KR20210005413A - Apparatus and method for subpixel line localization - Google Patents

Abstract

Disclosed are an apparatus and a method for determining subpixel line positions, capable of efficiently determining positions of line points corresponding to a line from positions of line pixels determined at a pixel level related to the line on an image, and a recording medium. The method for determining subpixel line positions according to an embodiment of the present invention is a method for determining subpixel line positions which determines positions of line points corresponding to lines from positions of line pixels determined at pixel levels related to the lines on an image comprising: a step of calculating gradients related to pixel value differences between adjacent pixels with respect to pixels including line pixels and peripheral pixels adjacent to the line pixels using a gradient calculation part; a step of calculating gradient magnitudes of the pixels and regularizing the gradients of the pixels based on the gradient magnitudes of the pixels to calculate subpixel level displacements with respect to the positions of the line pixels using a regularizing part; and a step of determining the positions of the line points at subpixel levels based on the positions of the line pixels and the subpixel level displacements using the subpixel line position determining part.

Description

Apparatus and method for determining sub-pixel line positioning {APPARATUS AND METHOD FOR SUBPIXEL LINE LOCALIZATION}

The present invention relates to an apparatus and method for determining a subpixel line position, and more particularly, to a subpixel level determining a position of a line point corresponding to the line from a position of a line pixel determined as a pixel level related to a line in an image. It relates to an apparatus and method for determining pixel line position.

In image processing terms, a line refers to a portion having a geometrically elongated shape with a brightness that is brighter or darker than the surroundings of both sides. As one of the most important and frequently observed features in various imaging technologies including indoor imaging, building exterior imaging, satellite imaging, medical imaging, and road imaging, line features are used to analyze and interpret scenes in images. do. Therefore, accurate line extraction is important in the field of image processing and computer vision.

Recently, with the development of various application fields including lane extraction for autonomous driving, blood vessel extraction from medical images, and road extraction from satellite images, demand for efficient extraction of line features from images is rapidly increasing. In order to determine the location of the linear features with high quality, it is necessary to determine the location of the linear features at the subpixel level. Steger proposed a method for determining subpixel line positioning based on a Taylor series Approximation (TA) of surface intensity and a normal vector derived from a Hessian matrix. However, this method has a problem in that the positions of linear features are determined unstable under a weak line signal and a high noise level. In addition, the method proposed by Steger checks only the positions within the three pixels in the vicinity of the current line diffusion direction and connects the positions, so when the line features have high curvature, the line positions cannot be properly connected. There is.

C. Steger, "An unbiased detector of curvilinear structures", IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 20, no. 2, pp. 113-125, 1998.

The present invention provides a subpixel line positioning apparatus and method capable of efficiently determining a position of a line point corresponding to a line at a subpixel level from a position of a line pixel determined at a pixel level related to a line in an image, and a recording medium. will be.

In addition, the present invention is to provide a subpixel line positioning apparatus and method, and a recording medium capable of generating a line by connecting line points so as to have straightness and omni-directionality.

In addition, the present invention is to provide a subpixel line positioning apparatus and method, and a recording medium capable of accurately connecting line points and extracting lines from an image with high quality even when line features have a high curvature.

Further, the present invention is to provide a subpixel line positioning apparatus and method, and a recording medium capable of extracting high-quality lines at high speed with a small computational throughput.

The problem to be solved by the present invention is not limited to the problems mentioned above. Other technical problems that are not mentioned will be clearly understood by those of ordinary skill in the art from the following description.

A method for determining a subpixel line position according to an aspect of the present invention includes determining a position of a line point corresponding to the line at a subpixel level from a position of a line pixel determined at a pixel level related to a line in an image. A method comprising: calculating, by a gradient calculator, gradients related to a pixel value difference between adjacent pixels for pixels including the line pixel and adjacent pixels adjacent to the line pixel; Normalizing, by a normalizing unit, the gradients of the pixels based on the sizes of the gradients of the pixels to calculate the sizes of the gradients of the pixels and calculate the subpixel level displacement with respect to the position of the line pixel; And determining, by the subpixel line position determining unit, the position of the line point at the subpixel level based on the position of the line pixel and the subpixel level displacement.

The determining of the position of the line point may include determining the position of the line point as a sub-pixel level by summing the position of the line pixel and the sub-pixel level displacement.

Normalizing the gradients of the pixels may include calculating, by a gradient summing unit, a summed gradient by summing the gradients of the pixels; Calculating sizes of gradients of the pixels by a gradient size calculator; Calculating a summed gradient size by summing the sizes of the gradients of the pixels by a gradient size summation unit; And calculating the subpixel level displacement by normalizing the summed gradient by the summed gradient size by a subpixel displacement calculator.

The method for determining a line position of a subpixel according to the present invention may further include determining, by a window setting unit, pixels for determining the position of the line point from the position of the line pixel in the image. The determining of the pixels includes setting a window having a (2K + 1) × (2K + 1) pixel size (K is a positive integer) around the line pixel, and determining pixels within the window. can do.

The calculating of the summed gradient may include calculating, by a row direction gradient summation unit, a row direction summed gradient by summing the row direction gradients of the pixels; And calculating a column direction sum gradient by summing the column direction gradients of the pixels by the column direction gradient summing unit. In the calculating of the sizes of the gradients, sizes of the row direction gradients and the sizes of the column direction gradients may be calculated. In the calculating of the summed gradient size, the summed gradient size may be calculated based on the sizes of the row direction gradients and the column direction gradients. The calculating of the subpixel level displacement may include calculating a row direction subpixel level displacement by dividing the row direction sum gradient by the gradient sum size by a row direction displacement calculator; And calculating a column direction subpixel level displacement by dividing the column direction sum gradient by the gradient sum size by a column direction displacement calculator. The determining of the position of the line point may include determining a row direction position of the line point as a subpixel level by summing the row direction position of the line pixel and the row direction subpixel level displacement; And determining the column direction position of the line point as a subpixel level by summing the column direction position of the line pixel and the column direction subpixel level displacement.

The normalizing of the gradients of the pixels may include: when the line pixel is ridge, normalizing the gradients of the pixels by applying a first sign to the sum of the gradients; And when the state of the line pixel is a valley, applying a second sign different from the first sign to the sum of the gradients to normalize the gradients of the pixels.

In the calculating of the summed gradient, the summed row direction gradient and the summed column direction gradient may be calculated according to Equation 1 below. In the calculating of the summed gradient size, the summed gradient size may be calculated according to Equation 2 below. In the calculating of the subpixel level displacement, the subpixel level displacement may be calculated according to Equations 3 and 4 below. In determining the position of the line point, the position of the line point may be determined according to Equation 5 below.

[Equation 1]

[Equation 2]

[Equation 3]

[Equation 4]

[Equation 5]

는 상기 행 방향 합산 그래디언트,

는 상기 열 방향 합산 그래디언트,

는 행 방향 그래디언트,

는 열 방향 그래디언트,

는 상기 그래디언트 합산 크기,

는 상기 라인픽셀의 상태가 릿지인지 밸리인지를 나타내는 정보,

는 상기 행 방향 서브픽셀 레벨 변위,

는 상기 열 방향 서브픽셀 레벨 변위, c는 상기 라인픽셀의 열 방향 픽셀값, r은 상기 라인픽셀의 행 방향 픽셀값,

는 상기 라인점의 행 방향 위치,

는 상기 라인점의 열 방향 위치를 나타낸다.In Equations 1 to 5,

Is the sum gradient in the row direction,

Is the summation gradient in the column direction,

Is the row direction gradient,

Is the column direction gradient,

Is the sum of the gradients,

Is information indicating whether the state of the line pixel is ridge or valley,

Is the subpixel level displacement in the row direction,

Is the column direction subpixel level displacement, c is the column direction pixel value of the line pixel, r is the row direction pixel value of the line pixel,

Is the row direction position of the line point,

Represents the position of the line point in the column direction.

