KR20090027550A - System and method for detecting object of image and computer readable medium stored thereon computer executable instruction recorded with time-series data structure - Google Patents

Abstract

An apparatus and a method detecting an image object and a recording medium thereof are provided to detect an object in an image by performing labeling technique effectively with small memory and simple calculation. An image object detecting apparatus comprises the followings: a detecting unit(210) detecting a value of a surrounding pixel based on the labeled pixel; a labeling unit(220) determining existence of a value of a binary image on the value of the surrounding pixel and converting the value of label into a new label value or the surrounding pixel; and a detection object determining unit(260) determining a detection object by using a statistical calculation value in which the label value is given.

Description

FIELD AND METHOD FOR DETECTING OBJECT OF IMAGE AND COMPUTER READABLE MEDIUM STORED THEREON COMPUTER EXECUTABLE INSTRUCTION RECORDED WITH TIME-SERIES DATA STRUCTURE}

The present invention relates to an object detecting apparatus and method in an image for detecting each target of a subject included in an image, and more particularly, to an object detecting apparatus and method using labeling using correlation with surrounding pixels.

Among various application techniques of image signal processing, labeling refers to an image processing technique used to distinguish objects (subjects) spaced apart from each other in an image.

Such a labeling technique is used to assign a unique number to each object through a labeling process when there are a plurality of objects in the image, and to process only a specific object of interest afterwards, and then, uniquely assigned to the specific object. It refers to a technique used by using a number.

Such a labeling technique may be applied to various applications in which image processing is used, and in particular, a face recognition system used for access control and the like may be used. That is, it may be used when a task of detecting and separating a face region from an image acquired from a predetermined optical device is to be performed as a preliminary step of face recognition for access authentication.

The labeling technique and the specific object detection technique in the image using the same as the above process for detecting and separating the face region in the image, together with the iris recognition, the recognition of a specific body part of the human body or in the automation line of the manufacturing industry It can be widely used in the process of identifying defective products.

Conventional techniques presented as labeling algorithms can be classified into recursive algorithms and algorithms using loops. The above method employs excessive self-calling or stack structure. This corresponds to a method of storing all information or line information in a memory and using the stored information.

This method may be applicable even if it takes some time when there are enough hardware resources, but it is determined in terms of hardware resources and processing time required for portable devices such as portable terminals with the purpose of portability, simplicity and mobility. There is a problem that it is quite difficult to apply.

The above problems are also included in the face detection technique that presupposes the labeling technique and application applications such as AF (Auto Focus), AWB (Auto White Balance), and AE (Auto Exposure) that are used in digital cameras. It can be said.

Recently, when the interactive video call service is launched, a mobile communication terminal including a function of detecting a face for the purpose of entertainment has been released, and this trend is being expanded as a mobile terminal for the Human Communication Interact (HCI) technology. It is evaluated to show trends.

According to this trend, target recognition techniques are applied to mobile communication terminals, etc., but terminals (mobile phones, digital cameras, optical devices, authentication systems, etc.) employing the above techniques process input images through frame memory or line memory. This technique requires a memory resource corresponding to this.

In other words, hardware terminals are disadvantageous in terms of economical efficiency as required hardware resources, and as described above, memory terminals have a problem of lengthening processing time such as data storage, reading, and operation since they are based on the utilization of stored and stored information. can do.

Disclosure of Invention The present invention has been made to overcome the above-mentioned problems, and an object thereof is to provide an efficient apparatus and system for performing an effective labeling technique even with simple calculation and use of a small memory, and detecting an object in an image based on this. have.

Another object of the present invention is to provide a recording medium on which a computer-readable program is executed so as to implement the method in a computer.

Other objects and advantages of the present invention will be described below and will be appreciated by the embodiments of the present invention. In addition, the objects and advantages of the present invention can be realized by the configuration and combination of configurations shown in the claims.

