KR20120133529A - Dmethod for deciding of lens distortion correction parameters - Google Patents

Abstract

PURPOSE: A method for determining a lens distortion correction parameter is provided to make a corrected image similar to a perfect image taken by a pin hole camera model. CONSTITUTION: A vertical line pixel calculating unit(150) calculates the number of pixels of each vertical line passing a reference point based on the reference point. An up/down position determining unit(160) determines an up position and a down position on left and right sides of the reference point. A distance calculating unit(170) calculates a distance between the up position and the down position. A distortion rate calculating unit(180) calculates a distortion rate of a vertical direction and a horizontal direction by using a distortion rate function. A main control unit(110) calculates a distortion rate parameter to minimize the distortion rate function. [Reference numerals] (110) Main control unit; (120) Display unit; (130) Horizontal line selecting unit; (140) Reference point determining unit; (150) Vertical line pixel calculating unit; (160) Up&down position determining unit; (170) Distance calculating unit; (180) Distortion rate calculating unit; (190) Image DB

Description

Lens distortion correction parameter determination technique {DMETHOD FOR DECIDING OF LENS DISTORTION CORRECTION PARAMETERS}

The present invention relates to a technique for determining lens distortion correction parameters, and more particularly, to restore photographs or videos taken by concave or convex distortion by low-cost cameras or special cameras using geometrical invariability, and are parallel and vertical to each other. In a template composed of a plurality of horizontal and vertical straight lines, find a reference point of intersection of one horizontal straight line closest to the center of distortion and a plurality of vertical straight lines, and upwardly about the horizontal straight line passing through the center of distortion. Alternatively, calculate the number of pixels of the vertical line that intersects the intersection of the vertical and horizontal straight lines and the horizontal and vertical angles, or calculate the coordinates of the intersection to restore the distorted image to the most ideal image. Lens distortion correction parameter determination technique .

In fact, most cameras cause lens distortion, which depends not only on the price of the camera, but also on a particular application.

To date, many lens distortion correction methods have relied on the invariant intrinsic properties of the projective geometry to find the distortion parameter.

Although the straight line in the background is a common property in the image, previous work has overlooked the condition when the straight lines are parallel to each other in determining lens distortion parameters, and even if the distorted image is corrected due to the overlook of the condition There is a problem that it falls short of the ideal image when no distortion occurs.

Lens distortion caused by general cameras is important in image processing, which is an inevitable physical phenomenon caused by spherical lenses. Detecting and correcting lens distortion is an essential step in achieving the correct content of the image for the following steps.

On the other hand, the following steps produce the final image with minimal deformation from low cost camera lenses.

Models of lens distortion include radial distortion, dispersion distortion and prism distortion.

Radial distortions such as convex or concave distortions are important, but other distortions are acceptable.

The methods applied to radial distortion used three-dimensional reference points to find the distortion parameters of the lens.

These methods faced the challenge of accurately setting up the experiments and accurately extracting control points under the influence of noise.

In another class of lens distortion correction, there are many methods that are limited based on the geometric invariance of some features for obtaining distortion parameters.

Straight lines in the background The straight properties of the lines in the image are widely used.

However, these models overlook these limitations in determining the distortion parameters of the lens when the straight lines are parallel to each other.

This may result in searching for incorrect distortion parameters for the final solution within the search algorithm.

Finally, other approaches have been proposed to correct lens distortion by collecting all the contents of the image to obtain distortion parameters.

These approaches can use image content without requiring any measurement of the pattern. However, the conclusion is still limited to producing satisfactory results.

Accordingly, the present invention includes a lens that includes a condition that can be easily overlooked in the process of correcting a distorted image to determine a parameter capable of minimizing distortion, so that the corrected image can be almost similar to an ideal image taken with a pinhole camera model. It is an object to provide a distortion correction parameter determination technique.

In order to achieve the above object, the lens distortion correction parameter determination technique by the lens correction system according to the present invention comprises the steps of the main control unit of the lens correction system to display a distorted image stored in the image DB on the display, binary display Converting a distorted image of a binary image format to a gray image format, wherein the horizontal selection unit 130 of the lens correction system passes through the center of distortion of the plurality of horizontal lines in the black and white image, or Selecting a near horizontal line, determining a point generated by crossing a plurality of vertical lines on the selected horizontal line by the reference point determining unit of the lens correction system, and each vertical line passing through the reference point by the pixel calculation unit of the lens correction system based on the reference point. Calculating the number of pixels of the lens, determining the up and down positions of the lens correction system Determining an up position and a down position at the left and right points of the reference point, respectively, calculating a distance between the up position and the down position from which the distance calculating unit of the lens correction system is determined, and the distortion rate calculating unit of the lens correction system using a distortion ratio function in a horizontal direction. And calculating a distortion rate in the vertical direction, and calculating a distortion factor parameter that minimizes the distortion function in the main control unit.