The method for determining a line position of a subpixel according to the present invention may further include connecting the line points to generate a line in the image by a line connector. The connecting of the line points may include coordinate distances between a first line point included in the line and second line points in the vicinity of the first line point among the line points, and the first line point and the A line point to be connected to the first line point among the second line points may be determined based on angles between the second line points.

The connecting of the line points may include calculating a coordinate distance between a first line point included in the line and a second line point in the vicinity of the first line point among the line points by a distance calculator; Calculating a first angle of the line at the first line point by an angle calculator, and calculating a second angle between the first line point and the second line point; Calculating an angle difference between the first angle and the second angle by an angle difference calculator; Calculating, by a connection distance calculator, a connection distance between the first line point and the second line point based on the coordinate distance and the angle difference; And determining, by a line point determining unit, a line point having a smallest connection distance to the first line point among the second line points, and connecting the line point to the first line point.

In the calculating of the coordinate distance, the coordinate distance may be calculated according to Equation 6 below. In the calculating of the second angle, the second angle may be calculated according to Equation 7 below. In the calculating of the angle difference, the angle difference may be calculated according to Equation 8 below. In the calculating of the connection distance, the connection distance may be calculated according to Equations 9 and 10 below.

[Equation 6]

[Equation 7]

[Equation 8]

[Equation 9]

[Equation 10]

은 상기 좌표 거리,

는 상기 제1 라인점의 행 방향 좌표,

는 상기 제1 라인점의 열 방향 좌표,

는 상기 제2 라인점의 행 방향 좌표,

는 상기 제2 라인점의 열 방향 좌표,

는 상기 각도차,

는 상기 제1 각도,

는 상기 제2 각도,

는 설정된 각도차 반영 인자,

는 상기 각도차에 따라 결정되는 각도거리,

는 상기 연결거리를 나타낸다.In Equations 6 to 10,

Is the coordinate distance,

Is the row direction coordinate of the first line point,

Is the column direction coordinate of the first line point,

Is the row direction coordinate of the second line point,

Is the column direction coordinate of the second line point,

Is the angle difference,

Is the first angle,

Is the second angle,

Is the set angle difference reflection factor,

Is the angular distance determined according to the angular difference,

Represents the connection distance.

According to another aspect of the present invention, there is provided a computer-readable recording medium in which a program for executing the subpixel line positioning method is recorded.

A subpixel line position determining apparatus according to another aspect of the present invention includes a subpixel line for determining a position of a line point corresponding to the line as a subpixel level from a position of a line pixel determined as a pixel level related to a line in an image. A positioning device, comprising: a gradient calculator configured to calculate gradients related to a pixel value difference between adjacent pixels for pixels including the line pixel and adjacent pixels adjacent to the line pixel; A normalization unit that normalizes the gradients of the pixels based on the sizes of the gradients of the pixels to calculate the sizes of the gradients of the pixels and to calculate the subpixel level displacement with respect to the position of the line pixel; And a subpixel line position determining unit configured to determine the position of the line point at a subpixel level based on the position of the line pixel and the subpixel level displacement.

The normalization unit may include: a gradient summing unit for calculating a sum gradient by summing the gradients of the pixels; A gradient size calculator that calculates the sizes of the gradients of the pixels; A gradient size summing unit for calculating a summed gradient size by summing the sizes of the gradients of the pixels; And a subpixel displacement calculator configured to calculate the subpixel level displacement by normalizing the summed gradient based on the summed gradient size. The subpixel line position determining unit may determine the position of the line point as the subpixel level by summing the position of the line pixel and the subpixel level displacement.

The subpixel line position determining apparatus according to the present invention may further include a window setting unit that determines pixels for determining the position of the line point from the position of the line pixel in the image. The window setting unit may set a window having a (2K + 1) × (2K + 1) pixel size (K is a positive integer) centered on the line pixel, and determine pixels in the window.

The gradient summing unit may include a row directional gradient summing unit for calculating a row directional sum gradient by summing the row directional gradients of the pixels; And a column direction gradient summing unit configured to calculate a column direction sum gradient by summing the column direction gradients of the pixels. The gradient size calculator may calculate sizes of the row direction gradients and the column direction gradients. The gradient summation unit may calculate the summed gradient size based on the sizes of the row direction gradients and the column direction gradients. The subpixel displacement calculator may include a row direction displacement calculator configured to calculate a row direction subpixel level displacement by dividing the row direction sum gradient by the gradient sum size; And a column direction displacement calculator configured to calculate a column direction subpixel level displacement by dividing the column direction sum gradient by the gradient sum size. The subpixel line position determining unit may determine a row direction position of the line point as a subpixel level by summing a row direction position of the line pixel and the row direction subpixel level displacement; Further, the column direction position of the line point may be determined as a subpixel level by summing the column direction position of the line pixel and the column direction subpixel level displacement.

When the state of the line pixel is a ridge, the normalization unit normalizes the gradients of the pixels by applying a first sign to the sum of the gradients; Further, when the state of the line pixel is a valley, the gradients of the pixels may be normalized by applying a second sign different from the first sign to the sum of the gradients.

The subpixel line positioning apparatus according to the present invention may further include a line connection unit connecting the line points to generate a line in the image. The line connection unit may include coordinate distances between a first line point included in the line and second line points in the vicinity of the first line point among the line points, and the first line point and the second line point A line point to be connected to the first line point among the second line points may be determined based on the angles between them.

The line connection unit may include a distance calculator configured to calculate a coordinate distance between a first line point included in the line and a second line point near the first line point among the line points; An angle calculator configured to calculate a first angle of the line at the first line point and to calculate a second angle between the first line point and the second line point; An angle difference calculator configured to calculate an angle difference between the first angle and the second angle; A connection distance calculator configured to calculate a connection distance between the first line point and the second line point based on the coordinate distance and the angle difference; And a line point determining unit connecting a line point having a smallest connection distance to the first line point among the second line points with the first line point.

According to an embodiment of the present invention, a subpixel line positioning apparatus and method for efficiently determining a position of a line point corresponding to a line from a position of a line pixel determined at a pixel level related to a line in an image at a subpixel level, and recording Media is provided.

In addition, according to an embodiment of the present invention, line points can be connected to have straightness and omni-directionality, and even when line features have high curvature, line points can be accurately connected, and low computational throughput As a result, high-quality lines can be extracted at high speed.

The effect of the present invention is not limited to the above-described effects. Effects not mentioned will be clearly understood by those of ordinary skill in the art from the present specification and the accompanying drawings.

