JP2683026B2 - Line figure vectorization device - Google Patents

Description

Description: TECHNICAL FIELD The present invention converts a binary line figure image read by a digital scanner into vector data and makes it suitable for processing of CAD, computer, etc. Character recognition, drawing recognition, etc. The present invention relates to a line graphic vectorization device.

2. Description of the Related Art Generally, when various line drawing images such as maps, drawings, and characters are handled by a computer or the like, if binary image data read by a digital scanner is used as it is, not only the memory becomes huge but also the enlargement It is inconvenient for graphic editing including reduction. In this respect, if such binary image data is converted into a combination of vector data regulated at the end and the end,
This is convenient because the memory capacity can be reduced and the graphic editing operation is easy. More specifically, it can be used as an input device for a digital scanner output CAD or the like.

From such a point of view, it has been conventionally considered to vectorize a line figure. Here, there are a thinning method and a skeleton method using a contour vector as a conversion method for converting binary image data of a line figure into vector data, which is generally well known in the past.

That is, the thinning method is a method in which the center line is extracted by narrowing the width one by one so that the line width becomes one. This does not change the connectivity of the original figures. That is, the figure is not cut or a hole is not formed. However, it takes a long time to process and requires a large amount of memory.
Also, as shown in FIG. 16 showing an example of thinning, the original core wire 1
In addition to this, there is a drawback that short beards 2 and loops 3 are easily generated.

On the other hand, a skeletonization method using a contour vector is known, for example, in Japanese Patent Laid-Open Nos. 62-286176 and 62-286177, and after the contour of an image is tracked and chain-coded. , This is a vectorization method, a vector forming a line forming a line is extracted from the vector, and a core line is formed by obtaining a center line of these pair vectors. According to this method, software can perform high-speed and flexible processing. However, as shown in FIG. 17 showing an example of this method, in addition to the original core line 4, vector separation at a branching portion as shown by reference numeral 5 and lack of a short vector as shown by reference numeral 6 Occurs, and the connectivity of the original figure is changed.

The present invention has been made in view of such a point, by using the core line forming method by the contour vector, without the processing time unlike the thinning processing method, and,
An algorithm can be simplified, and a line segment vectorization device that can accurately vectorize one line segment is not divided into several vectors as much as possible without impairing the connectivity of the original graphic. With the goal.

Structure In order to achieve the above object, the present invention first performs linear approximation by converting a contour portion of binary image data obtained by reading a line figure image with a digital scanner into chain code data and then performing linear approximation. Assuming a line figure vectorization device by a skeletonization method using a contour line vector, which obtains skeleton vector data from the contour vector data, first,
A bending point detecting means for detecting a bending point which is an end point or a branch point based on an angle between the continuous contour line vectors is provided, and the contour line vector group between the continuous bending points detected by the bending point detecting means is made into one element. And a single figure extracting means for extracting a single figure element based on the bending point data from the contour element shaped by the contour shaping means. Provided is a single figure skeletonizing means for skeletonizing the single figure extracted by the single figure extracting means.

Alternatively, a loop figure extracting means for extracting a loop figure element from the contour element shaped by the contour shaping means on the basis of the bending point data is provided, and a loop figure skeleton for converting the loop figure extracted by the loop figure extracting means into a skeleton vector Provide means.

Furthermore, a single figure extracting means for extracting a single figure element from the contour element shaped by the contour shaping means based on the bending point data and a loop figure extracting means for extracting a loop figure element are provided, and the single figure extracting means is provided. A single figure skeletonizing means for converting the extracted single figure into a skeleton vector and a loop figure skeletonizing means for changing the loop figure into a skeleton vector are provided, and the contour element shaped by the contour shaping means is a single figure element and a loop figure element. A graphic dividing means for dividing into and is provided.

An embodiment of the present invention will be described below with reference to FIGS. 1 to 15.

First, important terms used in the present invention will be defined.

The "line figure" means a figure written with a line width within a certain range, such as characters and drawings. For example, FIG.
It is as shown in (c).

The "single figure" is represented by one line, and is a line figure having endpoints at both ends as shown in FIG.

The “loop figure” means a line figure forming a closed loop as shown in FIG.

The combination of these as shown in FIG. 2 (c) is referred to as "other figure" in the line figure.

The configuration and schematic processing for line drawing vectorization according to this embodiment will be described with reference to FIG. First, the binary line image data 10a read by a digital scanner (not shown) and stored in the memory 10 is input to the contour line tracing means 11, and the contour portion is traced to the contour line chain code data. Converted to 11a. The chain code data 11a of the contour line is input to the straight line approximation means 12, is vectorized, and is converted into the contour line vector data 12a. The contour line vector data is input to the bending point detecting means 13, the bending point is detected from the angle formed by the continuous vectors, and it is determined from the property of the bending point whether the bending point is an end point or a branch point. Then, the end point data 13a and the branch point data 13b are output.
Next, the processing by the contour shaping means 14 is performed. This contour shaping means 14 uses the contour line vector data 12a output by the straight line approximation means 12, the end point data 13a output by the bending point detection means 13, and the branch point data 13b to form a contour line vector group between consecutive bending points. Those that can be integrated into one vector within a certain error range are integrated and determined as straight lines, and those that cannot be integrated are determined as curves, and the contour element data 14a is output. As a result, the contour line is shaped. The contour element data 14a, the end point data 13a, and the branch point data 13b are used to perform the skeletonization by the skeletonization means 15, and the skeleton vector data 15a is output. Finally,
The skeleton vector shaping means 16 is made to input the skeleton vector data 15a, the end points of the vector within a certain minute range are integrated into one, and the skeleton vector data is corrected to obtain the vector data.
You get 16a.