An image object detecting apparatus for achieving the above object comprises: a detector configured to detect a value for a surrounding pixel belonging to a transformation matrix region for labeling transformation based on a target pixel, which is a target of labeling transformation, among pixels of a binarization image; A labeling unit converting a labeling value of the target pixel into a new labeling value or a labeling value of the surrounding pixel according to whether a value other than the value representing the binarized image exists in the value; And a detection object determination unit which determines a detection object using a statistical calculation value of each pixel to which the label value is assigned.

On the other hand, the image object detection method according to another aspect of the present invention, the detection step of detecting a value for the surrounding pixels belonging to the transformation matrix region for labeling conversion on the basis of the target pixel of the labeling transformation of the pixels of the binarization image ; A labeling step of converting a labeling value of the target pixel into a new labeling value or a labeling value of the surrounding pixel according to whether a value other than a value representing the binarized image exists in the surrounding pixel value; And a detection object determination step of determining a detection object using a statistical calculation value of each pixel to which the label value is assigned.

In the present invention, the auto focusing function, the auto white balance function, and the auto exposure function used in a digital optical device such as a digital camera are utilized based on the face area.

It is possible to implement a more effective and small resource optical device based on the labeling technique and the object detection technique using the same according to the present invention having various functional applications.

As described above, the labeling and object detecting apparatus of the present invention can be easily implemented in hardware at the same time to increase the efficiency of each function, and when mounted on a mobile terminal in the future, it creates an effect that can implement a high utilization equipment based on high economic efficiency do.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, terms or words used in the specification and claims should not be construed as having a conventional or dictionary meaning, and the inventors should properly explain the concept of terms in order to best explain their own invention. Based on the principle that can be defined, it should be interpreted as meaning and concept corresponding to the technical idea of the present invention.

Therefore, the embodiments described in the specification and the drawings shown in the drawings are only the most preferred embodiment of the present invention and do not represent all of the technical idea of the present invention, various modifications that can be replaced at the time of the present application It should be understood that there may be equivalents and variations.

Prior to the detailed description of the present invention will be described for terms and definitions necessary for the description of the present invention.

In general, the color space of an image that is the basis of image processing is expressed in various spaces such as RGB, CMYK, HS, CIE, and Y depending on the viewpoint of color mixing, such as the viewpoint of color mixing and similarity with human visual system. It is apparent to those skilled in the art that the conversion of a color space expressed in a heterogeneous form into another color space can be converted by a simple mathematical conversion equation.

In addition, the input image is expressed as a sum of a plurality of pixels, and the pixels have image information (brightness, color, saturation, etc.) corresponding to the pixel. In general, the image information is divided step by step from 0 to 255 and is represented by 8-bit information, but the video information may be modified in various ways, such as 10 and 12 bits, depending on the applied environment.

Therefore, in describing the present invention, one color space coordinate system represented as an embodiment may be applied to the same and similar color coordinates in the above viewpoints, and the image information of one pixel in the input image may be It should be understood that the bit size is only one embodiment.

1 is a block diagram showing the configuration of an image labeling apparatus according to a preferred embodiment of the present invention, and a flowchart showing the process of the image labeling method according to the preferred embodiment of the present invention. Explain through 3.

The labeling apparatus 100 according to the present invention includes a block converting unit 130, a detecting unit 110, a labeling unit 120, and a label information calculating unit 140.

The detection unit 110 performs labeling based on a target pixel, which is a target of labeling conversion, among pixels of a binarized image having a first value corresponding to a value smaller than a threshold value and a second value corresponding to a value larger than a threshold value. A value of surrounding pixels belonging to a transform matrix region for transform is detected (S320).

The binarized image refers to an image obtained by dividing a value of a corresponding pixel based on a predetermined threshold value that is a standard of binarization. For example, when the threshold value is 50, a pixel having a value smaller than 50 is 0. A pixel with a value greater than 50 means an image represented by one.