1 is a flowchart of a method of determining a line position of a subpixel according to an exemplary embodiment of the present invention.
2 is a configuration diagram of a subpixel line positioning apparatus according to an embodiment of the present invention.
3A is a diagram illustrating a window set according to an embodiment of the present invention.
3B is a configuration diagram of a gradient calculator constituting a subpixel line positioning apparatus according to an embodiment of the present invention.
4 is a flowchart of step S300 of FIG. 1.
5 is a block diagram of a normalization unit constituting a subpixel line positioning apparatus according to an embodiment of the present invention.
6 is a flowchart of step S360 of FIG. 5.
7 is a flowchart of step S500 of FIG. 1.
8A is a block diagram of a line connection part of a subpixel line positioning apparatus according to an embodiment of the present invention.
8B is a diagram for describing a process of calculating a connection distance between a line point and a candidate line point according to an embodiment of the present invention.
9 is a diagram showing connection distance characteristics of a method for determining a line position of a subpixel according to an embodiment of the present invention.
10 is a diagram illustrating images including various line features used to evaluate the performance of a method for determining a subpixel line position according to an embodiment of the present invention.
11 is a diagram illustrating line points positioned in units of subpixels according to an embodiment of the present invention.
12 is a view showing a comparison of the square root mean error performance of the Example of the present invention and Comparative Example.
13 is a diagram showing a comparison of average error performance of an embodiment of the present invention and a comparative example for images having various noise sizes and line thicknesses.
14 is a diagram showing natural images used in an experiment for verifying the performance of a method for determining a subpixel line position according to an embodiment of the present invention.
15 illustrates line pixels detected in natural images according to an embodiment of the present invention.
16 is a diagram showing a result of determining the position of a line point in an embodiment of the present invention and a comparative example for natural images.
17 is a diagram showing a result of determining the position of a line point in an embodiment of the present invention and a comparative example for natural images to which noise is added.
18 is a diagram showing a result of generating a line by connecting line points according to an embodiment and a comparative example of the present invention.

Other advantages and features of the present invention, and a method of achieving them will become apparent with reference to embodiments to be described later in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, and the present invention is only defined by the scope of the claims. Even if not defined, all terms (including technical or scientific terms) used herein have the same meaning as commonly accepted by universal technology in the prior art to which this invention belongs. General descriptions of known configurations may be omitted so as not to obscure the subject matter of the present invention. In the drawings of the present invention, the same reference numerals are used as much as possible for the same or corresponding configurations. In order to help the understanding of the present invention, some configurations in the drawings may be somewhat exaggerated or reduced.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present application, terms such as "comprise", "have" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or a combination thereof described in the specification. It is to be understood that the possibility of the presence or addition of other features, numbers, steps, actions, components, parts, or combinations thereof, or any further features, is not excluded in advance.

The'~ unit' used throughout this specification is a unit that processes at least one function or operation, and may mean, for example, a hardware component such as software, FPGA, or ASIC. However,'~ part' is not limited to software or hardware. The'~ unit' may be configured to be in an addressable storage medium, or may be configured to reproduce one or more processors. As an example,'~ unit' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, and subs. Routines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The components and functions provided by the'~ unit' may be performed separately by a plurality of elements and the'~ units', or may be integrated with other additional elements.

The subpixel line position determination method according to an embodiment of the present invention estimates the subpixel line position based on the regularized sum of gradients (RSG) of pixels adjacent to the line pixel detected in advance and determined as the pixel level in an image. In addition, it is possible to determine the line position with sub-pixel accuracy accurately, quickly and with low computational throughput, and robustly and accurately determine the line position with sub-pixel accuracy even under weak line signal conditions or strong noise conditions. The method for determining a line position of a subpixel according to an embodiment of the present invention is based on a RSG process in which the sum of gradient vectors of pixels adjacent to the line pixel determined at the pixel level is divided by the sum of the gradient sizes of the pixels. Position can be accurately determined at the subpixel level.

The method for determining a subpixel line position according to an embodiment of the present invention includes a first line based on a coordinate distance and an angle difference between a line point (first line point) belonging to a line and candidate line points (second line points). A connection distance between the point and the second line point may be calculated, and a second line point to be connected to the first line point among the second line points may be determined and connected based on the connection distance calculated based on the coordinate distance and the angle difference. . According to an embodiment of the present invention, a line can be created by connecting line points to have straightness and omni-directionality, and by accurately connecting line points even when line features have high curvature, Lines can be extracted with high quality from an image.

1 is a flowchart of a method of determining a line position of a subpixel according to an exemplary embodiment of the present invention. 2 is a configuration diagram of a subpixel line positioning apparatus according to an embodiment of the present invention. 1 and 2, the subpixel line position determining apparatus 10 according to an exemplary embodiment of the present invention subtracts a position of a line point corresponding to a line from a position of a line pixel determined as a pixel level related to a line in an image. To determine the pixel level, it may include a window setting unit 100, a gradient calculator 200, a normalization unit 300, a subpixel line position determining unit 400, and a line connection unit 500.

The window setting unit 100 may determine pixels for determining the position of the line point from the position of the line pixel detected at the pixel level in the image (S100). The window setting unit 100 sets a local window having a pixel size of (2K + 1) × (2K + 1) pixel (K is a positive integer) centered on the line pixel, and (2K + 1) in the window. × (2K + 1) pixels can be determined.

3A is a diagram illustrating a window set according to an embodiment of the present invention. 1 to 3A, in order to determine the subpixel level line position for the line pixel LP, the window setting unit 100 centers on the line pixel LP corresponding to the (r,c) coordinate. (2K + 1) × (2K + 1) By setting a window (LW) with a pixel size (K is a positive integer), 4K (K+1) adjacent to the line pixel (LP) and the line pixel (LP) A total of (2K + 1) × (2K + 1) pixels including the neighboring pixels PP may be selected.

3A shows an example of a window LW with K = 1, that is, a window LW having a size of 3×3 pixels, but the window setting unit 100 sets the K value of the window LW to 2 or more. Pixels near the line pixel LP may be extracted. In addition, the window setting unit 100 may not necessarily set a square window LW, but may extract pixels near the line pixel LP by setting a window having a polygonal structure of a rectangle or more than a square.

The gradient calculator 200 may calculate gradients related to a pixel value difference between adjacent pixels for pixels including the line pixel LP and the surrounding pixels PP adjacent to the line pixel LP ( S200). 3B is a configuration diagram of a gradient calculator constituting a subpixel line positioning apparatus according to an embodiment of the present invention. 1 to 3B, the gradient calculator 200 may include a row direction gradient calculator 220 and a column direction gradient calculator 240.

The row direction gradient calculator 220 may calculate a row direction gradient by calculating a pixel value difference between pixels adjacent in the row direction. The column direction gradient calculator 240 may calculate a column direction gradient by calculating a pixel value difference between pixels adjacent in the column direction.

The normalization unit 300 may normalize the gradients of the pixels based on the sizes of the gradients of the pixels in order to calculate the sizes of the gradients of the pixels and calculate the subpixel level displacement with respect to the position of the line pixel (S300). 4 is a flowchart of step S300 of FIG. 1. 5 is a block diagram of a normalization unit constituting a subpixel line positioning apparatus according to an embodiment of the present invention. 4 and 5, the normalization unit 300 may include a gradient summing unit 320, a gradient size calculation unit 340, a gradient size summing unit 360, and a subpixel displacement calculation unit 380. have.

