KR20130036511A - Method for validation of object boundaries and system of thereof, method for detection of object - Google Patents

Abstract

PURPOSE: A method for determining validation of an object area, a device thereof, and an object extraction method thereof are provided to determine the validation of the object area based on a symmetry property, thereby efficiently determining an object with the symmetry property. CONSTITUTION: A validation determination device receives an image, which includes an object, through a photographing device(S201). The validation determination device automatically detects an object area from the image(S202). When a symmetry value for the extracted object is smaller than a predetermined first threshold value, the validation determination device determines that extraction of the object is valid and removes an area excluding the object area(S204-S206). [Reference numerals] (AA) Start; (BB,DD) No; (CC,EE) Yes; (FF) End; (S201) Obtaining an image using a photographing device; (S202) Automatically detecting an object area; (S203) Rotating an image; (S204) Calculating a symmetry value(R); (S205) R < Threshold value 1; (S206) Removing an area except for an object; (S207) R < Threshold value 2; (S208) Designating a part of the object/entire area; (S209) Requesting the re-photography of an input image;

Description

METHOD FOR VALIDATION OF OBJECT BOUNDARIES AND SYSTEM OF THEREOF, METHOD FOR DETECTION OF OBJECT}

The present invention relates to a method of extracting an object by receiving an image obtained by using a photographing device of a computer, a portable terminal, and an electronic device.

The technology of recognizing objects from images has been actively researched in various fields such as image processing, pattern recognition, computer vision and neural networks, and has numerous application fields commercially and legally. Recently, an image may be acquired by using a camera embedded in a mobile device such as a PDA or a mobile phone, and as the use of the mobile device increases, the demand for object recognition using a mobile device for the user's convenience is increasing. to be.

Object recognition in computer vision is a process of retrieving a model from a database that most closely resembles an object in an image, given a database of images and models. However, even in the same object, the size, direction, position, and pose of the object may be different in each image. Also, if a single image includes multiple objects or the object is hidden by other objects, only a part of them may be viewed. In some cases, many researches have been conducted on recognition techniques that work robustly under all these conditions.

[Related Technical Literature]

1. Method and apparatus for object recognition using region of interest in image and contour image (Patent Application No. 10-2009-0109727)

2. Apparatus and method for recognizing flowers in digital image processor (Patent Application No .: 10-2008-0134964)

An object of the present invention is to provide a method and apparatus for extracting an object from an image and determining the validity of the object based on the symmetry of the extracted object. Another object of the present invention is to provide a method and apparatus for preprocessing an image and automatically extracting an object region from an image background and an object boundary point.

According to a first aspect of the disclosed technology to achieve the above technical problem, a method of determining the validity of the extracted object region by extracting an object from an image, the method comprising: receiving an image including an object; Extracting the object by detecting an outline of the object in the image; Calculating a symmetry measure value for the extracted object; And determining the validity of the extracted object based on the symmetry measure value, and determining that the extraction of the object is valid when the symmetry measure value is smaller than a first predetermined threshold value. Provide a method of judgment.

The second aspect of the disclosed technology to achieve the above technical problem, in the method of extracting the object by detecting the contour of the object in the image, pre-processing the image to remove the edge (edge) component in the background region other than the object Reducing; Generating vertical and horizontal lines through the object in the image; Selecting four points where the vertical line and the horizontal line meet a boundary between the object and the background; And detecting an outline by tracking a boundary of the object from each of the four points.

According to a third aspect of the disclosed technology, an apparatus for extracting an object from an image and determining the validity of the extracted object region may include: an image input unit configured to receive an image including an object; An object extracting unit extracting the object by detecting an outline of the object in the image; A symmetry measure value calculator for calculating a symmetry measure value for the extracted object; And a validity determination unit that determines the validity of the extracted object based on the symmetry measure value, and determines that the extraction of the object is valid when the symmetry measure value is smaller than a first predetermined threshold value. Provide an apparatus for determining validity.

The disclosed technique may have the following effects. It is to be understood, however, that the scope of the disclosed technology is not to be construed as limited thereby, as it is not meant to imply that a particular embodiment should include all of the following effects or only the following effects.

The method of determining the validity of the object region according to an embodiment determines the validity of the object region based on the symmetry, and thus has an advantage of efficiently determining an object having symmetry. In addition, when only a partial region of the object is recognized and the object is extracted asymmetrically, the user may draw and select an unextracted region to reduce the process of re-evaluating the validity of the entire object. .

The object extraction method according to an embodiment extracts an object using an IS (Intelligent Scissor) algorithm at each of four boundary points of the object and the background, so that the object extraction method can be performed faster than the existing extraction method.