The binarized image is generally expressed by using bits of 0 and 1, but in the present invention, the binarized image is represented by 0 and 1 as well as various symbolic expressions that can be divided and expressed based on a reference threshold. Of course, it is also possible to express, and in describing the present invention, the first value is 0 and the second value is 1, which is only one example. have.

In the image processing, the transformation matrix refers to a matrix having a constant rectangular shape, and as illustrated in FIG. 6, which is a diagram illustrating a labeling filter for the transformation matrix of the preferred embodiment of the present invention, the transformation matrix refers to a matrix.

Although the transformation matrix shown in FIG. 6 illustrates a 3 × 3 matrix, various embodiments are possible depending on the form used and the degree of labeling desired by the user. In the case of having a distance between the pixels, the pixels may be distinguished even though they are spaced one by one.

Among the nine pixels illustrated in FIG. 6, a pixel denoted by b corresponds to a peripheral pixel of the present invention, and a pixel denoted by a becomes a target pixel to be subjected to labeling conversion, and the peripheral pixel existing around the target pixel is The detection unit 110 detects a value that has a value.

When the value of the surrounding pixel b is detected by the detector 110, the pixel value of the detected surrounding pixel is transmitted to the labeling unit 120 (S330), and the labeling unit 120 performs the object. If the value of the pixel is a second value and the value of the detected surrounding pixel includes only the first value or the second value, the value of the target pixel is converted into a label value distinguished from an existing label value, and the detected ambient If there is a label value other than the first value and the second value among the pixel values, the target pixel value is converted into a label value other than the first value and the second value (S340).

In step S340, the labeling value of the target pixel is converted into a new label value or a label value of the surrounding pixel according to whether or not a value other than the value representing the binary image is present in the surrounding pixel value. In this way, a more effective and simple labeling method can be realized.

That is, when the value of the target pixel is the second value 1, it corresponds to a pixel value larger than the predetermined threshold value, so it may be determined that the object pixel corresponds to an object existing in the image. Therefore, a labeling operation is performed when the value of the target pixel is the second value.

If the pixel value of the surrounding pixel is composed of only the first value and the second value for representing the binarized image, that is, if there is no value other than 0 and 1, the new pixel value is assigned to the target pixel. Naturally, the label value to be given should be a label value distinguished from a label value previously given.

Also, if the surrounding pixel is given a labeling value other than 0 and 1, that is, if the labeling value has been assigned to the pixel closest to the target pixel, it is adjacent to the pixel to which the labeling value was assigned in the previous operation. Since it corresponds to an object, a label value other than the first value and the second value is assigned to the target pixel.

This process can be calculated by reading the information of the input image sequentially according to the input order, so it is not necessary to store the frame or the line like the recursive method or the repetitive method used in the general labeling technique. Labeling can be given more easily.

Hereinafter, a blocker for blocking a target image for labeling will be described so as to improve the operation speed and lower the memory charge while maintaining the labeling operation efficiency.

The block converting unit 130 according to the present invention divides the labeling target image into a predetermined number of blocks according to the block rate, and the number of pixels corresponding to the threshold range among the pixels corresponding to each of the divided block areas is larger than the reference number. In this case, the second image is assigned to the block, and the first value is assigned to the block, and the target image is converted into a block image having the value assigned to the block as the pixel value (S300).

When the block image is converted into the block image as described above, the detection unit 110 described above receives the block image converted by the block converter 130 as the binarized image (S310).

A configuration of the block transformation, detection of a blocked image, and labeling through a specific example will be described below.

For convenience of explanation, assuming that the size of the labeling image is 100 × 100, and the target image having the 100 × 100 size is converted into a 10 × 10 image, the block rate is 10%, and the predetermined number The block is 100 blocks. It is apparent that the block rate and the predetermined number based on the block rate are determined by various application environments and various user environments based on hardware resources.

That is, as shown in FIG. 5, the size a of the column of the block image is 10, and the size b of the row is 10, and the value representing the binarized image is different from the previous example. 15, the first value was set to zero.

