JP2010015294A - Object domain extracting apparatus, object domain extracting method, object domain extracting program, and recording medium which record the same program - Google Patents

Abstract

An object region is extracted from an image taken under a background that does not change with time.
An imaging apparatus C1. Image data obtained by imaging a finger at C2 is input to the target area extraction apparatus. The imaging device C1 images a finger under a background that does not include color and does not change over time. The target area extracting apparatus acquires representative individual skin color information from the captured image of the imaging apparatus (S01). Further, the target region extraction device converts the captured image of the imaging device C2 into an image including a skin color region and an undetermined region based on the representative individual skin color information (S02). An area expanded image obtained by performing area expansion processing on an undetermined area in the converted image is acquired (S03). A labeling process is performed on the skin color area of the area expanded image to extract a finger area (S04).
[Selection] Figure 5

Description

  The present invention relates to a technique for recognizing an arbitrary object shape having a specific single color from an image, for example, a technique for recognizing a hand movement from an image taken by an imaging device.

  Currently, various methods and apparatuses have been proposed for man-machine interface between a person and a computer. For example, instead of operating an AV device or personal computer using a remote control, a mouse, a keyboard, or the like, based on motion information obtained by recognizing the movement of a finger from an image such as a gesture with a bare hand imaged by a camera. The concept of operation has been proposed. According to this, since the operation can be performed without touching the device, it is suitable for an interface in the field of medical and food from the viewpoint of hygiene.

As a technique for recognizing a finger in an image necessary for realizing this concept, for example, a “hand shape recognition method” of Patent Document 1 is known. In this method, a necessary finger region is extracted based on a difference between a background image acquired by capturing in advance and an image acquired by capturing a finger.
JP 7-105371 A

  However, the simple application of the background difference method as in Patent Document 1 cannot handle the case where the background changes in principle, and when there is a moving object other than a finger, the interframe difference method is used. However, it may not be possible in principle.

  The present invention has been made in view of such a conventional problem, and an object of the present invention is to extract a detection target region from an image captured under a case where the background changes.

  The present invention is a technical idea created to solve the above-described problem, and the invention according to claim 1 is an object for extracting a detection target region from an image obtained by capturing the detection target with a plurality of imaging devices. In the region extraction device, based on the acquired feature, the detection target, the means for acquiring the feature of the detection target from an image captured under a background that does not include color and does not change over time, Means for extracting a region of the detection target from another image obtained by imaging the detection target.

  The invention according to claim 2 is characterized in that the feature of the detection target is an individual color obtained by collecting representative hues determined for each luminance when the color of the detection target region included in the image is decomposed into luminance and hue. It is characterized by information.

  According to a third aspect of the present invention, there is provided means for converting the other image into an image including a color region and an undetermined region based on the individual color information, and a hue of the undetermined region adjacent to the color region in the converted image. Means for acquiring an extended color area image including an undetermined area in a color area when approximating the hue of the individual color information, and performing a labeling process on the color area in the extended color area image to detect the detection target Means for extracting the region.

  The invention according to claim 4 is a target region extraction method in which an image processing unit extracts a region to be detected from an image obtained by capturing a detection target with a plurality of imaging devices, and the image processing unit includes the detection target. A first step of acquiring a feature of the detection target from an image captured under a background that does not include a hue and does not change over time, and the image processing unit detects the detection target based on the acquired feature. And a second step of extracting the detection target region from another captured image.

  According to a fifth aspect of the present invention, in the first step, individual color information obtained by collecting representative colors determined for each luminance when the color of the detection target area included in the image is decomposed into luminance and hue is collected. It is characterized by acquisition.

  According to a sixth aspect of the present invention, the second step is a step of converting the other image into an image including a color region and an undetermined region based on the individual color information, and adjacent to the color region in the converted image. Obtaining an extended color area image including the undetermined area in the color area when the hue of the undetermined area approximates the hue of the individual color information; and labeling the color area in the extended color area image And extracting the region to be detected.

  A seventh aspect of the present invention relates to a target region extraction program that causes a computer to function as each unit of the target region extraction device according to any one of the first to third aspects.

  The invention described in claim 8 relates to a computer-readable recording medium in which the target area extracting program described in claim 7 is recorded.