The lens distortion correction parameter determination technique according to the present invention has the effect of reconstructing the distorted image closer to the ideal image, thereby producing the same effect as the image taken by a sufficiently expensive camera even with a low cost camera.

1 is a block diagram of a lens distortion correction parameter determination system according to the present invention;
2 is a flowchart of a lens distortion correction parameter determination technique according to the present invention;
3 is a detailed flowchart for calculating a parameter according to the present invention;
4 is a coordinate diagram for explaining radial distortion and tangential distortion occurring between distorted and undistorted points.
5 is a graph showing the distortion ratio of each method after the distortion correction,
6 is a view showing a distorted image and a preserved image under the influence of noise;
7 is a diagram illustrating a distorted image linear method for explaining radial distortion correction, the present invention, and
8 and 9 are comparison diagrams of a radial image and a corrected image photographed by a specific camera.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, terms and words used in the present specification and claims should not be construed as limited to ordinary or dictionary terms, and the inventor should appropriately interpret the concepts of the terms appropriately The present invention should be construed in accordance with the meaning and concept consistent with the technical idea of the present invention.

Therefore, the embodiments described in this specification and the configurations shown in the drawings are merely the most preferred embodiments of the present invention and do not represent all the technical ideas of the present invention. Therefore, It is to be understood that equivalents and modifications are possible.

The lens correction system for determining the lens distortion correction parameter according to the present invention includes a main control unit 110, a display unit 120, a horizontal selection unit 130, an intersection determination unit 140, and a vertical pixel calculation unit 150. do.

First, the main controller 110 displays the image stored in the image DB 190 on the display 120 (S10).

The main controller 110 converts the displayed binary image into a black and white image (S20).

There are a plurality of horizontal lines and a plurality of vertical lines in the binary image, wherein the widths of the horizontal lines and the vertical lines are 1 pixel, and the center of distortion is initialized to the center of the displayed image.

 In the horizontal direction, since the straight line passing through the center of distortion of the plurality of horizontal lines or vertical lines is a straight line even in the distorted image, the horizontal line selecting unit 130 of the lens correction system is closest to the horizontal line or center of distortion passing through the center of distortion. A step of selecting a horizontal line is performed (S30).

The horizontal line selected by the horizontal line selector 130 is called a drive line.

When the horizontal line is selected in step S30, the reference point determination unit 140 performs a step of determining the point where the horizontal line and a plurality of vertical lines intersect as the reference point (S40), wherein the reference point is the number of vertical lines (n) and same.

The horizontal line selected by the horizontal line selecting unit 130 is called a drive line. The reference point determined at the intersection of the drive line and the vertical line in the drive line is equal to the number n of vertical lines.

In the next step, the pixel calculator 150 calculates the number of pixels of each vertical line passing through the reference point based on the reference point (S50). Not located on

The up and down position determiner 160 determines a point in the near range of each reference point, that is, one point to the left of the reference point and one point to the right (S60). Bigger than pixels In the present invention, it is executed by setting `e` to 4.

At this time, the point on the right determined based on the reference point is an up position point, and the point on the left is a down position point.

Thus, each reference point is located in a range between the up position point and the down position point, respectively.

)에서 다운 포지션 포인트(

)까지 각각의 거리를 측정하는 단계(S70)를 수행함으로써 왜곡률 산출부(180)가 수평방향과 수직방향으로의 왜곡률을 산출하는 단계를 수행한다(S80).Distance calculation unit 170 is the up position point (

Down position from

By measuring the respective distances up to S70, the distortion rate calculating unit 180 calculates the distortion rates in the horizontal direction and the vertical direction (S80).

에 의해 산출하는데, 상기 는 아래의 [수학식 1]과 같다.The degree of distortion, i.e., the distortion rate calculator 180 functions as a function of the distortion rate.

Calculated by Is shown in Equation 1 below.

