CPU 101 may load the program stored in the external storage device 104 into the RAMU 103 and execute the program by the CPU 104.
The image data obtained by the camera 108 is stored in the external storage device 104 or the RAM 103. The device for inputting an image is not limited to the above-described camera 108, but may be, for example, a scanner or the like. 1
Reference numeral 10 denotes a face position detection unit that detects the position of a face from image data.

FIG. 2 is a block diagram showing a functional configuration related to a face position detecting unit in the image processing apparatus according to the first embodiment. In FIG. 1, reference numeral 10 denotes an image data supply unit which supplies image data to the system. The image data supply unit 10 may be an image data input unit configured by a camera 108 (or a scanner or the like), an external storage device 104 that stores image data, a RAM 103
May be an image data storage unit constituted by
Reference numeral 20 denotes a band division unit, 30 denotes an edge detection unit, 40 denotes a symmetry axis detection unit, and 50 denotes a matching unit for determining a face position. Details of the function of each unit will be described later. The output of the matching unit 50 is output as data indicating the position of the face.

Next, the operation of the functional configuration of FIG. 2 will be described in detail along the signal flow with reference to FIG. FIG. 3 is a flowchart showing the procedure of face position detection according to the first embodiment. Image data is input from an image data supply unit 10 such as an image input unit or an image data storage unit (Step S10). Note that the band dividing unit 20 divides the image into each spatial frequency band, and further reduces the image size by downsampling (decimation). This is to reduce the amount of image data to be processed by downsampling and to make the subsequent processing faster.

Hereinafter, for simplicity, the input image is represented by f (x, y),
Downsampled sub-images are denoted by fLL, fLH, fH
Notated as L, fHH. Here, fLL is a sub-image composed of horizontal and vertical low-frequency components of f (x, y), fLH is a sub-image composed of horizontal low-frequency components and vertical high-frequency components of f (x, y), fHL is a sub-image consisting of a horizontal high-frequency component and a vertical low-frequency component of f (x, y), and fHH is f (x, y)
Is a sub-image composed of high-frequency components in the horizontal and vertical directions.
FIG. 4 is a diagram illustrating an example of the sub-image. In FIG. 4, the upper left is f
LL, upper right corresponds to fLH, lower left corresponds to fHL, and lower right corresponds to fHH. Both fLH and fHL are input to the edge detection unit 30 and the symmetry axis detection 40. Further, fLL is input to the matching unit 50 that determines the position of the face. Note that fHH is not used in this embodiment because image information is scarce.

The fHL and fLH obtained as described above are supplied to the edge detection unit 30, where edge detection is performed, and an effective block is extracted as a region to be processed (step S11). FIG. 5 is a flowchart showing an outline of the processing in the edge detection unit 30. FIG.
5 is a flowchart showing a more detailed procedure of processing by the edge detection unit 30. 70 is an edge detection process;
An edge image is obtained from the input sub-images fHL and fLH. According to the flowchart of FIG. 5, first,
The sub-images fHL and fLH are input (step S701), threshold processing is performed on these sub-images, and binarization is performed.
f′HL and f′LH are obtained (step S702). Subsequently, in step S703, the logical sum (OR) of the binarized f'HL and f'LH as described above is calculated.
The image thus obtained is an edge image.

FIG. 7 is a diagram specifically showing a processing example of the edge detection processing 70. The upper left sub-image is fLL, and the upper right sub-image is a sub-image f'LH obtained by binarizing fLH with a predetermined threshold. The lower right sub-image is a sub-image f'HL obtained by binarizing fLH with a predetermined threshold, and the lower right is a logical sum of these two sub-images (f'LH and f'HL).
That is, it represents an edge image.

Subsequently, the edge image obtained by the edge detection processing 70 is supplied to an edge block extraction processing 80. In the edge block extraction processing 80, the edge image is converted into a small block (for example, 8 ×
(8 pixels) (step S801), and the number of edges included in each block is counted (step S8).
02). Then, a block including a certain number of edges or more is extracted by performing threshold processing on the counted number of edges (step S803). The block extracted in this way is called an edge block.