1 is a block diagram illustrating an apparatus for determining validity of an object region according to an embodiment of the present invention.
2 is a flowchart illustrating a step of determining the validity of an object region according to an embodiment of the present invention.
3 is a diagram illustrating an object extracted from an input image.
4 is a diagram illustrating a center axis of an object and a distance from the center axis to an outline of the object in order to calculate a symmetry measure value according to an embodiment of the present invention.
5 is a graph showing a distribution of symmetry scale values calculated using distance information from an object's central axis to an object's contour.
FIG. 6 is a diagram illustrating an image of contours extracted from an object and an image from which an area other than the object is removed according to an embodiment of the present invention.
7 is a graph showing a horizontal line passing through an object and a local cost (LC) value of the horizontal line.
FIG. 8 is a graph illustrating a horizontal line passing through an object and a local cost (LC) value of the horizontal line in an image in which preprocessing is performed according to an embodiment of the present invention.

The description of the disclosed technique is merely an example for structural or functional explanation and the scope of the disclosed technology should not be construed as being limited by the embodiments described in the text. That is, the embodiments may be variously modified and may have various forms, and thus the scope of the disclosed technology should be understood to include equivalents capable of realizing the technical idea.

On the other hand, the meaning of the terms described in the present invention will be understood as follows.

Singular expressions should be understood to include plural expressions unless the context clearly indicates otherwise, and terms such as "comprise" or "have" refer to a feature, number, step, operation, component, part or implementation thereof. It is to be understood that the combination is intended to be present, but not to exclude in advance the possibility of the presence or addition of one or more other features or numbers, steps, operations, components, parts or combinations thereof.

Each step may occur differently from the stated order unless the context clearly dictates the specific order. That is, each step may occur in the same order as described, may be performed substantially concurrently, or may be performed in reverse order.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the disclosed technology belongs, unless otherwise defined. Terms defined in commonly used dictionaries should be interpreted to be consistent with meaning in the context of the relevant art and can not be construed as having ideal or overly formal meaning unless expressly defined in the present application.

1 is a block diagram illustrating an apparatus for determining validity of an object region according to an embodiment of the present invention. Referring to FIG. 1, the apparatus 100 for determining the validity of an object includes an image input unit 110, an object extractor 120, a symmetry measure value calculator 130, and a validity determiner 140.

The image input unit 110 includes a photographing element such as a camera or an image sensor, and receives an image including an object by photographing. Herein, the object is a subject corresponding to a photographing purpose, and various subjects may be input as objects according to embodiments. However, since the present invention has an advantage of extracting a region of a flower from an image of a flower, an image is described in the following description. As an image including a flower, an object is described as referring to a flower. The image input unit 110 receives an image by real-time shooting, but according to an embodiment, the image input unit 110 may receive image data from a DB (not shown) that stores the previously photographed image and receive the image.

The object extractor 120 extracts the contour of the object by using the image input from the image input unit 110. The detection of the contour creates vertical and horizontal lines through the object in the image, and selects four points where the vertical and horizontal lines meet the boundary points of the object and the background. The contour is then detected by tracking the boundary of the object from each of the four points.

Detecting the contour in detail, according to an embodiment, the object extractor 120 reduces the edge component of a portion of the image corresponding to the background other than the object by preprocessing the input image. Here, the edge component means a portion in which the color component values R, G, and B of the pixels constituting the image change rapidly, and are calculated by a local coast value (LC value). The calculation of the LC value will be described in the description of FIG. 7. In an embodiment, the reduction of edge components is performed using an adaptive filtering algorithm. The reason for reducing the edge component by preprocessing is to prevent an object from being incorrectly extracted by an edge component existing in an area other than the boundary of the object in the image, that is, a part corresponding to the background.

Herein, reference will be made to FIGS. 7 and 8 to easily describe the edge component.

FIG. 7 (a) shows a horizontal line passing through the image, and FIG. 7 (b) shows a LC value of pixels forming the horizontal line. The LC value is represented by equation (1).

In Equation 1, G denotes a gradient size value of a pixel, and MG denotes a gradient value G having a maximum size in all pixels. In this case, the G value is calculated by Equation 2,

은 컬러영상에서 Red성분의 Gradient,

과

은 각각 Red성분의 수평 및 수직성분의 Gradient 값, I는 픽셀값을 의미하며 수학식 3에 의해 계산된다.

Gradient of Red component in color image,

and

Are the gradient values of the horizontal and vertical components of the red component, respectively, and I is the pixel value, and is calculated by Equation 3.