은 수평방향에서의 업 포지션 포인트이고,

은 수평방향에서의 다운 포지션 포인트이며,

은 왜곡 중심을 통과하는 수평선 또는 왜곡중심에서 가장 가까운 수평선과 교차하는 수직선의 개수이다.In [Equation 1]

Is the up position point in the horizontal direction,

Is the down position point in the horizontal direction,

Is the number of vertical lines that cross the horizontal line through the center of distortion or the horizontal line closest to the center of distortion.

에 의해 산출하는데,

는 아래의 [수학식 2]와 같다.On the other hand, the distortion rate calculation unit 180 functions in the same manner as the process of calculating the distortion rate in the horizontal direction also in the vertical direction

Calculated by

Is shown in Equation 2 below.

는 수직방향에서의 업 포지션 포인트이고,

은 수직방향에서의 다운 포지션 포인트이며,

은 왜곡 중심을 통과하는 수직선 또는 왜곡중심에서 가장 가까운 수직선과 교차하는 수평선의 개수이다.In Equation (2)

Is the up position point in the vertical direction,

Is the down position point in the vertical direction,

Is the number of horizontal lines that intersect the vertical line passing through the center of distortion or the nearest vertical line from the center of distortion.

)과 수직방향으로의 왜곡률(

)을 산출하면, 상기 주제어부(110)는 수평방향으로의 왜곡률과 수직방향으로의 왜곡률의 평균 왜곡률을 아래의 [수학식 3]을 통해 계산함으로써 영상 전체의 왜곡률을 계산한다.Therefore, the distortion rate in the horizontal direction calculated by the distortion rate calculation unit 180 (

) And distortion in the vertical direction (

), The main controller 110 calculates the distortion rate of the entire image by calculating the average distortion rate of the distortion rate in the horizontal direction and the distortion rate in the vertical direction through Equation 3 below.

The distortion rate (or error) reflects the farthest distance among pixels of each line (vertical line and horizontal line) in each axis (horizontal axis and vertical axis) at each reference point.

를 최소화시키는 왜곡 파라미터를 산출하는 단계를 수행한다(S90).Finally, the main controller 110 has a distortion function

Computing a distortion parameter to minimize the (S90).

를 최소화시키는 이상적인 경우는 모든 선이 직선이고 서로 평행한 경우임을 참고로 하여 이하에서는, 영상에서의 왜곡을 복원하기 위한 왜곡변수를 최적화한다.Distortion parameter is distortion factor

In the following, the ideal case of minimizing is to optimize the distortion variables for restoring the distortion in the image by referring to the case that all the lines are straight and parallel to each other.

First, before describing the optimization of the distortion variable to restore the distortion in the image, a process of mapping the distorted coordinates to the non-distorted coordinates will be described with reference to FIG. 4.

로부터 물리적으로 측정할 수 없는 왜곡되지 않은 이미지 좌표

로 매핑한다.The standard model for radial distortion and jetted (or tangential) distortion is measured measurable distorted image coordinates.

Undistorted image coordinates that cannot be physically measured from

Map to.

here,

그리고,

는 방사형 외곡의 계수이고,

는 분산형 외곡의 계수이며,

는 위에서

로 정의된 왜곡 중심으로부터 영상 포인트까지의 반경이다.

And,

Is the coefficient of the radial crop,

Is the coefficient of the distributed distortion,

From above

The radius from the center of distortion to the image point, defined by.

In general, in the present invention, only two coefficients of radial distortion are considered and the other coefficients are omitted without considering the following Equation 5 below.

그리고, 왜곡 중심

이 복원되는 것이다.Thus, lens distortion correction is of substantial importance for distortion coefficients.

And distortion center

This will be restored.

In the following description, optimization of distortion variables for restoring distortion in an image will be described.

Nonlinear optimization algorithm is an iterative process, and the focus value plays an important role in reaching the correct value.

If the initial value is too far from the global minima, the converged value can be a local minima, resulting in a false solution. .

있다고 가정하는 단계를 수행하고(S910), 상기 선

에 위치한 모든 점

을 만족시키는 아래의 [수학식 6]으로 표시하는 단계(S920)를 수행한다.In the lens compensating technique according to the present invention, the main control unit 110 is arranged on the distorted image plane.

Assuming that there is a step (S910), the line

All points in

In step S920, the following Equation 6 satisfies is performed.

을 특정하기 위한 상수들이다.In Equation 6, a, b, and c denote slopes of -a / b.

Constants for specifying

상의 각 점은 상기 [수학식 4]에 의해 왜곡된 이미지 평면상에서 점

과 관련이 있다.line

Each point on the image is a point on the image plane distorted by Equation 4 above.

