JP2706711B2 - Method and apparatus for correcting positional distortion of image data - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for correcting positional distortion of image data used for an automatic input device for drawing data.

[0002]

2. Description of the Related Art An automatic drawing input device for electronically scanning a drawing with an image scanner or the like and performing vectorization processing on the obtained binary image data and recognition of graphic elements such as characters and symbols is known. Are known. In such an automatic input device, the expansion and contraction of the medium on which the drawing is drawn (for example, paper or mylar paper), the finite resolution of an image scanner,
Alternatively, a minute positional error (distortion) of the graphic data due to the feeding accuracy of the medium or the like occurs throughout the automatically input drawing data.

As a method of correcting such positional distortion, for example, a method using a well-known affine transformation (“Digital image processing technology based on electronics technology concentration basic account”)
The Japan Industrial Technology Center (published in 1983), a method based on the "projection of points" of solid geometry ("Mathematics One-point Sosoku Solid Geometry" published by Kyoritsu Shuppan in 1979), or positional distortion existing in the horizontal or vertical direction A method of correcting in a stepwise manner in a direction perpendicular to the direction of the positional distortion (see
1-253796).

[0004]

When correcting distortion of automatically input drawing data, four points at the top, bottom, left and right of the drawing are used.
It is necessary to match the position of (for example, the four vertices of the drawing frame) with a predetermined position. However, since the first method in the above-described conventional technique is basically a three-point position distortion correction, it is difficult to correct the position of the fourth reference point correctly in the correction of the non-linear distortion included in the drawing data. There is a problem that is.

Further, according to the second method of the prior art, it is possible to correct four-point position distortion, but on the other hand, due to the mechanism of its occurrence, when correcting position distortion that is extremely small compared to the drawing size, it is necessary to perform projection. The reference point may be a point at infinity depending on the calculation accuracy, and correction may not be possible.

Further, a third method in the prior art is basically a method of correcting positional distortion of raster data.
Also, it is effective only when one side of the drawing frame after the automatic input is already horizontal or vertical, the length thereof is equal to the reference value, and the direction of the distortion exists in the direction orthogonal to the drawing frame serving as the reference. There are restrictions. For this reason, 4D for drawing data having a distortion in an arbitrary direction such as after automatic input.
There is a problem that it cannot be applied to point correction. Accordingly, an object of the present invention is to provide a simple position distortion correction method capable of performing four-point correction on drawing data having position distortion in an arbitrary direction such as after automatic input.

[0007]

According to the method for correcting positional distortion of image data according to the present invention, image data to be subjected to positional distortion correction is first subjected to three-point positional distortion correction by affine transformation, and then the three-point position distortion correction is performed. A correction amount required to match one uncorrected vertex of the four vertexes of the rectangle surrounding the image data after the position distortion correction with the corrected vertex is calculated, and the correction amount is calculated by three-point position distortion. A final four-point distortion correction is performed by proportionally distributing the corrected pixels in accordance with the positional relationship between the position and the one vertex.

[0008]

FIG. 3 to FIG. 6 show processing contents of an embodiment of the present invention.
This will be described with reference to the conceptual diagram of FIG. In each drawing, the coordinate origin is assumed to be at the lower left corner of each drawing.

First, as shown in FIG. 3, a quadrangle (rectangle in this example) abcd surrounding image data to be subjected to position distortion correction, for example, drawing data immediately after automatic input is defined. This square abcd may be the drawing frame itself added to the drawing read by an image scanner or the like,
Such drawing data having no drawing frame may be virtual data set so as to surround the whole or a part thereof after automatic input. In the following, for convenience of description, the square abcd is also referred to as a drawing frame. This square ab
cd is distorted due to various undesirable phenomena associated with automatic input. Each vertex of this distorted rectangle abcd is called an "old vertex".