The gradient summing unit 320 may calculate a sum gradient by summing the gradients of pixels (S320). The gradient summing unit 320 may include a row direction gradient summing unit 322 and a column directional gradient summing unit 324. The row direction gradient summing unit 322 may calculate the row direction sum gradient by summing the row direction gradients of the pixels. The column direction gradient summing unit 324 may calculate the column direction sum gradient by summing the column direction gradients of pixels. In an embodiment, the gradient summing unit 320 may calculate the summed row direction gradient and the summed column direction gradient according to Equation 1 below.

[Equation 1]

는 행 방향 합산 그래디언트,

는 열 방향 합산 그래디언트,

는 행 방향 그래디언트,

는 열 방향 그래디언트를 나타낸다. (r,c)는 라인픽셀의 좌표이다. 그래디언트 합산부(320)는 (2K + 1) × (2K + 1) 개(K는 양의 정수)의 픽셀들의 크기를 가지는 로컬 윈도우(local window) 내에서 행 방향 그래디언트들의 합인 행 방향 합산 그래디언트

와, 열 방향 그래디언트들의 합인 열 방향 합산 그래디언트

를 각각 계산할 수 있다.In Equation 1,

Is the summation gradient in the row direction,

Is the summation gradient in the column direction,

Is the row direction gradient,

Represents the column direction gradient. (r,c) is the coordinates of the line pixel. The gradient summing unit 320 is a row direction summation gradient that is the sum of row direction gradients within a local window having a size of (2K + 1) × (2K + 1) pixels (K is a positive integer).

Wow, the summation gradient in the column direction, which is the sum of the gradients in the column direction

Each can be calculated.

The gradient size calculator 340 may calculate the sizes of gradients of pixels (S340). The gradient size calculator 340 may calculate the sizes of the row direction gradients and the column direction gradients. The gradient size summing unit 360 may calculate the summed gradient size by summing the sizes of the gradients of the pixels (S360). The gradient size summing unit 360 may calculate the summed gradient size based on the sizes of the row direction gradients and the column direction gradients. In an embodiment, the gradient size summing unit 360 may calculate the sum of the gradient sizes in the local window according to Equation 2 below to calculate the summed gradient size.

[Equation 2]

는 행 방향 그래디언트,

는 열 방향 그래디언트,

는 그래디언트 합산 크기를 나타낸다. (r,c)는 라인픽셀의 좌표이다. 그래디언트 합산 크기

는 서브픽셀 라인 위치들의 그래디언트들의 합(합산 그래디언트)을 정규화하는데 사용된다.In Equation 2,

Is the row direction gradient,

Is the column direction gradient,

Represents the gradient sum size. (r,c) is the coordinates of the line pixel. Gradient sum size

Is used to normalize the sum of the gradients of the subpixel line positions (sum gradient).

6 is a flowchart of step S360 of FIG. 5. 2 and 6, when the state of the line pixel is ridge, the normalization unit 300 normalizes the gradients of the pixels by applying a first sign to the sum of the gradients (S362), and When the state is a valley, the gradients of the pixels may be normalized by applying a second code different from the first sign to the sum of the gradients (S364).

Referring back to FIGS. 4 and 5, the subpixel displacement calculator 380 may calculate the subpixel level displacement by normalizing the sum gradient based on the summed gradient size (S380). The subpixel displacement calculation unit 380 may include a row direction displacement calculation unit 382 and a column direction displacement calculation unit 384. The row direction displacement calculator 382 may calculate the row direction subpixel level displacement by dividing the row direction sum gradient by the gradient sum size. The column direction displacement calculator 384 may calculate the column direction subpixel level displacement by dividing the column direction sum gradient by the gradient sum size. In an embodiment, the subpixel displacement calculator 380 may calculate the subpixel level displacement according to Equations 3 and 4 below.

[Equation 3]

[Equation 4]

는 행 방향 합산 그래디언트,

는 열 방향 합산 그래디언트,

는 라인픽셀의 상태가 릿지인지 밸리인지를 나타내는 정보,

는 행 방향 서브픽셀 레벨 변위,

는 열 방향 서브픽셀 레벨 변위, c는 라인픽셀의 열 방향 좌표, r은 라인픽셀의 행 방향 좌표를 나타낸다. 라인픽셀의 상태를 결정하기 위하여, 아래와 같이 2개의 커널(kernel)이 정의되고, 2개의 커널을 기반으로 하기의 수식에 따라

가 산출될 수 있다.In Equation 3 and Equation 4,

Is the summation gradient in the row direction,

Is the summation gradient in the column direction,

Is information indicating whether the state of the line pixel is ridge or valley,

Is the row direction subpixel level displacement,

Is the subpixel level displacement in the column direction, c is the column direction coordinate of the line pixel, and r is the row direction coordinate of the line pixel. In order to determine the state of the line pixel, two kernels are defined as follows, and based on the two kernels, according to the following equation

Can be calculated.

,

,

값이 0 보다 크면 라인픽셀은 릿지로 분류되고, 그렇지 않으면 라인픽셀은 밸리로 분류될 수 있다. 서브픽셀 변위 산출부(380)는

값을 기반으로 수식 3에 따라 정규화값의 부호를 결정하여 그래디언트 합산 크기에 제1 부호(예를 들어, 양의 부호) 또는 제2 부호(예를 들어, 음의 부호)를 적용하고, 수식 4에 따라 행 방향 합산 그래디언트 및 열 방향 합산 그래디언트를 각각 정규화값의 부호가 적용된 그래디언트 합산 크기 D(r,c)로 나누어 정규화함으로써, 서브픽셀 레벨 변위를 산출할 수 있다. 서브픽셀 레벨 변위

,

는

,

, 및

에 따라 결정될 수 있다. 서브픽셀 레벨 변위는 라인픽셀의 중심에 대한 상대 변위값을 의미한다.

If the value is greater than 0, the line pixel may be classified as a ridge, otherwise the line pixel may be classified as a valley. The subpixel displacement calculation unit 380

Based on the value, the sign of the normalized value is determined according to Equation 3, and the first sign (for example, positive sign) or the second sign (for example, negative sign) is applied to the sum of the gradients, and Equation 4 According to, the subpixel level displacement can be calculated by dividing the summation gradient in the row direction and the summation gradient in the column direction by the summed gradient size D(r,c) to which the sign of the normalization value is applied. Subpixel level displacement

,

Is

,

, And

Can be determined according to. The subpixel level displacement means a relative displacement value with respect to the center of a line pixel.

The subpixel line position determiner 400 may determine the position of the line point at the subpixel level based on the position of the line pixel and the subpixel level displacement (S400). The subpixel line position determiner 400 may determine the position of the line point as the subpixel level by summing the position of the line pixel and the subpixel level displacement. The subpixel line position determining unit 400 determines the row direction position of the line point as the subpixel level by summing the row direction position of the line pixel and the row direction subpixel level displacement, and determines the column direction position of the line pixel and the column direction subpixel level. The position of the line point in the column direction may be determined as the sub-pixel level by summing the pixel-level displacement. In an embodiment, the subpixel line position determiner 400 may determine the position of the line point (subpixel line coordinate) according to Equation 5 below.

[Equation 5]

는 행 방향 서브픽셀 레벨 변위,

는 열 방향 서브픽셀 레벨 변위,

는 라인점의 행 방향 위치,

는 라인점의 열 방향 위치를 나타낸다.In Equation 5, c is the column direction coordinate of the line pixel, r is the row direction coordinate of the line pixel,

Is the row direction subpixel level displacement,

Is the column direction subpixel level displacement,

Is the row direction position of the line point,

Indicates the position of the line point in the column direction.