  According to the first to eighth aspects of the present invention, it is possible to extract a target area from an image even under a background that does not change over time, and it is possible to handle even if the detection target is a moving object other than a finger. it can.

  FIG. 1 shows a target area extraction apparatus according to an embodiment of the present invention. A group of image data picked up by the two photographing devices C1 and C2 is input to the target area extracting device 1 in time series.

  The target area extracting apparatus 1 is composed of a computer, and the two photographing apparatuses C1. Data communication means 2 for inputting image data from C2 via the network, image processing means 4 for processing the inputted image data, storage means 3 such as a memory (RAM) for temporarily storing the processing data, and the processing data Storage means 5 such as a hard disk drive device for storing data, display means 6 such as a monitor capable of displaying processing data, and input means 7 such as a keyboard and a mouse.

  Specifically, as a result of cooperation between a processor (CPU: Central Processor Unit, etc.) and software, the image processing means 4 is a feature of a detection target from the captured image data of the imaging device C1, as shown in FIG. And a means 9 for extracting a detection target region from the captured image data of the photographing apparatus C2 based on the extracted features.

  As an example, a digital video camera is used for the photographing devices C1 and C2. Here, as shown in FIG. 3 and FIG. 4, the photographing apparatus C1 is installed on a ceiling where a table top can be imaged substantially vertically downward, that is, on the ceiling, and the table surface as an imaging background has a skin color. It can be regarded as a blue back that does not contain and does not change over time. At this time, in order to prevent the imaging background from containing a skin color component and not to change over time, a top plate that can be regarded as a blue background may be placed on the table as shown in FIG.

  On the other hand, the imaging background of the imaging device C2 does not include a skin color component and changes with time. Here, as an example, it is assumed that the imaging devices C1 and C2 are imaging an operator's finger as a detection target. Hereinafter, a target area extraction method performed by the image processing unit 4 on the input image data captured by the photographing apparatuses C1 and C2 will be described with reference to FIG.

  S01: First, representative individual skin color information (individual color information) is acquired from the input image data captured by the photographing apparatus C1 as a feature of a detection target (operator's finger).

  That is, since the photographing apparatus C1 images a finger under a background that does not include skin color and does not change over time, a finger region can be extracted from the input image data shown in FIG. 6 using a background difference method. . Here, since the pixels near the boundary of the extracted finger region may have a color mixture of the skin color component and the background color component, this extracted finger region is set as a provisional finger region shown in FIG. A region obtained by performing erosion processing on the provisional finger region is defined as a pure finger region shown in FIG.

  The frequency distribution of luminance (Y) and hue (U, V) shown in FIG. 9 is obtained for all pixels included in the obtained pure finger region data. In this embodiment, the luminance (Y) is 64 gradations from “0” to “63”, and the hue (U, V) is 256 gradations from “0” to “255”. FIG. 9 is a diagram schematically representing the frequency distribution of luminance (Y) and hue (U, V). In practice, the frequency distribution of hue (U, V) for each luminance (Y) is The shape is as shown in FIG. In FIG. 10, “F” indicates a frequency.

  Then, from the frequency distribution of luminance (Y) and hue (U, V) shown in FIG. 9, a hue (U, V) value that is the mode of the distribution is obtained for each luminance (Y), and this is shown in FIG. It is stored in the storage means 3 as representative individual skin color information to be shown. At that time, the distribution for each luminance (Y) shown in FIG. 10 may be preliminarily subjected to convolution processing using a Gaussian kernel or the like.

  The typical individual skin color information shown in FIG. 11 obtained here is “1” at the position of the black point B and “0” at other points, and the luminance (Y) and hue (U, V) are variables. Can be considered as a three-variable function. This function is a representative individual skin color function shown in Expression (1).

  S02: Next, a skin color area and an undetermined area are acquired from the input image data captured by the photographing apparatus C2 and the representative individual skin color information. Here, the converted image shown in FIG. 13 is obtained from the input image data imaged by the photographing apparatus C2 shown in FIG. 12 and the representative individual skin color function according to the equation (2).