In the case of blocking by the blocking ratio as described above, one block represents 100 pixels of the original image, and each 100 blocks represents 100 pixels, respectively, representing 100,000 pixels of the original target image.

In the blocking conversion process, it is preferable to use one of the image information of each pixel as a threshold value of the conversion. That is, as described above, the pixel may have information expressed in various color spaces such as color information and luminance information. On the premise of detecting a human face, among the various image information, in particular, Cb, Cr It is desirable to base the information on.

That is, the number of pixels corresponding to the threshold range that is the range where the Cb and Cr values represent the color of the human body is counted based on the threshold value determined among the image information of the pixel. Since the values of Cb and Cr do not vary greatly among whites, blacks, and yellow races, there is an advantage that they can be used universally for various races. Therefore, a predetermined threshold range for Cb and Cr values is set, and the corresponding pixel is If the number of pixels corresponding to the threshold range among the pixels in charge of one block through the process of determining whether it corresponds to the threshold range is large through comparison with the reference number set by a predetermined reference, the second value 15 In the case of a small value), it can be operated in a manner of giving a first value (0).

The reference number is also understood as a reference value that can be changed by various application examples such as various image environments and object characteristics of the image.

In the manner as described above, the labeling target image is converted into a block image blocked by 10 × 10 by having a block having a second value 15 and a first value 0 as the corresponding pixel value.

The block image converted by the above process is shown in Table 1 below.

In Table 1, the second value is illustrated as 15, and the first value is illustrated as 0. This means that it should be interpreted as one of various application examples that can express the first value and the second value of the binary image as described above.

It is most preferable to set the block rate to 32 × 24 for the size of the block image in order to distinguish visual recognition and labeling of the target image. In the case of the block image set to the size, about 10 objects can be distinguished when judged based on the characteristics of the normal image, which means that the number of labels assigned to each object is required.

Therefore, the labeling unit 120 is preferably configured to be assigned the label value using a 4-bit bit string as described above. In the case of 4 bits represented by a binary number, since 15 bits may have 15 values, it may be said that the number of label values assigned to distinguish the object is most preferable.

The blocking rate is preferably set to about 3% to 10% of the target image for the efficiency of the arithmetic processing module and memory while maintaining the performance of labeling and object detection based on the same.

As shown in FIG. 5, the reason for assigning the binarization value to the first value to 0 and the second value to 15 is that when the new labeling is applied, an algorithm is applied to give a value between 0 and 15. This is to prevent a bit from being generated and to skip an operation of determining whether a given labeling value is duplicated.

When the pixel located at (2, 2) in Table 1 is the target pixel, the target pixel corresponds to the second value, and the first value and the second value among the pixel values of the surrounding pixels belonging to the 3x3 transformation matrix region. Since there is no value other than the value, a labeling value is assigned to the target pixel. Assuming that the pixel of (2,2) is the beginning of a labeling operation, a new labeling 1 is exemplarily described.

In the next step, as shown in Table 2, the transformation matrix is shifted 1 to the right so that the pixel located at (2,3) becomes the target pixel, and the pixel value of (2,2) is a value of "1" in the labeling process. Since the surrounding pixels are enumerated in order from the top left to the bottom right, there are eight pixels as 0,0,0,1,15,15,15,15.

Since there is a label value of "1" other than the first value and the second value (0,15) among the eight pixels in the conversion matrix area, the (2,3) target pixel is converted into a "1" label value. do. That is, the label value given to the nearest pixel is extended to the label value of the adjacent pixel. Applying the same steps as above, the result of the labeling operation for the second row is shown in Table 3 below.

As shown in Table 3, the (2,8) pixel is a conversion target because the target pixel is 15 (second value), and in the case of the (2,6) and (2,7) pixels, the value of the target pixel is 0. Because labeling is not converted. In the case of the (2,8) pixel, there is no value other than 0 and 15 in the surrounding eight pixels, and since the value of the label "1" has already been given, a new label value "2" is given.