Next, as shown in FIG. 3, a rectangular ABCD surrounding the image data after the positional distortion correction is defined. This square ABCD corresponds to an ideal drawing frame or the like in the case where drawing data is automatically input without distortion. Each vertex of this square ABCD is called a “new vertex”. The (X, Y) coordinates of each of the new and old vertices are represented by a (Xa, Ya), b (Xb, Yb), c (Xc, Y
c), d (Xd, Yd) A (XA, YA), B (XB, YB), C (XC, YC)
), D (XD, YD).

Next, any three of the four old vertices a, b, c, and d are selected, and the affine transformation coefficients required to match each of the three old vertices with the three nearest new vertices are selected. Is calculated. In this example, as shown in FIG. 3, each of the three old vertices a, b, and d is replaced with the three nearest new vertices A,
B and D should be matched. That is,
For any point, the coordinates before affine transformation are (X, Y),
If the coordinates after the affine transformation are (X ₁ , Y ₁ ) and the affine transformation coefficients are α ₀₁ , α ₁₀ , α ₀₀ , β ₀₁ , β ₁₀ , and β ₀₀ , X ₁ = α ₀₁ X + α ₁₀ Y + α _00. .. (1) Y ₁ = β ₀₁ X + β ₁₀ Y + β ₀₀ (2) The six affine transformation coefficients included in the equations (1) and (2), that is, the six unknowns, are obtained by converting three arbitrary points (in this example, three old vertices a, b, and d) before and after the transformation. (X, Y) coordinates can be substituted into each equation to define six equations.

After the six affine transformation coefficients are determined in this manner, the figure frame ab is
The position coordinates of all the elements of the image data in the cd are converted into new position coordinates. In the example of FIG. 3, the position coordinates of any element (X, Y) points indicating the P (X, Y) is the position coordinates after correction (X _1, Y ₁₎ points indicating the P ₁ (X _1, Y ₁ ). Here, the elements constituting the vectorized image data to be subjected to the position distortion correction include a line segment defined by the position coordinates of the start point and the end point, a circle or an arc defined by the position coordinates and the radius of the center point, and a vertex. , Or a position coordinate of the center point of a recognized character or symbol.

As a result of the three-point correction by the affine transformation,
As shown in FIG. 4, three of the four old vertices of the drawing frame match the newest vertex nearest to the corrected drawing frame, but one old vertex C ₁ (XC ₁ , YC ₁ ) remain displaced from the nearest new vertex C (XC, YC). Therefore, the correction amount of the position required to match the old vertex C ₁ after the affine transformation to the new vertex C, that calculates the distance between points C ₁ and C. Next, a final 4 Ten'yugami corrected by prorating according to calculate already correction on a positional relationship between each point and the vertex C ₁ after the affine transformation. As a result of the four-point distortion correction, as illustrated in FIG.
The point P ₁ (X ₁ , Y ₁ ) after the affine transformation is the point P
₂ (X ₂ , Y ₂ ).

A specific example of the four-point position distortion correction based on the proportional distribution of the correction amount will be described with reference to FIGS. In this example, first, the distance (correction amount) between the point C ₁ and the new vertex C is decomposed into its X component (ΔX) and Y component (ΔY). Then, as shown in FIG.
Are proportionally distributed to perform four-point distortion correction in the X direction, and then, as shown in FIG. 5, four-point position distortion correction in the Y direction is performed by proportionally distributing ΔY to each point.

First, a method of proportionally distributing the correction amount ΔX to each point will be described with reference to FIG. First, a region surrounded by the drawing frame ABC ₁ D after three-point position distortion correction by affine transformation is a first region surrounded by a trapezoid AEC ₁ D,
It is divided into two parts: a second area surrounded by a right triangle C ₁ EB. For the first area, pay attention to the right triangle DC ₁ C ₂ , and calculate the X component of the correction amount to be proportionally distributed to the point F having Y ₁ as the Y coordinate on the hypotenuse, that is, the length α of the line segment FG. Determined as the following equation. α = [L (DG) / L (DC ₂ )] ΔX (3) where ΔX is the X component of the correction amount for the point C ₁ , that is, the length of the line segment C ₁ C ₂ . L (DC ₂ ), L
(DG) means the length of the line segments DC ₂ and DG, and the same symbols are used for the lengths of the line segments.