However, “I” is the value of each pixel of the converted image data shown in FIG. 13, and is limited to “0” or “1”. Further, “I _Y ”, “I _U ”, and “I _V ” are values of luminance and hue of each pixel of the input image data imaged by the imaging device C2 shown in FIG. The converted image data of FIG. 13 may be stored in the storage unit 3 or may be stored in the storage unit 5.

  S03: The area expansion process shown in FIG. 17 is applied to the skin color area and the undetermined area included in the converted image data of FIG. That is, for the converted image data shown in FIG. 13, the area-extended image data shown in FIG.

  Here, the representative individual skin color function is a three-variable function with luminance (Y) and hue (U, V) as variables, but a two-variable function with hue (U, V) as a variable for each luminance (Y). Can be regarded as layered. This layer will be referred to as a luminance layer. Further, as shown in FIG. 16, the hue (U, V) corresponding to the position of the black point B of the representative individual skin color information is the main hue of the luminance layer, and the number of elements is one. Let it be the hue distribution of the luminance layer. Hereinafter, the algorithm of the area expansion process will be described in detail with reference to FIG.

  First, with respect to the converted image data shown in FIG. 13, it is confirmed whether or not there is an undetermined region pixel in the vicinity of 8 of any skin color region pixel (S11). If there is an undetermined area pixel, the difference (ΔU, ΔV) between the existing main color and the color of the undetermined area pixel is calculated in the luminance layer corresponding to the luminance value of the undetermined area pixel (S12).

  Next, based on the comparison between the norm of the difference calculated in S12 and a predetermined reference value, it is determined whether or not to integrate the hue distribution of the luminance layer (S13). This reference value is described as a threshold value in the program.

  At this time, if the norm of the difference calculated in S12 is smaller than a predetermined reference value, the difference is integrated, and the hue of the undetermined region pixels is added to the hue distribution of the luminance layer to obtain a new hue distribution, and this new hue distribution Is stored in the storage means 3 as a skin color region pixel (S14).

  On the other hand, if the norm of the difference calculated in S12 is larger than the predetermined reference value, the undecided area pixels are not integrated and stored in the storage means 3 as non-skin color area pixels (S15). If there is no undetermined area pixel in the vicinity of 8 of any skin color area pixel, the area expansion process is terminated.

  S04: A labeling process for adding a label attribute to each pixel included in the skin color area in the area expanded image data shown in FIG. 14 obtained by the process of S03 is performed to obtain the finger area data shown in FIG. As a result, it is possible to acquire a finger region from an image even under a background that changes with time, and in this respect, the range of an image used for finger authentication or the like is expanded.

  At this time, for the finger region selection, for example, the region having the largest area may be selected, or the region having the highest correlation with the finger silhouette templates of various postures in various scales prepared in advance is used. Also good. The finger region data acquired here may be stored in the storage unit 5 or the like. In addition, you may perform the process of S01-S04, displaying on the display means 6, such as a monitor.

  Hereinafter, another example of S01 will be described. Here, a finger region is extracted from image data of a finger captured under a background that does not include a skin color and does not change with time, regardless of the background difference method.

  First, the background image of FIG. 18 is imaged by the imaging device C1. A frequency distribution of luminance (Y) and hue (U, V) shown in FIG. 19 is obtained for the input data of the captured image. With respect to the obtained distribution frequency, the distribution frequency is “0” when the distribution frequency is “1” or more, and the distribution frequency is “1” when the distribution frequency is “0”.

  Next, an area having a distribution frequency of “0” and an area of “1” are set as a mask area and a non-mask area, respectively, and an erosion non-mask area obtained by performing an erosion process on the non-mask area for each luminance layer. To get the background mask function. This background mask function f can be obtained using equations (3) and (4).

  Then, provisional finger region image data shown in FIG. 7 is obtained from the input image data obtained by imaging the finger shown in FIG. 6 and the background mask function according to the equation (5).

Here, “I _P ” is the value of each pixel of the provisional finger region image data shown in FIG. 7, and is limited to “0” or “1”. “I _Y ”, “I _U ”, “I _V ” Are values of luminance (Y) and hue (U, V) of each pixel of the input image data obtained by imaging the finger shown in FIG. At this time, in the provisional finger region image data shown in FIG. 7, a region having a pixel value “1” is set as a provisional finger region. The pure finger region shown in FIG. 8 is extracted from the provisional finger region, the representative skin color information of FIG. 11 is obtained from this, and the processes of S02 to S04 are performed.