The line connector 500 may connect line points to create a line in an image. The line connection unit 500 determines coordinate distances between a first line point included in the line, second line points in the vicinity of the first line point among the line points, and angles between the first line point and the second line points. Based on the second line point, a line point to be connected to the first line point may be determined.

7 is a flowchart of step S500 of FIG. 1. 8A is a block diagram of a line connection part of a subpixel line positioning apparatus according to an embodiment of the present invention. 8B is a diagram for describing a process of calculating a connection distance between a line point and a candidate line point according to an embodiment of the present invention. 7 to 8B, the line connection unit 500 includes a distance calculation unit 510, an angle calculation unit 520, an angle difference calculation unit 530, a connection distance calculation unit 540, and a line point determination unit ( 550) may be included.

The distance calculating unit 510 calculates a coordinate distance between the first line point P1 included in the line LN and the second line points P2 and P3 in the vicinity of the first line point P1 among the line points. It can be calculated (S510). The distance calculator 510 may calculate the coordinate distance according to Equation 6 below.

[Equation 6]

은 좌표 거리,

는 제1 라인점(P1)의 행 방향 좌표,

는 제1 라인점(P1)의 열 방향 좌표,

는 제2 라인점(후보 라인점)(P2, P3)의 행 방향 좌표,

는 제2 라인점(P2, P3)의 열 방향 좌표를 나타낸다.

는 제1 라인점(P1)의 좌표이고,

는 제2 라인점(P2, P3)의 좌표이다.In Equation 6,

Is the coordinate distance,

Is the row direction coordinate of the first line point P1,

Is the column direction coordinate of the first line point P1,

Is the row direction coordinate of the second line point (candidate line point) (P2, P3),

Denotes the column direction coordinates of the second line points P2 and P3.

Is the coordinates of the first line point P1,

Is the coordinates of the second line points P2 and P3.

The angle calculation unit 520 calculates a first angle of the line LN at the first line point P1, and calculates a second angle between the first line point P1 and the second line points P2 and P3. It can be calculated (S520). In an embodiment, the angle calculator 520 may calculate the second angle (candidate direction angle) according to Equation 7 below.

[Equation 7]

는 제1 라인점(P1)의 행 방향 좌표,

는 제1 라인점(P1)의 열 방향 좌표,

는 제2 라인점(P2, P3)의 행 방향 좌표,

는 제2 라인점(P2, P3)의 열 방향 좌표,

는 제2 각도를 나타낸다. 각도차 산출부(530)는 제1 각도(현재 라인의 방향 각도)와 제2 각도(후보 방향 각도) 간의 각도차를 산출할 수 있다(S530). 각도차 산출부(530)는 하기의 수식 8에 따라 각도차를 산출할 수 있다.In Equation 7,

Is the row direction coordinate of the first line point P1,

Is the column direction coordinate of the first line point P1,

Is the row direction coordinate of the second line point (P2, P3),

Is the column direction coordinate of the second line point (P2, P3),

Represents the second angle. The angle difference calculator 530 may calculate an angle difference between a first angle (a current line direction angle) and a second angle (a candidate direction angle) (S530). The angle difference calculator 530 may calculate the angle difference according to Equation 8 below.

[Equation 8]

는 각도차,

는 제1 각도,

는 제2 각도를 나타낸다. 연결거리 산출부(540)는 좌표 거리 및 각도차를 기반으로 제1 라인점(P1)과 제2 라인점(P2, P3) 간의 연결거리를 산출할 수 있다(S540). 연결거리 산출부(540)는 하기의 수식 9 및 수식 10에 따라 연결거리를 산출할 수 있다.In Equation 8,

Is the angle difference,

Is the first angle,

Represents the second angle. The connection distance calculation unit 540 may calculate a connection distance between the first line point P1 and the second line point P2 and P3 based on the coordinate distance and the angle difference (S540). The connection distance calculator 540 may calculate the connection distance according to Equations 9 and 10 below.

[Equation 9]

[Equation 10]

는 각도차,

는 설정된 각도차 반영 인자,

는 각도차(

)에 따라 결정되는 각도거리,

는 연결거리를 나타낸다.

는 연결 거리에 대한 각도 거리의 영향을 조절하기 위한 인자이다.

는 예를 들어, 3.0으로 설정될 수 있다. 라인점 결정부(550)는 제2 라인점들(후보 라인점들) 중 제1 라인점(P1)과 가장 연결거리가 작은 라인점(최소 연결거리를 가지는 라인점)을 제1 라인점(P1)과 연결할 수 있다(S550). 다음으로, 선택된 제2 라인점의 방향 및 위치(좌표)는 다음 라인점을 연결을 위한 제1 라인점의 방향 및 위치로 갱신될 수 있다.In Equation 9 and Equation 10,

Is the angle difference,

Is the set angle difference reflection factor,

Is the angular difference (

Angular distance determined according to ),

Represents the connection distance.

Is a factor for controlling the influence of the angular distance on the connection distance.

May be set to 3.0, for example. The line point determination unit 550 selects a line point (a line point having a minimum connection distance) with the smallest connection distance to the first line point P1 among the second line points (candidate line points). It can be connected to P1) (S550). Next, the direction and position (coordinate) of the selected second line point may be updated to the direction and position of the first line point for connecting the next line point.

, 세로축은 각도차(각도거리)

, 색상 스케일은 연결거리 d를 나타낸다. 반투명한 백색 영역은 연결거리가 3.0 이하인 영역을 나타낸다. 도 9의 (b)에 도시된 바와 같이, 본 발명의 실시예에 따른 방법은 좌표거리

가 커질수록(예를 들어, 0.8 픽셀보다 큰 경우), 좌표거리

보다 직진성(straightness)을 더 중요한 인자로 고려한다.9 is a diagram showing connection distance characteristics of a method for determining a line position of a subpixel according to an embodiment of the present invention. FIG. 9A is a diagram showing the connection distance characteristics according to the Steger method, and FIG. 9B is a diagram showing the connection distance characteristics according to Equation 10 above. In the graph shown in Fig. 9(b), the horizontal axis is the coordinate distance (geometric distance)

, The vertical axis is the angle difference (angle distance)

, The color scale represents the connection distance d. The translucent white area represents the area where the connection distance is 3.0 or less. As shown in (b) of Figure 9, the method according to the embodiment of the present invention is a coordinate distance

The larger is (for example, greater than 0.8 pixels), the coordinate distance

Consider straightness as a more important factor.

In addition, the Steger method does not consider line points deviating from the current line direction by more than 90° as candidate connection positions, but the present invention provides line points with a relatively small geometric distance from the current line point (for example, less than 0.8 pixels). In the case of, even if these line points deviate from the line direction of the current line point by 90° or more, they may be considered as candidate connection positions. Accordingly, it is possible to prevent a phenomenon in which the line points are disconnected by connecting the line points so as to have straightness and omni-directionality.

A simulation for evaluating the performance of the method for determining subpixel line positions according to an embodiment of the present invention was performed. For the test of the subpixel line positioning method, as shown in FIG. 10, the blurring factor is 1, and various line thicknesses, line normal angles, and noise strengths are determined. Branches created images to include line features. 10 is an exemplary diagram of four simulated line images having a size of 101×101 pixels. The line thickness of the four simulated line images shown in FIGS. 10A to 10D is 6 pixels, the noise intensity is 8/255, and the line angles are 0°, 23°, 47°, and 82°, respectively.