According to FIG. 7 (b), since the LC value shows a sharp change in value (edge component) in the background region or the internal region of the object in addition to the boundary between the image and the object, it acts as an obstacle to detecting the boundary between the background and the object. do.

Therefore, an adaptive filtering algorithm is used to reduce the edge component of the background to facilitate detection of the boundary between the background and the object.

FIG. 8 is a graph showing a local cost (LC) value of a horizontal line and a horizontal line passing through an object in an image in which preprocessing is performed according to an embodiment of the present invention.

FIG. 8 (a) shows an image of pre-processing by applying an adaptive filtering algorithm according to Equation 4. FIG.

는 필터링 후의 Gradient 값을 의미하여,

는 입력 Gradient 값,

은 m에 따라 정해지는 상수 값을 의미한다. 적응적 필터링 알고리즘을 적용하여 도 8 (a)에서 도시하는 바와 같이 배경의 엣지 성분들이 제거되었음을 알 수 있다.In equation (4)

Means the gradient value after filtering,

Is the input gradient value,

Is a constant value determined by m. It can be seen that the edge components of the background have been removed by applying the adaptive filtering algorithm as shown in FIG.

8 (b) is a graph showing a local cost (LC) value of a horizontal line passing through an object in an image in which the preprocessing of the adaptive filtering algorithm is performed. Edge components are removed from the background to clearly see the boundary between the background and the object.

FIG. 8 shows that two boundary points of the background and the object are selected by a horizontal line passing through the object, but the object extractor 120 generates two vertical lines passing through the object in the same manner to select two boundary points of the background and the object. Select four points for both the background and the boundary of the object. According to the embodiment, the selection of each point is to select the point with the minimum LC value on the horizontal line and the vertical line.However, the LC value may change the maximum value and the minimum value according to the formula, so each point is a point where the LC value changes rapidly. Can be selected as the boundary point.

The object extractor 120 then tracks the boundary of the object from each of the four points to detect the contour. At this time, tracking is performed by an intelligent scissor (IS) algorithm. By performing the IS algorithm, the lines connecting the four points are connected to each other, the contour is detected, and the object is extracted. In this case, the extraction of the object means that the outline of the object forms a closed loop so that the object has an area separated from the background of the image. On the other hand, the IS algorithm is performed based on the equation (5).

는 비용함수로서 포인트 p로부터 다음 포인트 q를 찾는데 사용되는 값이다.

는 canny edge의 결과 값

와 그레디언트(Gradient) 값

의 가중치

의 결합으로 계산된다.

값을 이용하여 영상의 두 점 사이의 최소 비용을 가지는 최적 경로의 연산은 Two-Dimensional Dynamic Programming 알고리즘에 의하여 수행된다. here

Is a cost function that is used to find the next point q from point p.

Is the result of the canny edge

And gradient values

Weight of

Calculated by the combination of.

The calculation of the optimal path with the minimum cost between two points in the image using the value is performed by two-dimensional dynamic programming algorithm.

Referring back to FIG. 1, the symmetry measure value calculator 130 calculates a symmetry measure value for the extracted object. On the other hand, the image received by the image input unit 110 in order to check the horizontal level of the object and to validate the symmetry scale value according to the embodiment the validity determination apparatus 100 of the object further includes an image rotating unit (not shown) can do.

According to an embodiment, the image rotating unit (not shown) may have a straight line connecting a coordinate having a maximum distance value from the center point of the extracted object to the contour of the object (the boundary between the object and the background) and the vertical or horizontal axis of the image. Rotate the image by the calculated angle by calculating the angle. According to another embodiment, the image rotating unit (not shown) may rotate only the extracted object.

According to one embodiment, the calculation of the symmetry measure value is calculated by the average of the normalized distance difference values from the center axis of the extracted object to the contour of the object.

Reference is made to FIG. 4 to clarify the average of the distance values.

4 is a diagram for measuring a distance from the central axis and the central axis of the object to the contour of the object. The horizontal axis W means width and the vertical axis H means height. D1 and d2 of FIG. 4 (a) show distances to the contour of the object based on the vertical axis of the object.

Equation 6 calculates a symmetry measure value based on the vertical axis.

Equation 7 calculates a symmetry scale value based on the horizontal axis.

In this case, the horizontal axis W means width and the vertical axis H means height. In FIG. 4B, d3 and d4 (not shown) represent distances to the contour of the object based on the horizontal axis of the object.