  In addition, this invention is not limited to the said embodiment, For example, it can also construct | assemble as a program which functions a computer as the image processing means 4 of the said target area extraction apparatus. This program may cause the computer to execute all the processes of S01 to S04, or may cause the computer to execute only the area expansion process of S11 to S15.

  This program can be provided to a computer by downloading from a website or the like. The program is stored in a recording medium such as a CD-ROM, DVD-ROM, CD-R, CD-RW, DVD-R, DVD-RW, MO, HDD, Blu-ray Disk (registered trademark). It may be provided to a computer. Since the program code itself read from the recording medium realizes the processing of the above embodiment, the recording medium also constitutes the present invention.

1 is a schematic configuration diagram of a target area extraction device according to an embodiment of the present invention. Schematic of the image processing means. The schematic of the imaging condition of the imaging device C1, C2. The schematic of the imaging background of the imaging device C1. The flowchart of the same object area | region extraction method. The figure which shows the input image from the imaging device C1. The figure which shows the temporary finger area | region extracted from the input image of FIG. The figure which shows the pure finger area | region which performed the erosion process to the temporary finger area | region of FIG. The schematic diagram of the frequency distribution of the brightness | luminance and hue of the pure finger area | region of FIG. The frequency distribution chart of the hue for each luminance. The schematic diagram of the representative individual skin color information. The figure which shows the input image from the imaging device C2. The figure which shows the same conversion image. The figure which shows the same area expansion image. The figure which shows the same finger area | region. The hue distribution map of the luminance layer corresponding to the black spot position of the representative individual skin color information. Same area expansion process flowchart The figure which shows the background image used for the other example of S1 in FIG. The same luminance and hue frequency distribution chart.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Object area extraction apparatus 2 ... Data communication means 3 ... Memory | storage means 4 ... Image processing means 5 ... Storage means 6 ... Display means 7 ... Input means 8 ... Means to extract the characteristic of an object 9 ... Extract the area | region of an object Means C1. C2 ... Shooting device

Claims (8)

A target region extraction device that extracts a detection target region from an image obtained by capturing a detection target with a plurality of photographing devices,
Means for acquiring the characteristics of the detection target from an image captured under a background that does not include hue and does not change over time;
Means for extracting an area of the detection target from another image obtained by imaging the detection target based on the acquired feature;
A target area extracting apparatus characterized by comprising: The feature of the detection target is individual color information obtained by collecting representative hues determined for each luminance when the color of the detection target region included in the image is decomposed into luminance and hue. The target area extracting apparatus according to claim 1. Means for converting the other image into an image including a color area and an undetermined area based on the individual color information;
Means for obtaining an extended color area image including an undetermined area in the color area when the hue of the undetermined area adjacent to the color area approximates the hue of the individual color information in the converted image;
Means for performing a labeling process on a color region in the extended color region image and extracting the region to be detected;
The target area extracting apparatus according to claim 2, further comprising: A target area extraction method in which an image processing means extracts a detection target area from an image obtained by imaging a detection target with a plurality of imaging devices,
A first step in which the image processing means acquires the feature of the detection target from an image obtained by capturing the detection target under a background that does not include a hue and does not change over time;
A second step in which the image processing means extracts a region of the detection target from another image obtained by imaging the detection target based on the acquired feature;
A target region extraction method characterized by comprising: The first step is to acquire individual color information obtained by collecting representative hues determined for each luminance when the color of a detection target area included in the image is decomposed into luminance and hue. Item 4. The target region extraction method according to Item 4. Converting the other image into an image including a color area and an undetermined area based on the individual color information; and
Obtaining an extended color area image including the undetermined area in the color area when the hue of the undetermined area adjacent to the color area approximates the hue of the individual color information in the converted image;
Applying a labeling process to a color region in the extended color region image to extract the region to be detected;
The target region extracting method according to claim 5, further comprising:   A target area extraction program that causes a computer to function as each means of the target area extraction device according to claim 1.   A computer-readable recording medium on which the target area extracting program according to claim 7 is recorded.

Images