Prior to the experiment, first, line pixels were detected based on the gradients of pixels in the image. For example, line pixels may be selected based on the sum of gradient angle differences of adjacent pixels, but line pixels are not limited to this method and may be selected in various ways. Line pixels may be determined based on a pixel-by-pixel position. For comparison with the subpixel line position determination method (hereinafter, may be abbreviated as'RSG') according to an embodiment of the present invention, "An unbiased detector of curvilinear structures" (IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 20, no. 2, pp. 113-125, 1998, C. Steger), the Taylor series Approximation (hereinafter, may be abbreviated as'TA') as a comparative example Was determined.

11 is a diagram illustrating line points positioned in units of subpixels according to an embodiment of the present invention. 11A to 11F are simulated images with line thicknesses of 1, 2, 3, 4, 6, and 8, respectively, and noise intensity is all 30/255. In (a) to (f) of FIG. 11, the line angles are 0°, 15°, 30°, 45°, 60°, and 75°, respectively. In the images shown in FIGS. 11A to 11F, red triangles are line points positioned in sub-pixel units according to an embodiment (RSG) of the present invention, and blue squares are sub-pixels according to comparative example (TA). Line points and green arrows positioned in pixels indicate gradient vectors.

It can be seen from the results shown in FIG. 11 that the subpixel line positioning method according to the embodiment of the present invention has superior line point positioning performance than the comparative example TA in various cases. In particular, when the line thickness is 8 pixels and the line signal is relatively weak as shown in Fig. 11(f) where the line angle is 75°, the method according to the embodiment of the present invention is more straightforward than the comparative example, and accurately calculates line points. Was determined, and it can be seen that it has much higher accuracy than the comparative example.

Based on the root mean squared errors (RMSE), the performance of the Example (RSG) and Comparative Example (TA) of the present invention is compared and shown in FIG. 12. 12(a) to (d) show the case where the noise size is 10/255, FIGS. 12(e) to (h) show the case where the noise size is 30/255, and FIGS. 12(i) to ( l) is the case where the noise size is 60/255. 12(a), (e), and (i) show the case where the line thickness is 2 pixels, and FIGS. 12(b), (f), and (j) show the case where the line thickness is 4 pixels, and FIG. 12 (C), (g), and (k) of FIG. 12 are cases where the line thickness is 6 pixels, and (d), (h), and (l) of FIG. 12 are cases where the line thickness is 8 pixels. The method (RSG) according to the embodiment of the present invention exhibited a lower RMSE than the comparative example (TA) in various cases, and in particular, in the case of relatively weak line characteristics due to a large line thickness and a large noise size, the comparative example ( The positions of the line points were more accurately measured in units of subpixels than TA).

A simulation was conducted for evaluating the performance of the method (RGS) and the comparative example (TA) according to the embodiment of the present invention. The average of the RMSEs for various noise sizes and line thicknesses is calculated and shown in FIG. 13. (A), (d), (g), and (j) shown in the first column of FIG. 13 are results for Comparative Example (TA), and (b), (e), and (h) shown in the second column of FIG. ), (k) are the results for the embodiment of the present invention (RSG). (C), (f), (i), and (l) shown in the third column of FIG. 13 are differences between the average RMSE values of the comparative example (TA) and the example of the present invention (RSG). It means that the average RMSE value of the present embodiment (RSG) is low, that is, the performance of the present embodiment (RSG) is superior to that of the comparative example (TA).

(A), (b), and (c) of FIG. 13 are the cases where the smoothing factor is set to 0.5, and (d), (e), and (f) of Fig. 13 are the cases where the smoothing factor is set to 0.75. 13(g), (h), and (i) show the case where the smoothing factor is set to 1.0, and Fig. 13(j), (k), and (l) show the case where the smoothing factor is set to 1.25. In FIG. 13, two horizontal/vertical axes (σ _n , w) are noise magnitude and line thickness. From the illustration of FIG. 13, it can be seen that in various cases, the method RSG according to the embodiment of the present invention can more accurately determine the position of the line points than the comparative example TA. The method (RSG) according to the embodiment of the present invention exhibited significantly higher performance than the comparative example (TA), especially when the noise level is high.

Table 1 shows the comparison of the calculation time performance and the RMSE average of the method RSG and the comparative example TA according to an embodiment of the present invention for various smoothing factors (σ _s ). As the smoothing factor is increased, there is an advantage in that the operation time and RMSE are decreased. However, when the smoothing factor is increased, there is a problem that features in the image are decreased. The method (RSG) according to the embodiment of the present invention is 20% faster than that of the comparative example (TA), especially when the smoothing factor is low, and the position of the line points is accurately determined by subpixel units by 40% or more. Accordingly, according to an embodiment of the present invention, lines can be accurately detected while preventing features from being damaged in an image by applying a low smoothing factor.

An experiment was conducted to verify the performance of the method for determining a subpixel line position of the present invention using natural images shown in FIG. 14. 14A and 14B are well-known images of "Lena" and "Barbara", respectively. 14C and 14D are "Shrub" images and "Facade" images of shrubs and buildings, respectively, taken with a DSLR camera. The range of pixel values for the four images is [0, 255]. Prior to the subpixel line positioning experiment for natural images, first, line pixels in pixel units were detected using a line detection method based on the sum of gradient angle differences. 15 illustrates detected line pixels according to an embodiment of the present invention. Black triangles represent line pixels corresponding to ridges, and white triangles represent line pixels corresponding to valleys. Table 2 shows the number of detected line pixels.

As shown in Fig. 15, line features detected with pixel accuracy typically have a zigzag shape. Experiments were conducted to determine sub-pixel line positions based on line pixels detected for natural images, respectively, by the method (RSG) according to the embodiment of the present invention and the method (TA) of the comparative example. Images were denoised with a smoothing factor of 1.0. Gradients (first-order differences) were calculated by convolution of smoothed images using Sobel operator, and second-order differences were calculated by convolution of first-order difference images using Sobel operator.

16 is a diagram showing a result of line positioning in an embodiment of the present invention and a comparative example. In FIG. 16, red triangles facing upward/downward are detected according to an embodiment (RSG) of the present invention, and blue triangles facing upward/downward are detected according to Comparative Example (TA). In addition, in FIG. 16, blue/red triangles facing upward are line points of a ridge line, and blue/red triangles facing downward are line points of a valley line. In the embodiment of the present invention (RSG), the line points were detected in a continuous line, and the positions of the line points were detected as positions indicating straightness rather than a zigzag shape.

To compare the performance of the present invention (RSG) and the comparative example (TA) under severe noise conditions, random Gaussian noise having a noise size of 10 is added to the original images, and then the noise images are smoothed with a smoothing factor of 1.0. I did. Subsequently, the subpixel line positions were detected for smoothed images by the embodiment (RSG) and the comparative example (TA) of the present invention, and the detected line points are shown in FIG. 17. In FIG. 17, red triangles facing upward/downward are detected according to an embodiment (RSG) of the present invention, and blue triangles facing upward/downward are detected according to Comparative Example (TA). In addition, in FIG. 17, blue/red triangles facing upward are line points of a ridge line, and blue/red triangles facing downward are line points of a valley line.