FIG. 8 is a diagram for explaining the correspondence between an edge image and an edge block. 8A is the edge image shown in FIG. 7, and FIG. 8B is a result of extracting a block including a predetermined number or more of pixels of the edge portion from the edge image of FIG. 8A. . In this example, when the block size is 8 × 8 pixels and at least 8 pixels out of a total of 64 pixels in one block are edges, the block is displayed in black as an edge block. As is clear from FIG.
At this point, the already plain background will be excluded.

Next, in the face edge extraction processing 90, T-shaped edge blocks are extracted from the edge blocks extracted in the processing 80 (step S901). The processing in the face edge extraction processing 90 will be described in more detail with reference to FIGS.

FIG. 9 is a diagram showing T-shaped distributed edge blocks. Since the edges detected from the face are distributed in a T-shape, only the edge blocks distributed in such a T-shape are selectively extracted to restrict the face detection target area. Hereinafter, the edge block thus extracted is referred to as an effective edge block. The extracted effective edge blocks are output to the symmetry axis detection unit 40.

FIG. 10 is a diagram showing an output example of the face edge extraction processing 90. By selecting only the region where the edge blocks are distributed in a T-shape, isolated blocks and blocks located on the contour of an object having a flat background are removed, and the subsequent processing (processing by the symmetry axis detecting unit 40, In the processing by the matching unit 50), the area to be processed can be significantly limited.

In this example, a T-shape (FIG. 9) consisting of two blocks in the left, right, and downward directions around the block of interest is used. ,
It is changed according to the size of the block used. Further, not all the T-shaped neighboring blocks of interest (in this example, the attention block and the six blocks) need to be edge blocks. For example, if a certain number or more of the focused block and the neighboring blocks to be focused on are edge blocks, a loose reference setting may be made such that a valid edge block is selected and a corresponding edge block is selected. Note that these reference settings are set in the input unit 10.
Perform from step 5.

Further, in case that the size of the input face is completely unknown, the edge detection processing 70 and the edge block extraction processing are performed in several different sizes, for example, from the screen size to the minimum size of 12 × 12 pixels. 80, a face edge extraction process 90 can also be executed. in this way,
Perform each process in multiple stages of size, the size of the degree of symmetry and the value of the correlation coefficient σ that are high in the process of each size is selected as the appropriate size for the input image at that time, and the subsequent process Used. The output of the edge detection unit 30 as described above controls the calculation area of the symmetry axis detection unit 40 as a map indicating a target area in the subsequent processing.

In step S12 (FIG. 3), the degree of symmetry is detected in each part of the effective block area extracted as described above. Hereinafter, the symmetric axis detection unit 40
Will be described. FIG. 11 is a flowchart illustrating a processing procedure of the symmetry axis detection unit 40. Symmetry axis detector 40
Is executed only in the area validated by the edge detection unit 30.

First, fHL and fLH are input from the band separation unit 20, and an effective edge block is input from the edge detection unit 30 (step S401). Then, a symmetry axis is detected based on these data (step S4).
02). The detection of the axis of symmetry is performed based on the directionality of the gradient (gradient) of the image data. The direction θ of the gradient vector is expressed by the following equation (1).

[0030]

(Equation 1)

Here, the gradient on the right side of the above equation (1) is equivalent to the high frequency component of the image, and the following equation (2) is established.

[0032]

(Equation 2)

Therefore, in the present embodiment, the following equation (3) is executed instead of equation (1) by directly using the outputs fLH and fHL of the band dividing section 20 as the direction θ of the gradient vector. Ask.

[0034]

(Equation 3)

The above equation (3) may be executed by software, or may be executed by hardware using a look-up table.

Next, in step S403, the degree of symmetry γ at each position (x, y) of the image is determined. γ
(X, y) is defined by the following equation (4) using θ (x, y).

[0037]

(Equation 4)

Here, equation (4) calculates the difference between the left and right θ in this block. It should be noted that the calculation block range is made to substantially match the size of the face to be detected. If the size of the face is unknown, B having a different size in several stages may be used. FIG.
2 is a diagram showing the correspondence between the band-separated image data fLL (a) and the degree of symmetry γ (b). When the symmetry is high, both the first and second terms of equation (4) take smaller values for the cancellation within the term. Further, although the expression (4) is obtained by further squaring the absolute value, the degree of symmetry γ may be simply a sum of absolute values.