The result of the labeling image labeled with the 10 × 10 block image by the above method is as follows.

As described above, all pixels of the target image are calculated, and when the pixel on the edge is the target pixel, a limited conversion matrix area can be utilized, and the user is converted to a default value and operated. Various modifications are possible accordingly.

As described in detail above, after the labeling operation for identifying each object in the target image is completed, the characteristics of the region labeled with the same labeling are identified.

The characteristics of the labeling area are in charge of the label information calculating unit 140 of the present invention, and receives the label image given the label value from the labeling unit 120 (S350). Label information including at least one of the information and the area information is calculated for each label area (S360).

As shown in Table 4, when the label information calculating unit 140 receives the label image, the number of regions to which the same label value is assigned is five, and the label is image characteristic information for the five label regions. Will yield information.

In the case of the label area (ROI area Range Of Interest) to which the label value "1" has been assigned, the start point is (2,2) and the end point is (4,5). The information about the end point may be determined based only on the area to which the label value is actually assigned, or the area using the maximum outline to which the same label value is assigned may be used. The end point becoming (4, 5) is an example in which the maximum outline area is utilized.

If the area information is determined based on the area to which the actual label value is assigned, the area information is 10, and the area information is 12 when it is based on the outermost outline. By the above method, information on characteristics of each ROI is calculated, and this information is used as reference data used in various applications in the future.

In addition, the labeling calculation unit 140 of the present invention may include an integrating unit 142. The integrating unit 142 integrates the overlapping label areas when the label areas corresponding to the label information overlap each other. And calculate label information of the integrated label area.

If a label image as shown in Table 5 is generated, two regions, ROI-1 and ROI-2, to which different label values "1" and "2" are assigned, overlap each other. In this case, even if the label value is not the same, the object represented in the image is preferably configured to be included in the same division area.

That is, it is preferable to integrate ROI-1, which is the label region designated as "1", and ROI-2, which is the label region designated as "2", and newly calculate label information of the integrated label region. When the integrated ROI is called ROI-M, the start point of the label information of the ROI-M is (2,2) and the end point is (6,8). The area information is 22 when the area information is calculated in the application example based only on the area to which the actual label value is given except for the pixel represented by 0.

The newly merged label area ROI-M may be labeled with a "1" or "2" label value, and another new label value may be assigned. This is because the 1 and 2 label values are not used in other label areas, so they can be assigned to the integrated label area. It is of course possible to assign a new label value (e.g., a value of 10 or more) as a value that is not used as another label value so that it is represented by a symbol that promises to be an integrated area.

Referring to FIG. 7, which illustrates a merge process according to an embodiment of the present invention, the image a of FIG. 7 denotes image information before the merge process, and b denotes image information after the merge process. it means.

Hereinafter, an apparatus and method for detecting an image object according to another object of the present invention will be described in detail with reference to FIGS. 2 and 4.

The object detecting apparatus and method are application examples employing the above-described labeling technique and, as described above, relate to a technique for detecting a specific object in an image such as a face region, a specific region of a body, a defective article, etc. of a human body.

In the description of the object detecting apparatus and method, a description overlapping with the above-described labeling apparatus and method is omitted, and reference numerals for indicating the components to be referred to are only given for convenience of explanation, so reference is made. Even if the signs are not the same, the functions they are responsible for are the same and similar components mean the same and similar configurations.

Referring to FIG. 2, which is a block diagram showing the configuration of an image object detecting apparatus according to an exemplary embodiment of the present invention, and FIG. 4, which is a flowchart illustrating a process of the method of detecting an image object according to an exemplary embodiment of the present invention, The image object detecting apparatus 200 according to the present invention includes a block converting unit 230, a detecting unit 210, a labeling unit 220, a label information calculating unit 240, a filtering unit 250, and a detection target determining unit 260. ).