As shown in FIG. 17, when the subpixel line positions are detected according to the embodiment of the present invention (RSG), the positions of the line points correspond well to line features that can be visually detected in the smoothed image. In order to evaluate the effect of the fluctuation of the noise level, noise images having various noise levels (5, 10, 15, 20, 30) were generated by adding random Gaussian noise. Table 3 shows the average distances for natural images.

Each of d1 to d6 is as follows.

은 노이즈가 없는 이미지에 대해 비교예(TA)에 의해 검출한 라인점의 위치,

은 노이즈가 없는 이미지에 대해 본 발명의 실시예(RSG)에 의해 검출한 라인점의 위치,

은 노이즈가 있는 이미지에 대해 비교예(TA)에 의해 검출한 라인점의 위치,

은 노이즈가 없는 이미지에 대해 본 발명의 실시예(RSG)에 의해 검출한 라인점의 위치를 나타낸다. 표 3에서, d1 내지 d6 앞에 기재된 'R', 'V'는 각각 릿지 라인과 밸리 라인을 나타낸다. 표 3에 도시된 바와 같이, 평균 거리 d1 은 Lena, Barbara, Shrub 이미지들에서 약 0.5 픽셀, Facade 이미지에서 약 0.36 이었다. 평균 거리 d3 은 모든 노이즈 및 자연 이미지들에서 평균 거리 d2 보다 약 30% 정도 작고, 평균 거리 d5 는 모든 경우에서 평균 거리 d4 보다 작으며, 평균 거리 d6은 0.52 ~ 1.17 픽셀 범위인 것을 알 수 있다. 이는 본 발명의 실시예(RSG)에 의하면 다양한 노이즈 레벨에서 비교예(TA) 보다 정확하고 안정적으로 라인위치를 결정할 수 있으며, 다양한 신호 및 노이즈 레벨 하에서 정확한 이미지 처리가 가능한 것을 의미한다.

Is the position of the line point detected by Comparative Example (TA) for an image without noise,

Is the position of the line point detected by the embodiment of the present invention (RSG) for an image without noise,

Is the position of the line point detected by Comparative Example (TA) for the image with noise,

Represents the position of the line point detected by the embodiment of the present invention (RSG) for a noise-free image. In Table 3,'R'and'V' described in front of d1 to d6 represent a ridge line and a valley line, respectively. As shown in Table 3, the average distance d1 was about 0.5 pixels in the Lena, Barbara, and Shrub images, and about 0.36 in the Facade images. It can be seen that the average distance d3 is about 30% smaller than the average distance d2 in all noise and natural images, the average distance d5 is less than the average distance d4 in all cases, and the average distance d6 is in the range of 0.52 to 1.17 pixels. This means that according to the embodiment (RSG) of the present invention, the line position can be determined more accurately and stably than the comparative example (TA) at various noise levels, and accurate image processing is possible under various signal and noise levels.

An experiment was performed to connect line points for the Lena image and the Barbara image. 18 is a view showing a result of generating a line by connecting line points according to an embodiment and a comparative example of the present invention. In FIG. 18, the upper four images represent lines generated by connecting line points according to Comparative Example (TA), and the lower four images are generated by connecting line points according to the embodiment of the present invention (RSG). This is the displayed lines. In FIG. 18, a green line represents a ridge line and a yellow line represents a valley line. Lines consisting of less than 5 line points are not shown in FIG. 18.

From the illustration of FIG. 18, it can be seen that when lines are connected by connecting line points according to an embodiment of the present invention, longer lines than the lines generated by the comparative example TA can be generated. In the case of the comparative example TA, a number of portions in which lines are omitted are found compared to the embodiment of the present invention. The reason that the lines generated by the embodiment of the present invention are longer than that of the comparative example is that they have higher straightness and omni-directionality than that of the comparative example.

At least some of the methods according to the embodiments of the present invention can be written in a program that can be executed on a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium. Computer-readable recording media include volatile memories such as SRAM (Static RAM), DRAM (Dynamic RAM), SDRAM (Synchronous DRAM), Read Only Memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Nonvolatile memory such as Electrically Erasable and Programmable ROM (EEPROM), flash memory device, phase-change RAM (PRAM), magnetic RAM (MRAM), resistive RAM (RRAM), ferroelectric RAM (FRAM), floppy disk, hard disk, or The optical reading medium may be, for example, a storage medium in the form of a CD-ROM or a DVD, but is not limited thereto.

It should be understood that the above embodiments have been presented to aid the understanding of the present invention, and do not limit the scope of the present invention, and various deformable embodiments are also within the scope of the present invention. The technical protection scope of the present invention should be determined by the technical idea of the claims, and the technical protection scope of the present invention is not limited to the literal description of the claims per se, but the invention of the scope of which the technical value is substantially equal. It should be understood that it extends to.

10: subpixel line positioning device
100: window setting unit
200: gradient calculation unit
300: normalization unit
320: gradient summing section
340: gradient size calculation unit
360: gradient size summing section
380: subpixel displacement calculation unit
400: subpixel line position determining unit
500: line connection
510: distance calculation unit
520: angle calculation unit
530: angle difference calculation unit
540: connection distance calculation unit
550: line point determination unit

Claims (18)

A subpixel line position determination method for determining a position of a line point corresponding to the line at a subpixel level from a position of a line pixel determined at a pixel level related to a line in an image,
Calculating gradients related to a pixel value difference between the adjacent pixels for pixels including the line pixel and adjacent pixels adjacent to the line pixel, by a gradient calculator;
Normalizing, by a normalizing unit, the gradients of the pixels based on the sizes of the gradients of the pixels to calculate the sizes of the gradients of the pixels and calculate the subpixel level displacement with respect to the position of the line pixel; And
And determining, by a subpixel line position determining unit, a position of the line point at a subpixel level based on the position of the line pixel and the subpixel level displacement. The method of claim 1,
The step of determining the position of the line point,
And determining the position of the line point as a sub-pixel level by summing the position of the line pixel and the sub-pixel level displacement. The method of claim 1,
Normalizing the gradients of the pixels,
Calculating a sum gradient by summing the gradients of the pixels by a gradient summing unit;
Calculating sizes of gradients of the pixels by a gradient size calculator;
Calculating a summed gradient size by summing the sizes of the gradients of the pixels by a gradient size summation unit; And
And calculating the subpixel level displacement by normalizing the sum gradient based on the summed gradient size by a subpixel displacement calculator. The method of claim 3,
Further comprising, by a window setting unit, determining pixels for determining the position of the line point from the position of the line pixel in the image,
The determining of the pixels includes setting a window having a (2K + 1) × (2K + 1) pixel size around the line pixel, and determining pixels within the window, wherein K is a positive A method of determining subpixel line position, which is an integer. The method of claim 4,
The step of calculating the summation gradient,
Calculating a row direction sum gradient by summing the row direction gradients of the pixels by a row direction gradient summing unit; And
Comprising a column direction summation gradient by summing the column direction gradients of the pixels by a column direction gradient summing unit,
In the calculating of the sizes of the gradients, the sizes of the row direction gradients and the column direction gradients are calculated,
In the calculating of the summed gradient size, the summed gradient size is calculated based on the sizes of the row direction gradients and the column direction gradients,
The step of calculating the subpixel level displacement,
Calculating a subpixel level displacement in a row direction by dividing the summed gradient in the row direction by the summed gradient size by a row direction displacement calculator; And
Comprising a column direction subpixel level displacement by dividing the column direction sum gradient by the gradient sum size by a column direction displacement calculation unit,
The step of determining the position of the line point,
Determining a row direction position of the line point as a subpixel level by summing the row direction position of the line pixel and the row direction subpixel level displacement; And
And determining the position in the column direction of the line point as a sub-pixel level by summing the position in the column direction of the line pixel and the level displacement in the column direction. The method of claim 5,
Normalizing the gradients of the pixels,
When the state of the line pixel is a ridge, normalizing the gradients of the pixels by applying a first sign to the sum of the gradients; And
And normalizing the gradients of the pixels by applying a second sign different from the first sign to the sum of the gradients when the state of the line pixel is a valley. The method of claim 6,
In the calculating of the summation gradient, the row direction summation gradient and the column direction summation gradient are calculated according to Equation 1 below,
In the calculating of the summed gradient size, the summed gradient size was calculated according to Equation 2 below,
In the calculating of the subpixel level displacement, the subpixel level displacement is calculated according to Equations 3 and 4 below,
The step of determining the position of the line point is to determine the position of the line point according to Equation 5 below,
[Equation 1]