The dark vertical stripes in FIG. 12B indicate the axis of symmetry of the face. In FIG. 12B, γ is calculated and displayed for the entire image for the sake of explanation. However, in practice, the calculation for determining the degree of symmetry is performed in the area where the effective edge block is extracted by the edge detection unit 30. Only executed. FIG. 13 is a diagram showing a state in which the degree of symmetry γ has been calculated in the effective edge block. That is, FIG. 13 is equal to the logical product of FIG. 10 and FIG.

When the degree of symmetry γ is calculated as described above, the process proceeds to step S404 (FIG. 10), and a portion where the calculated γ has a value equal to or larger than a predetermined threshold is extracted. FIG. 13 is a diagram illustrating a result of performing threshold processing on γ and further limiting an area (face candidate area) that is likely to be a face. In the black island portion in FIG.
0 (FIG. 2) is executed, and the position of the face is determined (steps S13 and S14).

The matching section 50 includes an edge detection section 30,
As a result of the processing by the symmetry axis detecting unit 40, the most likely face position is determined within the limited face candidate area (FIG. 14). Specifically, a typical face is stored as a template,
Search for a location that is most similar to the template.

As a criterion for similarity determination, a normalized cross-correlation coefficient that is strong against changes in illumination conditions is used. The coefficient σ (x, y) is defined by the following equation (5).

[0043]

(Equation 5)

Where f (x, y) is an image, w (x, y) is a template, and f bar is f (x, y) overlapping ω (x, y).
And w bar is the average value of w (x, y).

FIG. 15 is a diagram showing a region where the normalized cross-correlation coefficient σ is 0.35 or more in the region of FIG. 14 (σ is scaled from −1 to +1 by the equation (5)). And the closer to 1, the more similar).

The black island shown in FIG. 15 shows an area where the possibility of being a face is extremely high as a result of processing according to the above equation (5). The white square in FIG. 16 is a view in which the place where σ is the maximum is displayed on the fLL.
It can be seen that the best matching point correctly detects the position of the face (between both eyes).

The face determination process performed by the matching unit 50 is as follows.
In addition to the above-described correlation (correlation calculation), any existing method such as elastic matching that allows some deformation of the face pattern and Fourier transform method may be used, and is not particularly limited to the above embodiment.

As described above, according to the above-described embodiment, the processing amount of the process of detecting the symmetry is reduced by extracting the valid edge blocks, and the pattern matching is performed only on the portion having a high degree of the symmetry. By determining the position of the face, the processing amount of pattern matching is greatly reduced. Therefore, high-speed face position detection becomes possible.

Further, according to the first embodiment: 1. Since the sampling rate is reduced by utilizing the band division unit 20 (FIG. 2), the face search can be executed efficiently. The output of the band dividing unit 20 is output to the subsequent edge detecting unit 3
0, since it is directly used by the symmetry axis detection unit 40 (FIG. 2), the calculation is performed at high speed and the hardware can be easily realized. 2. It is particularly effective in removing a plain background and a low contrast area. Since the symmetry axis detection unit 40 uses the direction of the gradient of the image data, it is hardly affected by changes in illumination conditions,
Stable axis of symmetry detection is possible. In addition, since a unique area having a left / right symmetry is searched for, there is an effect that the face area can be significantly limited even in a complicated background.

Further, by repeatedly applying the processing by the band dividing section 20, it is possible to generate images having different resolutions for one input image. The processing of the edge detection unit 30 and the symmetry axis detection unit 40 is performed on each of these images having different resolutions, and by selecting appropriate information from the images, the above-described multi-step block size is used. The same effect can be obtained as in the case where images with different scalings are input.

(Second Embodiment) In the first embodiment, the edge detection unit 30 is used as a method for limiting the face area.
(FIG. 2) and a symmetric axis detection unit 40 (FIG. 2). However, in an environment where the background is plain, the face area can be limited by the edge detection unit 30 alone. Conversely, when a complex background is targeted, the face area may be limited by the symmetric axis detection unit alone. As described above, according to the second embodiment, unnecessary processing can be removed according to the background environment to be handled, and more efficient and high-speed face detection can be realized.