The block converter 230, the detector 210, the labeling unit 220, and the label information calculating unit 240 have been described in the above-described image labeling apparatus and method, and thus detailed descriptions thereof will be omitted. .

The detection object determination unit 260 of the image object detecting apparatus 200 of the present invention receives information on the label value to which the label value is assigned from the labeling unit 220, and each pixel to which the label value is assigned. The detection target is determined using the statistical calculated value of.

The statistical calculation value of each pixel to which the label value is assigned is given a label value corresponding to the object to be detected by using various information such as average, variance, distribution, standard deviation, or information on the number of pixels to which the label value is assigned. Detected pixels are detected.

 In order to increase the precision or efficiency of the detection, the image object detecting apparatus 200 includes one or more of start point information, end point information, and area information of a label area to which the same label value is assigned in the label image to which the label value is assigned. Preferably, the label information calculating unit 240 calculates label information for each label area.

When the label information calculation unit 240 is additionally configured as described above, the detection object determination unit 260 receives label information from the label information calculation unit 240 and corresponds to reference information for determining the detection object. According to the classification of the label information for each label area and using the classified label information includes a determination unit 262 for determining the detection target region to determine the detection target.

That is, the detection target region is detected according to the degree of matching with the reference information by receiving various information of each label region to which the same label value is given and using the reference information for detecting the desired detection target.

Specific elements of the reference information will be described later. Each of the label areas is classified according to the degree of conformity with the reference information, and the classified top label areas are determined as the detection targets or the label areas of the upper predetermined order are detected as the detection target areas. It can also be screened.

As described above, when the block converting unit 230 is additionally configured in the image object detecting apparatus 200, the block converting unit 230 converts the target image into the block image (S400), and the detecting unit 210. In operation S420, the block image converted by the block converter 230 is input as the binarized image, and the surrounding pixel value is detected based on the target pixel of the transmitted binarized image.

In addition, the label information calculating unit 240 of the image object detecting apparatus 200 integrates the overlapped label regions and calculates the label information of the integrated label regions when the label regions corresponding to the label information overlap each other. An integration unit 242 may be included.

That is, the labeling unit 220 receives the pixel value detected from the detection unit 210 in operation S430, and performs a labeling process using the detected surrounding pixel value and the target pixel value in operation S440.

When the labeling process is completed, the labeling unit 220 transmits a label image to the labeling calculator 240 (S450), and includes an integrated unit 242 included in the labeling calculator 240 and the labeling calculator 240. ) Calculates label information of the label region of the label image or the integrated label region (S460).

In addition, the filtering unit 250 is a block corresponding to NR (Noise Reduction), and receives the label information from the labeling calculation unit 240 (S470), in the label area that is outside the filtering reference range of the transmitted label information The corresponding label information is removed (S480).

As illustrated in FIG. 8, the filtering unit 250 performs a function of excluding a region that can be excluded from a candidate region in a desired target detection using a reference range, for example, the number of pixels.

FIG. 8 illustrates a process in which five label regions shown in (a) are filtered into one label region as shown in (b) by a technique of removing a case where the number of pixels is smaller than a reference range.

As an example, the detection of the face area is supplementary. As described above, the pixels having the image information corresponding to the critical range of the face area, as well as the pixels corresponding to the actual face area, various backgrounds and subjects represented in the image Some of them may not exclude the possibility of including the pixel corresponding to the threshold range.

As such, in order to exclude an object having similar image information in advance even though it is not an actual face area, when the label area is out of the reference range, the label information corresponding to the label area is removed.

The reference range for filtering may be determined by using the number of label regions as well as the information of various label regions, and the characteristics of objects represented in various experiments or images to match the characteristics of the object to be detected. It is preferable to configure so that it is objectively calculated.

For example, when a face is typically detected, an image is acquired close to an optical machine, and thus, the reference range may be set to a range of a size occupied by the acquired image, and the acquired image is located at a centered area. Since the location may be generally located, the label area more than a predetermined size may be determined based on the distance from the center of the label, and the label area below the specific size may be excluded by absolutely comparing the size of the label area. Of course, it may be decided.