Related to

의 함수들임을 의미한다.This means that all coordinates of the line point are points

Means functions of.

에 위치한 모든 점

을 만족시키는 상기 [수학식 6]을 왜곡된 이미지의 평면상에 위치한 점

의 함수로 변환하는 단계를 수행한다(S930).Thus, the main control unit 110 is a line

All points in

[Equation 6] satisfying the point that is located on the plane of the distorted image

Convert to a function of (S930).

That is, the main control unit 1110 converts Equation 6 as shown in Equation 7 below.

은 왜곡된 영상 평면에서 상응하는 수정 방정식을 의미한다.In [Equation 7] above

Denotes the corresponding correction equation in the distorted image plane.

에 대한 함수

에서 변수의 치환을 통해 아래의 [수학식 8] 과 같이 편도함수식으로 표현하는 단계를 수행한다(S940).On the other hand, the main control unit 110 is any distorted image point

Functions for

In step S940, the expression is expressed as a partial function as shown in Equation 8 below.

Here, all four partial derivatives can be calculated directly from Equation 4.

에서의 접선 즉, 탄젠트(tangent)

의 기울기에 해당된다.Therefore, the slope of the line S (which must be equal to -a / b) in the distorted plane is pointed by Equation 9 below.

Tangent at, or tangent

Corresponds to the slope of.

In this case, the present invention selects a plurality of lines in the image template as reference lines in the linear solution.

과

을

로 표현될 수 있는 선형방정식의 솔루선만으로 추정한다.In addition, the present invention is

and

of

Estimate only the solution of the linear equation that can be expressed as.

의 변수와 계수들은 아래와 같다.Meanwhile, the equation

The variables and coefficients of are as follows.

는 접선의 기울기가

인 직선

상에서 i 번째 왜곡된 점을 표시하고,

은 직선

에 대응되는 선의 추정된 기울기이다.here,

The slope of the tangent

Phosphorus straight

Display the i-th distorted point on

Silver straight

The estimated slope of the line corresponding to.

은 2차 다항식에 의해 2W+1 크기의 영역 내의 커브 포인트의 근사치를 계산함으로써 추정되는데, 본 발명에서 상기 W는 5인 것이 바람직하다.inclination

Is estimated by calculating an approximation of curve points in a region of size 2W + 1 by a second order polynomial, where W is 5 in the present invention.

Hereinafter, to evaluate the lens distortion correction parameter determination technique according to the present invention, correction is performed on the composite image and the actual image.

The distorted composite images in the simulation are generated by distorting the undistorted composite template image with the radial distortion model represented by Equation 10 below using Equation 5 above.

Distortion coefficient k ₂ is fixed at k ₂ = 3.10 ^-13 because it has a weaker effect on distortion compared to k ₁ , while distortion coefficient k ₁ is k = -8.10 ^-7 : -4.8.10 ^-7 for testing Is changed to

In this case, a linear solution, a straight line method, and a lens distortion correction parameter determination technique according to the present invention may be used to restore an undistorted image from the distorted images.

After the distortion correction, the execution of each method is evaluated based on the distortion rate through Equation 3, and the distortion rate of each method is as shown in FIG. 5.

 First, it can be seen from FIG. 5 that the distortion of the linear method is the worst of the three methods.

In addition, when ignoring the parallel condition among the straight lines and considering only the linear condition of each line, FIG. 5 shows that the distortion correction based on the straight line method causes a larger distortion rate than the present invention.

However, by applying the present invention, since the lines are straight and parallel conditions are considered at the same time, the lines are straight and parallel are more strictly guaranteed in the distortion correction.

Nearly all vertical lines (horizontal lines) are coincident with the vertical axes (line horizontal axes) at each reference point.

As shown in Fig. 6, the present invention considers the influence of noise.

 The distorted composite image is altered by Gaussian noise with an average of zero pixels with a standard deviation of 0.2 pixels.

Through the simulations described above, the effect of noise makes the image content according to the invention very wrong compared to the original image content, but the corrected image is almost identical to the ideal image taken with a pinhole camera.

Therefore, the lens correction technique according to the present invention is actually another method for correcting lens distortion under the influence of noise.

Indeed, in order to evaluate the effect of the lens correction technique according to the present invention, Fig. 7 shows the radial distortion correction of a Canon SX20IX camera using the linear method and the lens correction technique according to the present invention.