FIG. 17 is a block diagram showing a functional configuration related to a face position detecting unit in the image processing apparatus according to the second embodiment. In the figure, the same components as those of the first embodiment (FIG. 2) are denoted by the same reference numerals, and description thereof is omitted here.
In the second embodiment, when detecting the position of the face, the mode
1, 2, and 3 can be set. In mode 1, as in the first embodiment, the symmetry axis detection unit 40 detects the degree of symmetry γ in the area of the effective block extracted by the edge detection unit 30 and has a γ value equal to or greater than a predetermined value. The matching unit 50 performs a matching process on the region. In the mode 2, the matching unit 50 executes the matching process on the area of the effective block detected by the edge detection unit 30. Further, in the mode 3, the symmetry axis detection unit 40 calculates the degree of symmetry γ over the entire region of the input image, and the matching unit 50 performs matching processing on a portion having a γ value equal to or greater than a predetermined value.

In FIG. 17, reference numeral 100 denotes a switch.
The signal for designating a region in which the symmetry axis detection unit 40 detects the symmetry axis (detects the degree of symmetry) is switched. Mode 1
Then, since the symmetry axis detection unit 40 performs processing on the area of the effective block from the edge detection unit 30, the switch is switched to the a side to supply the output of the edge detection unit 30 to the symmetry axis detection unit 40. In the mode 3, the symmetry axis detection unit 40 calculates the degree of symmetry γ for the entire region, so that the switch of the switch 100 is switched to the b side.

The switching of the switch 110 is also controlled according to the mode. That is, in the mode 2, the edge detection unit 3
Since the matching unit 50 performs processing on the area of the effective block extracted by 0, the switch of the switch 110 is switched to the c side. On the other hand, in the modes 1 and 3, since the symmetry axis detection unit 40 processes the portion where the degree of symmetry γ is equal to or more than the predetermined value, the switch of the switch 110 is switched to the d side.

FIG. 18 is a flowchart for explaining the flow of the processing of the second embodiment. In this figure, the same processing steps as those in FIG. 3 are denoted by the same step numbers.
As shown in the drawing, when mode 1 is designated (the mode is designated by the input unit 105), the switch 100 is switched to the a side, and the switch 110 is switched to the d side. As a result, the same processing as the procedure in FIG. 3 is executed via the branches of steps S21 and S22.
In the case of the mode 2, since the switch 110 is switched to the c side, the process proceeds from step S22 to step S23, and the matching unit 50 executes the matching process on the area of the effective block extracted by the edge detection unit 30. Then, in step S14, the part with the highest matching degree (best matching point) is determined as the position of the face. Further, when the mode 3 is designated, the switch 1
00 is switched to the b side, and the switch 110 is switched to the d side.
Therefore, the process proceeds from step S21 to step S24, and the degree of symmetry γ is calculated from fLH and fHL for the entire image. Then, in step S13, the matching unit 50 performs matching on a portion where the degree γ is equal to or more than a predetermined value, and determines the best matching point as the face position in step S14.

As described above, according to the second embodiment,
The content of the process for limiting the execution range of the matching process for face position detection can be set according to the state of the background or the like. Therefore, more efficient and faster face position detection processing is realized.

The face position detection process described in each of the above embodiments may be realized by dedicated hardware, or software for realizing the above-described control may be implemented by the CPU 101.
May be realized by executing, or may be realized by appropriately mixing software and hardware. The software may be stored in the ROM 102 or may be loaded from the external storage device 104 into the RAM 103.

(Third Embodiment) In the first embodiment, the directionality of the image gradient is used as an index of symmetry of the symmetry axis detection unit 40 (FIG. 2). However, the method of detecting the axis of symmetry is not limited to such a method. For example,
The left and right differences in the block are calculated by directly using the image gradient values themselves according to equation (6) for finding the left / right difference of the absolute value of the gradient and equation (7) for finding the left / right difference itself. More simply, or more simply, the image value is calculated by the equation (8) or (9) for calculating the sum of absolute values of the left and right differences, or the equation (10) of the cross correlation (correlation calculation) on the left and right. The symmetry may be calculated using (luminance value).