Next, the correction amount δX in the X direction to be proportionally distributed to an arbitrary point P ₁ (X ₁ , Y ₁ ) having the same Y coordinate as the point F in the first area is expressed by the following equation. Determine. δX = [L (G′P ₁ ) / L (G′F)] α (4) In the following, for simplicity of explanation, the sides AD and BC of the square ABCD are set parallel to the X axis. AB and CD with
Is set parallel to the Y axis, XA = XBBXmin, XC = XD≡Xmax YA = YD≡Ymin, YB = YC≡Ymax (5)

[0017] From equation (3) to (5), alpha = [(Y ₁ over Ymin) / (YC ₁ -Ymin)] [Delta] X · · · (6) [delta] X = [(X ₁ over Xmin) / (W −α)] α (7) where W is the length of the side AD, that is, Xmax−Xmin
It is. Normally, the condition that the position distortion rate is sufficiently smaller than 1, that is, the condition of α≪W is satisfied,
From the equation (7), the following equation can be obtained with sufficient approximation accuracy. [delta] X ≒ [(X ₁ over Xmin) / W] alpha = [(X ₁ over Xmin) / (Xmax -Xmin)] × [(Y ₁ over Ymin) / (YC ₁ -Ymin)] [Delta] X · · · (8 )

Next, in the second region surrounded by the right triangle C ₁ BE in FIG.
Focusing on ₁ BJ, the X component β of the correction amount to be proportionally distributed to the point H having the Y coordinate value Y ₁ on the hypotenuse is determined by the following equation. β = [L (HI) / L (C ₁ J)] ΔX = [(Ymax−Y ₁ ) / (Ymax−YC ₁ )] ΔX (9) Further, a point in the second area is set. X to be proportionally distributed to an arbitrary point P ₁ (X ₁ , Y ₁ ) having the same Y coordinate value Y ₁ as H
The direction correction amount δX is determined as in the following equation. δX = [L (H′P ₁ ) / L (H′H)] α (10)

Attention is again paid to the right triangle C ₁ BJ. L (H′H) L (BI) = (XC ₁ −Xmin) [(Ymax−Y ₁ ) / (Ymax−YC ₁ )] .. (11) holds, the following relationship is obtained from the expressions (9) to (11). δX = [(X ₁ −Xmin) / (XC ₁ −Xmin)] ΔX (12)

As is clear from equations (8) and (12),
By proportioning the arbitrary point on the basis of the X component ΔX correction amount of the position distortion for the vertex C ₁ on the positional relationship between any point and the vertex C _1, corrects the positional distortion of the X-direction of the arbitrary point can do.

After completion of the correction of the positional distortion in the X direction, the figure frame becomes a square ABC ₂ D as shown in FIG. Subsequently, the vertex C ₂ is moved in the Y direction so that it matches the new vertex C, and the correction amount ΔY
Is proportionally distributed to each point in the square ABC ₂ D. First, focusing on the right triangle C ₂ BC, the Y component of the correction amount to be proportionally distributed to the point K having the X coordinate value X ₂ on the hypotenuse,
That is, the length γ of the line segment KL is determined as follows. γ = [L (BL) / L (BC)] ΔY = [(X ₂ −Xmin) / (Xmax−Xmin)] ΔY (13)