After the label information outside the reference range is removed by the filtering unit 250 as described above (S480), only the undeleted label information is transmitted to the detection target determination unit 260, and the detection target determination unit 260, in particular, The determination unit 262 determines (S495) a detection subject region by using label information other than the removed label information.

The reference information used by the detection subject determination unit 260 including the determination unit 262 to detect the detection target region includes information on the size of the label region compared to the overall image size, The at least one selected from information on a corresponding position, aspect ratio information of the label area, and information on the number of second values existing in the label area may be set.

As described above, even when the filtering step is completed, it may not be possible to completely exclude the possibility that large areas in which the target detection area and the image information (color, etc.) are in a similar range remain.

Therefore, the determination unit 262 performs a process of selecting the closest detection target among the label areas ROI corresponding to the candidate group of the detection target.

The information on the size of the label region compared to the entire image size, which is one of the reference information, is more likely to have a proper size than the entire image, rather than to occupy a very small or large area of the target image. That is, a small area is likely to be another object outside the subject, and in an extremely large area, it is likely to be an object corresponding to a background.

Since the location information is more likely than the case where the face is placed at the center of the image when the face is photographed with an optical device, etc., the distance is separated from the central reference point by a predetermined range. It can be set as a standard for determining the detection target.

In addition, the aspect ratio information of the label area reflects the physical characteristics of the object to be detected and corresponds to the reference information determined by reflecting the actual shape characteristics of the object to be detected. For example, when the face of the human body is a detection target, the ratio of the aspect ratio is 0.7 or 0.8, which corresponds to the area of the human face more than the unique ratio of about 0.1 or 8, so that the reference information is used as the above-mentioned reference information. The face area is detected.

In addition, the number information of the second value existing in the label area can use the number information of the second value compared to the size of the label area. If the ratio of the area occupied by the elliptical shape to the rectangular shape of the square is used, that is, the number information of the second value is used, the precision of face area detection can be further improved.

The reference information as described above may be set to various information according to the physical characteristics of the detection target and the image characteristics reflecting the same, and may be configured to use one or more pieces of reference information by the selected combination of the reference information.

A score (score) is assigned to each label region based on the reference information as described above, and a label region having the highest score or a label region having the highest predetermined order can be detected as a desired detection subject region.

Each component of the image labeling apparatus and the image object detecting apparatus according to the present invention should be understood as a logical component rather than a physically divided component. That is, since each configuration corresponds to logically divided components to realize the technical idea of the present invention, even if each component is integrated or separated, the present invention can realize the function performed by the logical configuration of the present invention. Should be interpreted to be within the scope of.

The image labeling method or the image object detecting method according to the present invention described above may be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include.

The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. In addition, functional programs, codes, and code segments for implementing the image labeling method or the image object detection method may be easily inferred by programmers in the art to which the present invention belongs.

Although the present invention has been described above by means of limited embodiments and drawings, the present invention is not limited thereto and will be described below by the person skilled in the art to which the present invention pertains. Of course, various modifications and variations are possible within the scope of the claims.

The following drawings, which are attached to this specification, illustrate exemplary embodiments of the present invention, and together with the detailed description of the present invention serve to further understand the technical spirit of the present invention, the present invention includes matters described in such drawings. It should not be construed as limited to.

1 is a block diagram showing the configuration of an image labeling apparatus according to a preferred embodiment of the present invention;

2 is a block diagram showing the configuration of an image object detecting apparatus according to an embodiment of the present invention;

3 is a flowchart illustrating a process of an image labeling method according to an embodiment of the present invention;

4 is a flowchart illustrating a process of a method for detecting an image object according to an exemplary embodiment of the present invention;

5 is a diagram for a blocked block image according to one embodiment of the present invention;

6 illustrates a labeling filter for a transformation matrix of a preferred embodiment of the present invention;

7 illustrates a merge process according to an embodiment of the present invention;

8 is a diagram illustrating a filtering process according to an embodiment of the present invention.