[0059]

(Equation 6)

[0060]

(Equation 7)

[0061]

(Equation 8)

[0062]

(Equation 9)

[0063]

(Equation 10)

As described above, according to the third embodiment, by directly using the image gradient value, the direction θ of the gradient value is used.
Can be omitted, and a simpler circuit configuration (or program configuration) can be achieved. In the method using image values, for example, if cross-correlation is performed between the left and right in the block, the average value and the variance value in the block required for the correlation operation are determined by the final-stage matching unit 50 (FIG. 2). It is possible to share with.
Therefore, there is an effect of reducing the burden of the correlation operation amount requiring a large amount of calculation.

Even if the present invention is applied to a system composed of a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), an apparatus (for example, a copying machine, a facsimile, etc.) comprising one device Device).

Further, an object of the present invention is to supply a storage medium storing a program code of software for realizing the functions of the above-described embodiments to a system or an apparatus, and to provide a computer (or CPU) of the system or the apparatus.
And MPU) read and execute the program code stored in the storage medium.

In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD
-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

When the computer executes the readout program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer based on the instruction of the program code. ) May perform some or all of the actual processing, and the processing may realize the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, based on the instruction of the program code, It goes without saying that the CPU included in the function expansion board or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

When the present invention is applied to the storage medium, the storage medium stores program codes corresponding to the above-described flowcharts. Each module shown will be stored in a storage medium.

That is, the program codes of at least the "division processing module", "limitation processing module" and "detection processing module" may be stored in the storage medium. Here, the cholera module constitutes a control program for automatically detecting the position of a face image from image data, and the division processing module realizes division processing for dividing the image data into bands. In addition, the limitation processing module implements limitation processing for limiting an area in which a face image is to be searched using image data that has been band-divided by the division processing. Then, the detection processing module implements a detection process for detecting the position of the face in the area limited by the limitation process.

As described above, according to the above embodiments, it is possible to detect a face quickly and efficiently. Further, by the processing of the edge detection unit 30, a low-contrast region or a plain region can be removed at high speed. Furthermore,
By providing the symmetry axis detection unit 40, a unique region of left / right symmetry can be extracted even in a complicated background, and the region where the face exists can be largely limited. In particular, according to the above-described embodiment, since the detection of the symmetry axis by the symmetry axis detection unit 40 uses the direction of the gradient of the image data, it is hardly affected by changes in the illumination condition, and the left / right symmetry area is more effectively removed. It can be extracted reliably.

Further, by applying the present invention to a surveillance camera, a video conference system, and a video phone, it is possible to realize an automatic tracking function, or to enable transmission and compression of only a predetermined image, thereby improving transmission capability. it can.

By applying the present invention to a video camera, automatic focusing on a predetermined image and automatic adjustment of skin color can be realized.

Further, by applying the present invention to a video editing apparatus, only a predetermined image such as a human face can be automatically detected from a huge amount of data, and the editing work can be greatly shortened.

Further, the present invention makes it possible to prevent illegal access to buildings, offices, ATMs and terminals requiring security, to collate passport photos with a person at customs, to automate the search of a face database, and the like. Application to a man-machine interface is also possible.

[0078]

As described above, according to the present invention, it is possible to detect a predetermined image from one piece of image data at a high speed with a simple configuration without using background information.

Further, according to the present invention, it is hard to be affected by a change in illumination condition, and stable image detection can be performed.

Further, according to the present invention, it is possible to quickly remove a plain area and a low-contrast area where image information is poor from a face search target area by simple edge detection and edge processing. For this reason, an area to be subjected to face detection is limited, and high-speed face detection can be performed.

Further, according to the present invention, by extracting a symmetrical area from a complicated background, it is possible to greatly reduce the load of the subsequent face search processing, and high-speed face detection is realized.

[0082]

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an image processing apparatus according to an embodiment.

FIG. 2 is a block diagram illustrating a functional configuration related to a face position detection unit in the image processing apparatus according to the first embodiment.

FIG. 3 is a flowchart illustrating a procedure of face position detection according to the first embodiment.

FIG. 4 is a diagram illustrating an example of a sub image.

FIG. 5 is a flowchart showing an outline of processing in an edge detection unit 30;

FIG. 6 is a flowchart illustrating a more detailed procedure of a process performed by the edge detection unit 30;

FIG. 7 is a diagram specifically showing a processing example of edge detection processing 70;

FIG. 8 is a diagram illustrating the correspondence between an edge image and an edge block.