Before the correction, the distortion ratio is 4.0758, and the distortion ratios after the correction through the linear correction method and the lens correction technique according to the present invention are 0.6821 and 06421, respectively.

 8 and 9 show the radial distortion correction results of the lens correction technique according to the present invention for Canon SX20IS and Canon IXUS 95IS.

In the case of the SX20IS, the distortion correction parameters are k ₁ = 7.7079.10 ^-7 , k ₂ = -1.8791.10 ^-12 , c _u = 329.2941 and c _v = 228.1301.

For the IXUS 95IS, the distortion correction parameters are k ₁ = 5.4129.10 ^-7 , k ₂ = -1.1441.10 ^-12 , c _u = 293.4185, c _v = 221.4925.

As described above, although the present invention has been described by way of limited embodiments and drawings, the present invention is not limited thereto, and the technical idea of the present invention and the following by those of ordinary skill in the art to which the present invention pertains. Of course, various modifications and variations are possible within the scope of equivalents of the claims to be described.

100: lens correction system 110: main control unit
120: display unit 130: horizontal line selection unit
140: reference point determiner 150: pixel calculator
160: up and down position determination unit 170: distance calculation unit
180: distortion rate calculator 190: image DB

Claims (9)

Lens Distortion Correction Parameter Determination by Lens Correction System
(a) the main controller 110 of the lens correction system displaying the distorted image stored in the image DB 190 on the display 120;
(b) converting the distorted image of the binary image format displayed by the main controller 110 into a gray image format;
(c) selecting, by the horizontal line selecting unit 130 of the lens correction system, a horizontal line passing through the distortion center or near the distortion center among the plurality of horizontal lines in the black and white image;
(d) determining, by the reference point determiner 140 of the lens correction system, points generated by crossing a plurality of vertical lines on the horizontal line selected in step (c) as reference points;
(e) calculating, by the pixel calculator 150 of the lens correction system, the number of pixels of each vertical line passing through the reference point based on the reference point;
(f) determining an up position and a down position at the left and right points of the reference point by the up and down position determination unit 160 of the lens correction system;
(g) calculating, by the distance calculator 170 of the lens correction system, a distance between the up position and the down position determined in the step (f);
(h) calculating a distortion rate in a horizontal direction and a vertical direction by the distortion rate calculation unit 180 of the lens correction system; And
and (i) calculating, by the main controller (110), a distortion factor parameter that minimizes the distortion factor function in step (h).
The method of claim 1,
In the step (h), the distortion factor in the horizontal direction is

, Where

Is an up position point in the horizontal direction;

Is a down position point in the horizontal direction; And

Is the number of horizontal lines passing through the center of distortion or the horizontal lines intersecting with the horizontal line nearest to the center of distortion.
The method of claim 1,
In step (h), the distortion factor in the vertical direction is

Where,

Is an up position point in the vertical direction;

Is the down position point in the vertical direction,

Is the number of vertical lines passing through the center of distortion or the horizontal lines intersecting with the vertical lines nearest to the center of distortion.
The method of claim 1,
And a distortion ratio calculator 180 of the lens correction system calculates an average distortion ratio of the distortion ratio in the horizontal direction and the distortion ratio in the vertical direction to calculate the distortion ratio of the entire distorted image.
The method of claim 1,
Step (i) is
(i-1) An arbitrary line on the plane of the distorted image of the main controller 110 (

Hypothesizing;
(i-2) Lines assumed by the main controller 110 in (i-1)

All points at)

Displaying one equation whose slope is -a / b to satisfy?
(i-3) Points on the plane of the image in which the main controller 110 distorts the equation

For

Converting to a function;
(i-4) the main control unit 110 performs the function

Expressing the partial derivatives through substitution of the variables in the equation; And
(i-5) calculating, by the main control unit 110, a distortion function by calculating a partial deviation function.
6. The method of claim 5,
Equation in the step (i-2) is

Where a, b and c represent the slope as -a / b

The lens distortion correction parameter determination technique, characterized in that the constant for specifying.
6. The method of claim 5,
Equation in the step (i-3) is

Lens distortion correction parameter determination technique, characterized in that.
6. The method of claim 5,
The partial derivative formula in step (i-4) is

Lens distortion correction parameter determination technique, characterized in that.
The method of claim 8,
The main control unit 110 in the step (i-4) is the one-way function line (

Lens distortion correction parameter determination technique, characterized in that to calculate the parameter by transforming in the same form as the slope-a / b of

Images