Claims (15)

A detector configured to detect values of surrounding pixels belonging to a transformation matrix region for labeling transformation based on a target pixel, which is a target of labeling transformation, among pixels of the binarization image; A labeling unit converting a labeling value of the target pixel into a new labeling value or a labeling value of the surrounding pixel according to whether a value other than a value representing the binarized image exists in the surrounding pixel value; And And a detection object determination unit which determines a detection object by using a statistical calculation value of each pixel to which the label value is assigned. The method of claim 1, wherein the binarization image, And a second value corresponding to a value larger than the threshold value and a first value corresponding to a value smaller than the threshold value. The method of claim 2, wherein the labeling unit, When the value of the target pixel is a second value and the value of the detected surrounding pixel includes only the first value or the second value, the value of the target pixel is converted into a label value distinguished from an existing label value, and the detection is performed. And a label value other than the first value and the second value among the neighboring pixel values, wherein the target pixel value is converted into a label value other than the first value and the second value. . The method of claim 1, And a label information calculating unit configured to calculate label information for each label region, the label information including at least one of start point information, end point information, and area information of the label region to which the same label value is assigned in the label image. The detection subject determination unit, And a labeling unit for classifying label information for each label region according to a degree corresponding to the reference information for determining a detection target, and determining a detection target region by using the classified label information. . The method of claim 4, wherein Further comprising a filtering unit for removing the label information corresponding to the label area outside the filtering criteria range, The determination unit, And label information other than the removed label information. The method of claim 4, wherein the reference information for determining the detection target, At least one selected from information on the size of the label area compared to the entire image size, information on the position of the label area corresponding to the entire image, information on the aspect ratio of the label area, and information on the number of second values present in the label area. An image object detection apparatus, characterized in that determined using information. The method of claim 1, wherein the detection target, Image object detection apparatus, characterized in that the face area of the human body. A detection step of detecting a value of a surrounding pixel belonging to a transformation matrix region for labeling transformation based on a target pixel, which is a target of labeling transformation, among pixels of the binarization image; A labeling step of converting a labeling value of the target pixel into a new labeling value or a labeling value of the surrounding pixel according to whether a value other than a value representing the binarized image exists in the surrounding pixel value; And And a detection object determination step of determining a detection object using a statistically calculated value of each pixel to which the label value is assigned. The method of claim 8, wherein the binarization image, And a second value corresponding to a value larger than the threshold and a first value corresponding to a value smaller than the threshold. The method of claim 9, wherein the labeling step, When the value of the target pixel is a second value and the value of the detected surrounding pixel includes only the first value or the second value, the value of the target pixel is converted into a label value distinguished from an existing label value, and the detection is performed. And a label value other than a first value and a second value among the neighboring pixel values, wherein the value of the target pixel is converted into a label value other than the first value and the second value. . The method of claim 8, Comprising a label information calculation step of calculating for each label region the label information including at least one of the start point information, the end point information and the area information of the label region is given the same label value in the label image is given, The detecting target determination step, And classifying label information for each label region according to a degree of conformity with reference information for determining a detection target, and determining a detection target region by using the classified label information. Way. The method of claim 11, The method may further include a filtering step of removing label information corresponding to a label area outside the filtering criteria range. The determining step, And label information other than the removed label information. The method of claim 11, wherein the reference information for determining the detection target, At least one selected from information on the size of the label area compared to the size of the entire image, information on the position of the label area in the entire image, aspect ratio information of the label area, and information on the number of second values present in the label area. Image object detection method characterized in that determined using the information. The method of claim 8, wherein the detection target, Image object detection method characterized in that the face area of the human body. A non-transitory computer-readable recording medium having programmed thereon a method according to any one of claims 8 to 14.

Images