FIG. 9 is a diagram illustrating edge blocks distributed in a T shape.

10 is a diagram illustrating an output example of a face edge extraction process 90. FIG.

FIG. 11 is a flowchart illustrating a processing procedure of a symmetric axis detection unit 40;

FIG. 12 is a diagram showing a correspondence relationship between band-separated image data fLL and a degree of symmetry γ.

FIG. 13 is a diagram illustrating a result obtained by performing threshold processing on γ and further limiting an area (face candidate area) that is likely to be a face.

FIG. 14 is a diagram illustrating a state in which a portion having a degree of symmetry γ equal to or greater than a predetermined value is extracted from the state illustrated in FIG. 13;

FIG. 15 is a diagram illustrating a state in which a portion having a high degree of matching with a standard face pattern is extracted from the portion illustrated in FIG. 14;

16 is a diagram showing the correspondence between the highest matching part and the original image in the part shown in FIG.

FIG. 17 is a block diagram illustrating a functional configuration related to a face position detection unit in the image processing device according to the second embodiment.

FIG. 18 is a flowchart illustrating a flow of a process according to the second embodiment.

FIG. 19 is a diagram showing an example of a memory map of a storage medium storing a control program according to the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Image data supply part 20 Band division part 30 Edge detection part 40 Symmetry axis detection part 50 Matching part 60 Output terminal

Claims (13)

[Claims] 1. An image processing method for automatically detecting a predetermined image from image data, comprising: a dividing step of dividing the image data into bands; and performing the predetermined processing based on the image data band-divided by the dividing step. An image processing method comprising: a limiting step of limiting an area where an image is to be searched; and a detecting step of detecting a position of the predetermined image in the area limited by the limiting step. 2. The image processing method according to claim 1, wherein the predetermined image is a human face. 3. The method according to claim 1, wherein the limiting step includes:
3. The image processing method according to claim 2, wherein a region where edges are distributed in a T-shape is selectively extracted. 4. The limiting step includes extracting an edge portion of an image based on band-divided image data, dividing the image into a plurality of blocks each including a predetermined number of pixels, and extracting the edge portion of the extracted edge portion. 3. The image processing method according to claim 2, wherein a block including a predetermined number of pixels or more is determined as an area in which the predetermined image is to be searched. 5. The method according to claim 1, wherein the limiting step extracts a portion in which a block including a predetermined number or more of pixels of the edge portion is connected in a substantially T-shape, and uses the portion as a region in which the predetermined image is searched. The image processing method according to claim 4. 6. The image processing method according to claim 2, wherein the limiting step limits an area in which the predetermined image is to be searched based on a degree of left-right symmetry in each part of the image. 7. The image processing apparatus according to claim 6, wherein the degree of the target is detected using at least one of an image data value, a gradient value of the image data, and a direction of the gradient of the image data. Method. 8. The method according to claim 1, wherein the limiting step includes:
A region where edges are distributed in a T-shape is selectively extracted, a degree of left-right symmetry is detected for each part of the extracted region, and a region where the predetermined image is to be searched is determined based on the detected degree. 3. The image processing method according to claim 2, wherein the image processing is limited. 9. The image processing method according to claim 2, wherein the division unit divides the image data into a predetermined spatial frequency band. 10. The image processing device according to claim 1, wherein the dividing unit divides the image data into
The image data is divided into first image data including a high frequency component in a horizontal direction and a low frequency component in a vertical direction, and second image data including a high frequency component in a vertical direction and a low frequency component in a horizontal direction. 9. The image processing method according to 8. 11. The image processing method according to claim 8, wherein the dividing unit further performs a thinning process on the divided image data to reduce an image size of the divided image data. 12. An image processing apparatus for automatically detecting a predetermined image from image data, comprising: a dividing unit that divides the image data into bands; An image processing apparatus comprising: a limiting unit that limits an area in which an image is to be searched; and a detecting unit that detects a position of the predetermined image in an area limited by the limiting unit. 13. A computer-readable memory storing a control program for automatically detecting a predetermined image from image data, wherein the control program causes a computer to divide the image data into bands, Means for limiting an area to be searched for the predetermined image based on the image data band-divided by the means, and functioning as detection means for detecting the position of the predetermined image in the area limited by the limitation means A computer readable memory characterized by the following.