The validity determination unit 140 determines the validity of the extracted object based on the symmetry scale values VR to HR. According to an embodiment, when the object has only left and right symmetry based on the vertical axis, the validity determining unit 140 determines the validity of the extracted object based on the VR value. When the object has left and right symmetry, if the object is extracted correctly by the object extractor 120, d1 and d2 have a normal distribution in the extracted object. Therefore, the validity determination unit 140 compares the VR value with a predetermined threshold value (first threshold value) and determines that the extraction of the object is valid when the VR value is smaller than the first threshold value.

Reference is made to FIGS. 3 and 5 to clarify the validity judgment.

In FIG. 3B, since only a partial region of the object is extracted, the VR value is calculated to be larger than the first threshold value. Therefore, the validity determination unit 140 determines that the object extraction of FIG. 3B is not valid. 3 (a), the VR value is calculated to be smaller than the first threshold value because the object is well extracted according to the area of the object. Therefore, the validity determination unit 140 determines that the object extraction of FIG. 3 (a) is valid.

5 is a graph showing a distribution of symmetry scale values calculated using distance information from an object's central axis to an object's contour. When the object region is effectively extracted from the image, as shown by the blue line, d1 and d2 (or d3 and d4) show a symmetrical distribution. However, if the object area is not extracted effectively, the distance value shows a large distribution as indicated by the red line.

According to an embodiment, the validity may be determined based on a VR value when the object has left and right symmetry, but the validity may be determined based on an HR value when the object has up and down symmetry. In addition, when the object has symmetry both up, down, left and right, the validity of object extraction may be determined based on the average value of the VR value and the HR value.

At this time, the symmetry measure value is calculated by Equation 8.

According to an exemplary embodiment, when the symmetry measure value is larger than the first threshold but smaller than the second threshold, the validity determination unit 140 may receive at least some areas of the object by a user manipulation. Even when the algorithm is separated, if the background and the object are difficult to separate, the user can receive the object area directly from the user. According to an embodiment, when the symmetry measure value is greater than the first threshold value and greater than the second threshold value, the image may be newly received through the image input unit 110.

Since the apparatus for determining the validity of the object region according to the above-described exemplary embodiment determines the validity of the object region based on the symmetry, there is an advantage in that an object having symmetry can be efficiently determined. In addition, when only a part of the object is recognized and the object is asymmetrically extracted, the user can draw and select an unextracted area to reduce the process of re-evaluating the validity of the entire object. There is an advantage.

In addition, the object extraction method according to the above-described embodiment extracts the object using IS (Intelligent Scissor) algorithm at each of the four boundary points of the object and the background, so that the object can be extracted more quickly than the existing extraction method. have.

2 is a flowchart illustrating a step of determining the validity of an object region according to an embodiment of the present invention. Hereinafter, a method of determining the validity of the object area will be described with reference to FIG. 2. Even when the validity determination apparatus 100 of FIG. 1 is implemented in a time series, the image input unit 110, the object extractor 120, the symmetry scale calculation unit 130, and the validity determination unit 240 correspond to the present embodiment. The part described with respect to is also applied to this embodiment as it is.

In operation S201, the validity determination apparatus 100 receives an image using a photographing apparatus. Image input may receive an image by real-time shooting, and in accordance with an embodiment, may receive the image data received from a DB (not shown) that stores the pre-recorded image.

In operation S202, the validity determination device 100 automatically detects the object area. The automatic detection means that the validity determination device 100 detects and extracts an outline of an object included in the image by performing a data operation on the input image. Detection of contours creates vertical and horizontal lines through the object in the image. The validity judgment apparatus 100 selects four points where the vertical line and the horizontal line meet the boundary point between the object and the background. The validity judgment apparatus 100 detects an outline by tracking an interface of an object from each of four points.

In operation S203, the validity determination apparatus 100 rotates the image. The image received in step S201 performs the rotation of the image to check the horizontality of the object and to validate the symmetry measure value. According to one embodiment, the angle calculated by calculating the angle between the vertical axis or the horizontal axis of the image and a straight line connecting the coordinates having the maximum distance value from the center point of the extracted object to the contour of the object (the boundary between the object and the background) Rotate the image by According to another embodiment, only the extracted object may be rotated.

In operation S204, the validity determination device 100 calculates a symmetry measure value. The symmetry measure value is calculated by the above equations (6) to (8).

In operation S205, the validity determination apparatus 100 determines the validity of the extracted object. At this time, if the symmetry measure value is less than the first threshold value, the validity determination device 100 determines that the extraction of the object is valid.

When the symmetry measure value is greater than the first threshold value, the validity determination device 100 compares the predetermined second threshold value (S207), and when the symmetry measure value is smaller than the second threshold value, inputs at least a partial region of the object by a user's manipulation. If it is greater than the second threshold (S208), the image containing the object is input again (S209).