Subsequently, a correction amount δY in the Y direction to be proportionally distributed to a point P ₂ (X ₂ , Y ₂ ) having the same X coordinate as the point K in the square ABC ₂ D is determined as follows. δY = [L (L′ P ₂ ) / L (L′ K)] γ (14) where L (L′ K) = (Ymax−Ymin) −γ (15) . Normally, the condition that the position distortion rate is sufficiently smaller than 1 is satisfied, that is, the condition of γ≪ (Ymax−Ymin) is satisfied. Therefore, the following equations are obtained from equations (13) to (15) with sufficient approximation accuracy. Is obtained. δY ≒ [(Y ₂ −Y min) / (Y max −Y min)] × [(X ₂ −X min) / (X max −X min)] ΔY (16) As is apparent from the equation (16), the vertex C of the correction amount of the Y component ΔY position distortion for ₂ to vertex C ₂ and X direction 4
By performing proportional distribution to an arbitrary point based on the positional relationship with the arbitrary point after the correction of the point position distortion, it is possible to correct the four-point positional distortion in the Y direction of the arbitrary point.

The contents of the processing described above are summarized in the flowchart of FIG. Further, the processing content described as a specific example regarding step 6 of the flowchart of FIG. 1 is summarized in the flowchart of FIG.

As for the four-point position distortion correction, first, position distortion correction in the X direction is performed for all elements.
Subsequently, the case where the positional distortion correction in the Y direction is performed for all the elements is illustrated. However, the configuration may be such that the positional distortion correction in the Y direction is performed on all the elements first, and then the positional distortion correction in the X direction is performed on all the elements.
Alternatively, the processing of performing the position correction distortion in the X and Y directions for one element and then performing the position correction distortion in the X and Y directions for the next element may be repeated for all the elements.

FIG. 7 shows a typical example of the configuration of an automatic drawing input device incorporating the position distortion correction function of the above embodiment. The automatic drawing input device includes an image scanner 21 that optically scans a drawing to read two-dimensionally arranged dot pattern data, and an image processing device 2 that processes the dot pattern data and converts it into vectorized data.
2 and a workstation 23 for performing the position distortion correction of the above-described embodiment on the vectorized data.

The results of an experiment conducted on the above embodiment using the apparatus shown in FIG. 7 will be described below. The resolution of the dot pattern data read by the image scanner 21 was 16 dots / mm 2, the memory capacity of the image processing device 22 was 16 megabytes, and a personal computer was used as the workstation 23. As a vectorization algorithm, a method based on distance transformation and skeleton extraction was used.

The drawing to be input is A1 size (84
Using an architectural plan of 12 mm x 594 mm), the total number of vectors after vector processing was 4,835. This drawing data is displayed on the display device of the workstation 23, and the coordinates (unit: mm) of the four old vertices of the drawing frame selected by using the mouse are a (14.63, 19.31), b (1
6.31, 560.19), c (796.94, 559.44), d (79
6.75, 17.94). On the other hand, the coordinates of the corrected new vertex input from the keyboard are A (14.63, 19.31),
B (14.63, 559.31), C (794.63, 559.31), D (
794.63, 19.31). First, the three old vertices a, b,
d new vertices A, B, was subjected to only the affine transformation processing for position conversion into D, takes 9 seconds as the total processing time, one of the old coordinates of vertices C ₁ remaining points after conversion (793.13
, 559.94). Next, as a result of performing the affine transformation processing and the four-point correction processing according to the present invention, it took 10 seconds as the total processing time, and all the four old vertices coincided with the four new vertices.

[0028]

As described above in detail, by incorporating the present invention as a function of post-processing software for editing / correcting in, for example, an automatic drawing input device, it is possible to correct four points of distortion contained in drawing data. This makes it possible to improve the efficiency of the correction work and significantly improve the quality of the created drawing data.

[Brief description of the drawings]

FIG. 1 is a flowchart summarizing processing contents of a position distortion correction method according to an embodiment of the present invention.

FIG. 2 is a flowchart summarizing a specific example of the processing content of step 6 in the flowchart of FIG. 1;

FIG. 3 is a conceptual diagram for supplementarily explaining the processing content of the embodiment, and shows a rectangle surrounded by four old vertices a, b, c, and d and four new vertices A, B, C, and An example of a square surrounded by D.