[Equation 2]

[Equation 3]

[Equation 4]

[Equation 5]

In Equations 1 to 5,

Is the sum gradient in the row direction,

Is the summation gradient in the column direction,

Is the row direction gradient,

Is the column direction gradient,

Is the sum of the gradients,

Is information indicating whether the state of the line pixel is ridge or valley,

Is the subpixel level displacement in the row direction,

Is the column direction subpixel level displacement, c is the column direction coordinate of the line pixel, r is the row direction coordinate of the line pixel,

Is the row direction position of the line point,

Is a method of determining a subpixel line position indicating a position of the line point in the column direction. The method of claim 1,
The step of connecting the line points to generate a line in the image by a line connector, further comprising,
The step of connecting the line points,
Based on the coordinate distances between the first line point included in the line and the second line points in the vicinity of the first line point among the line points, and the angles between the first line point and the second line point And determining a line point to be connected to the first line point among the second line points. The method of claim 8,
The step of connecting the line points,
Calculating a coordinate distance between a first line point included in the line and a second line point near the first line point among the line points, by a distance calculator;
Calculating a first angle of the line at the first line point by an angle calculator, and calculating a second angle between the first line point and the second line point;
Calculating an angle difference between the first angle and the second angle by an angle difference calculator;
Calculating, by a connection distance calculator, a connection distance between the first line point and the second line point based on the coordinate distance and the angle difference; And
And determining, by a line point determining unit, a line point having a smallest connection distance to the first line point among the second line points and connecting the line point to the first line point. The method of claim 9,
In the calculating of the coordinate distance, the coordinate distance is calculated according to Equation 6 below,
In the calculating of the second angle, the second angle is calculated according to Equation 7 below,
In the calculating of the angle difference, the angle difference is calculated according to Equation 8 below,
In the calculating of the connection distance, the connection distance is calculated according to Equations 9 and 10 below,
[Equation 6]

[Equation 7]

[Equation 8]

[Equation 9]

[Equation 10]

In Equations 6 to 10,

Is the coordinate distance,

Is the row direction coordinate of the first line point,

Is the column direction coordinate of the first line point,

Is the row direction coordinate of the second line point,

Is the column direction coordinate of the second line point,

Is the angle difference,

Is the first angle,

Is the second angle,

Is the set angle difference reflection factor,

Is the angular distance determined according to the angular difference,

Is a method of determining a subpixel line position indicating the connection distance. A computer-readable recording medium on which a program for executing the subpixel line positioning method according to any one of claims 1 to 10 is recorded. A subpixel line position determining device for determining a position of a line point corresponding to the line at a subpixel level from a position of a line pixel determined at a pixel level related to a line in an image,
A gradient calculator configured to calculate gradients related to a pixel value difference between adjacent pixels for pixels including the line pixel and adjacent pixels adjacent to the line pixel;
A normalization unit that normalizes the gradients of the pixels based on the sizes of the gradients of the pixels to calculate the sizes of the gradients of the pixels and to calculate the subpixel level displacement with respect to the position of the line pixel; And
And a subpixel line position determining unit configured to determine a position of the line point at a subpixel level based on the position of the line pixel and the subpixel level displacement. The method of claim 12,
The normalization unit,
A gradient summing unit for calculating a sum gradient by summing the gradients of the pixels;
A gradient size calculator that calculates the sizes of the gradients of the pixels;
A gradient size summing unit for calculating a summed gradient size by summing the sizes of the gradients of the pixels; And
And a subpixel displacement calculator configured to calculate the subpixel level displacement by normalizing the summed gradient based on the summed gradient size,
The subpixel line position determining unit,
A subpixel line position determining apparatus configured to determine the position of the line point as a subpixel level by summing the position of the line pixel and the subpixel level displacement. The method of claim 13,
Further comprising a window setting unit for determining pixels for determining the position of the line point from the position of the line pixel in the image,
The window setting unit sets a window having a (2K + 1) × (2K + 1) pixel size around the line pixel, determines pixels in the window, and determines the subpixel line position where K is a positive integer. Device. The method of claim 14,
The gradient summing unit,
A row direction gradient summing unit for calculating a row direction sum gradient by summing the row direction gradients of the pixels; And
A column direction gradient summing unit configured to calculate a column direction sum gradient by summing the column direction gradients of the pixels,
The gradient size calculator calculates sizes of the row direction gradients and the column direction gradients,
The gradient size summation unit calculates the summed gradient size based on the sizes of the row direction gradients and the column direction gradients,
The subpixel displacement calculating unit,
A row direction displacement calculator configured to calculate a row direction subpixel level displacement by dividing the row direction sum gradient by the gradient sum size; And
A column direction displacement calculator configured to calculate a column direction subpixel level displacement by dividing the column direction sum gradient by the gradient sum size,
The subpixel line position determining unit,
Determining a row direction position of the line point as a subpixel level by summing the row direction position of the line pixel and the row direction subpixel level displacement; And
A subpixel line position determining device configured to determine a position in the column direction of the line point as a subpixel level by summing the position in the column direction of the line pixel and the level displacement in the column direction. The method of claim 15,
The normalization unit,
When the state of the line pixel is ridge, normalizing the gradients of the pixels by applying a first sign to the sum of the gradients; And
When the state of the line pixel is a valley, the subpixel line positioning apparatus normalizes the gradients of the pixels by applying a second sign different from the first sign to the sum of the gradients. The method of claim 12,
Further comprising a line connector connecting the line points to generate a line in the image,
The line connection part,
Based on the coordinate distances between the first line point included in the line and the second line points in the vicinity of the first line point among the line points, and the angles between the first line point and the second line point A subpixel line position determining device configured to determine a line point to be connected to the first line point among the second line points. The method of claim 17,
The line connection part,
A distance calculator configured to calculate a coordinate distance between a first line point included in the line and a second line point in the vicinity of the first line point among the line points;
An angle calculator configured to calculate a first angle of the line at the first line point and to calculate a second angle between the first line point and the second line point;
An angle difference calculator configured to calculate an angle difference between the first angle and the second angle;
A connection distance calculator configured to calculate a connection distance between the first line point and the second line point based on the coordinate distance and the angle difference; And
And a line point determining unit connecting a line point having a smallest connection distance to the first line point among the second line points with the first line point.

Images