The method of determining the validity of the object region according to the above-described exemplary embodiment has an advantage of efficiently determining an object having symmetry because the validity of the object region is determined based on the symmetry. In addition, when only a part of the object is recognized and the object is asymmetrically extracted, the user can draw and select an unextracted area to reduce the process of re-evaluating the validity of the entire object. There is an advantage.

In addition, the object extraction method according to the above-described embodiment extracts the object using IS (Intelligent Scissor) algorithm at each of the four boundary points of the object and the background, so that the object can be extracted more quickly than the existing extraction method. have.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention as defined by the following claims It can be understood that

100: device for determining validity of an object
110: video input unit
120: object extraction unit
130: symmetry scale value calculation unit
140: validity judgment unit

Claims (16)

In the method of extracting the object from the image to determine the validity of the extracted object area,
(a) receiving an image including an object;
(b) extracting the object by detecting an outline of the object in the image;
(c) calculating a symmetry measure value for the extracted object; And
(d) determining the validity of the extracted object based on the symmetry measure value,
And determining that the extraction of the object is valid when the symmetry measure value is smaller than a first predetermined threshold value. The method of claim 1, wherein step (d)
When the symmetry measure value is greater than the first threshold value, when compared with a second predetermined threshold value, when the value is greater than the second threshold value, the image including the object is input again, and when the value is smaller than the second threshold value. And receiving at least a partial area of the object by a user's manipulation. The method of claim 1, wherein after step (b)
And rotating the image by the calculated angle by calculating an angle between a straight line connecting a coordinate having a maximum distance value from a center point of the extracted object to a contour of the object and a vertical or horizontal axis of the image. How to determine the validity of the object area. The method of claim 1, wherein the detection of the contour
A method for determining validity of an object region performed by a preprocessing process by an adaptive filtering algorithm and an intelligent scissor (IS) algorithm. In the method for extracting the object by detecting the contour of the object in the image,
(a) pre-processing the image to reduce edge components in the background region other than the object;
(b) generating vertical and horizontal lines through the object in the image;
(c) selecting four points where the vertical line and the horizontal line meet a boundary between the object and the background; And
(d) detecting an outline by tracking a boundary of the object from each of the four points. 6. The method of claim 5,
Wherein said preprocessing is performed using an adaptive filtering algorithm. 6. The method of claim 5,
And the tracking is performed using an intelligent scissor (IS) algorithm. 6. The method of claim 5, wherein the selection of each of the four points is
Extracting an object for selecting a point at which the cost value (LC value) of the vertical line and the horizontal line changes rapidly. The method of claim 8,
Each of the points is the point at which the LC value is the minimum.
Wherein the LC value is calculated by the formula LC = 1 + MG-G, where G is the gradient size value of the pixel and MG is the G value having the maximum size in all pixels. The method of claim 8,
And calculating the LC value based on the color component values most included in the object among the color components (RED, BLUE, GREEN). The method of claim 1,
The method of determining the validity of the object region is calculated by the average of the distance difference value from the center axis of the extracted object to the contour of the object. 12. The method of claim 11, wherein said average is
The method of determining the validity of the object region which is an average obtained by normalizing the distance from the central axis to the contour of the object. 12. The method of claim 11,
And the central axis is a vertical axis or a horizontal axis with respect to the extracted object. 12. The method of claim 11 wherein the calculation of the symmetry measure value is
Equation

(Where R is the symmetry measure value, VR is the symmetry measure value for the longitudinal axis, HR is the symmetry measure value for the abscissa),
The VR and HR formula

;

and d1 and d2 are distance values from a vertical axis to the detected contour, and d3 and d4 are distance values from a horizontal axis to the detected contour. An apparatus for determining the validity of the extracted object region by extracting an object from an image,
An image input unit configured to receive an image including an object;
An object extracting unit extracting the object by detecting an outline of the object in the image;
A symmetry measure value calculator for calculating a symmetry measure value for the extracted object; And
The validity of the extracted object is determined based on the symmetry measure value.
And a validity determination unit that determines that the extraction of the object is valid when the symmetry measure value is smaller than a first predetermined threshold value. The method of claim 15, wherein the validity determination unit
When the symmetry measure value is larger than the first threshold value, compared to a second predetermined threshold value, when the symmetry measure value is larger than the second threshold value, the image including the object is input again, and when the value is smaller than the second threshold value. An apparatus for determining validity of an object area receiving at least a partial area of the object by a user's manipulation.

Images