FIG. 4 is a conceptual diagram for supplementarily explaining the processing content of the embodiment, and is an affine transformation so that three old vertices a, b, and d match three new vertices A, B, and D; Is determined, and a three-point-by-one distortion correction is performed based on this.

FIG. 5 is a conceptual diagram for supplementary explanation of the processing content of the embodiment, showing how four-point-one distortion correction is first performed in the X direction.

FIG. 6 is a conceptual diagram for supplementarily explaining the processing content of the embodiment, showing a state in which four-point-one distortion correction is subsequently performed in the Y direction.

FIG. 7 is a block diagram illustrating an example of the configuration of an automatic drawing input device for implementing the correction method according to the embodiment.

[Explanation of symbols]

 21 image scanner 22 image processing device 23 workstation

Claims (4)

(57) [Claims] 1. A position distortion correction method for performing position distortion correction on vectorized two-dimensional image data, comprising: a rectangle surrounding the image data to be subjected to the position distortion correction (each vertex is referred to as an “old vertex”). ) And a rectangle surrounding the image data after the position distortion correction (each vertex thereof is referred to as a “new vertex”), and arbitrary three of the four old vertices are selected, and each is selected as the nearest neighbor. The affine transformation coefficients required to match the three new vertices are calculated , and based on the calculated affine transformation coefficients ,
The distortion correction and facilities elements in 3-point displacement correction of all image data of the object, the fourth rectangle that encloses the image data after the three-point displacement correction of
In order to match one of the vertices that does not match the new vertex to the nearest new vertex, the image data after the correction
The nearest neighbor in the vertical and horizontal directions of the rectangle surrounding the data
Is matched with one vertex that does not match the new vertex of
The distance between the nearest new vertices is calculated, and the vertical and horizontal directions of the square surrounding the corrected image data are calculated.
For each element of the image data after the three-point position distortion correction ,
A position distortion correction method for image data, wherein four-point position distortion correction is performed by proportionally distributing the four points according to the positional relationship with one not- existing vertex. 2. A displacement correction method of the image data according to claim 1, wherein the image data of the displacement correction target comprising the recognized characters and symbols or the one of them. 3. A rectangle surrounding the image data after the positional distortion correction is a rectangle having two sides parallel to the horizontal direction and two sides parallel to the vertical direction on the display screen. 2. The method according to claim 1, wherein the proportional distribution of the correction amount is performed first for the horizontal component and then for the vertical component. 4. A position distortion correction device for performing position distortion correction on vectorized two-dimensional image data, comprising: a rectangle surrounding the position distortion correction target image data (each vertex is referred to as an “old vertex”). ) And a rectangle surrounding the image data after the position distortion correction (each vertex thereof is referred to as a “new vertex”), and arbitrary three of the four old vertices are selected, and each is selected as the nearest neighbor. 3 and coefficient calculating means for calculating an affine transformation coefficients necessary to match the new vertex, the old vertices inside based on the affine transformation coefficients of the calculation already
A three-point position distortion correcting means for performing three-point position distortion correction on all image data elements to be subjected to the distortion correction, and a square 4 surrounding the image data after the three-point position distortion correction.
In order to match one of the vertices that does not match the new vertex to the nearest new vertex, the image data after the correction
The nearest neighbor in the vertical and horizontal directions of the rectangle surrounding the data
Is matched with one vertex that does not match the new vertex of
A distance calculator that calculates the distance between the nearest new vertices.
Steps and vertical and horizontal directions of a square surrounding the corrected image data
For each element of the image data after the three-point position distortion correction ,
Prorated to in accordance with the positional relationship between one vertex not
Proportional distribution means for correcting four-point position distortion.
An apparatus for correcting positional distortion of image data.