Here, there is a particular feature in the configuration and processing of the skeletonizing means 15, and this skeletonizing means 15 has, depending on its function, a single figure extracting means 21, a single figure skeletonizing means 22, a contour element correcting means 23, a loop figure extracting means. 24, a loop figure skeletonizing means 25 and a figure dividing means 26.

First, the single figure extracting means 21 searches for a single figure in the contour element data 14a, and if there is a single figure, the single figure skeletonizing means 22
The skeletonization is performed on the single figure and the skeleton vector data 15a is output. Then, the contour element correction means 2
If not, the process proceeds to the process by the loop figure extracting means 24. The loop figure extracting means 24 searches the contour element data 14a for a loop figure, and if there is a loop figure, the loop figure skeletonizing means 25 performs skeletonization on the loop figure and outputs skeleton vector data 15a. After that, similarly, the processing proceeds to the processing by the contour element correction means 23. After that, if the data end has not been reached, the process proceeds to the processing of the figure dividing means 26. The figure dividing means 26 examines the contour element data 15a, divides a figure not extracted as a single figure or a loop figure into a short figure or a loop figure, and corrects the contour element data 15a, the end point data 13a and the branch point data 13b, The processing after the single figure extracting means 21 or the loop figure extracting means 24 is similarly repeated. In the contour element correcting means 23, the contour element connected by the branch point with the contour element cored by the single figure skeletonizing means 22 or the loop figure skeletonizing means 25 is extended to the skeleton line, and as an end point, The contour element data and the end point data are corrected. Such processing is repeated until all the contour element data 14a have been processed.

Next, the processing by each means will be specifically described.
The processing from contour line tracing to contour shaping is based on the method disclosed in Japanese Patent Laid-Open No. 62-2861776.

(A) Contour line tracing method The binary image data input and stored in the memory 10 is raster-scanned to find contour pixels which are not yet traced, and the third
As shown in the figure, the outer contour is counterclockwise (direction A),
The inner contour is tracked in the clockwise direction (direction B), and when returning to the tracking start point P ₀ , the chain code data 11a between them is output as the end of one contour loop. Such tracking processing is executed until there are no contour pixels that have not been tracked yet.

(B) Straight Line Approximation Method A certain chain code C as shown in FIG. 4 obtained by the contour tracing means 11 is used as a threshold value (allowable error d).
Vectorize to be inside. In FIG. 4, V ₁ represents the vector after linear approximation. A straight line of such vector V ₁ becomes one contour line vector V ₂ .

(C) Bending point detection method As shown in FIG. 5, if the absolute value of the angle θ ₁ between the continuous contour line vectors V ₂₁ and V ₂₂ is greater than a certain threshold value, both contour line vectors V ₂₁ and V ₂₂ It is considered that the space is not a straight line but a bending point P ₁ . Here, even if thickness is bending point P _1, there as in the case of end point branch points, contour vector directed in the vicinity of point P ₂ (bending point P ₁ with respect to the bending point P ₁ located as shown in Figure 6 For example, it is determined in a clockwise direction with respect to V ₂₁ that a pixel at a point assumed to be at an angle θ ₂ with the contour line vector V ₂₂ forming a pair) is black or white. For example, if the black pixel area is shown by hatching in FIG. 6, and if the neighboring point P ₂ is a white image as shown in FIG. 6A, the bending point is the end point P _11. judge. On the other hand, if the neighborhood points P ₂ as shown in FIG. 9 (b) is a black image, the bending point is a branch point P ₁₂
Is determined.

Here, when the line width of the line image is thin or when the reading density of the digital scanner is low, only one end point P ₁₁ is detected at one end of the line as shown in FIG. 7 (a). However, if not, two end points P ₁₁ are detected at one end of the line as shown in FIG. 7 (b).

(C) Contour shaping method As shown in FIG. 8 (a), a contour vector group between consecutive bending points that can be linearized within a certain threshold is linearized as shown in FIG. 8 (b). That is, (integrating into one vector), and those that cannot be linearized are registered as contour element data 14a as arcs. Further, when contour line vectors V ₂₁ , V ₂₃ , and V ₂₂ that bend by sandwiching a small vector V ₂₃ as shown in FIG. 9A are searched for,
The small vector V ₂₃ is ignored, and the intersection point of the vectors V ₂₁ and V ₂₂ before and after the small vector is newly set as a bending point P ₁ so that the contour line vectors V ₂₁ and V ₂₂ shown in FIG. 9B are obtained. Divide the contour element data. This is the same even in the case of the arc shape as shown in FIG. 10 (a), the contour is shaped as shown in FIG. 10 (b), and a new bending point is set.

(E) Single figure extraction method and skeletonization method First, as shown in FIG. 11 (a), a certain end point is set as a tracking start end point _P11A, and the data of the contour element E is sequentially tracked from this end point _P11A. If it arrives at the other end point P _11B , and then proceeds from the other end point P _11C on the other side in a direction substantially opposite to the outward path and returns to the tracking start end point P _11A via the other end point P _11D on the start side,
The element is determined to be a continuous line segment having endpoints at both ends, and extracted as a single figure. D in FIG. 11 (b) shows the extracted single figure element. Then, a vector passing through the center is obtained from such a single figure element D and set as a core line vector V ₃ . Even if there is a branch point P ₁₂ as shown in FIG. 11 in the course of such a tracking, it goes straight in the pursuing direction until then, searches for the next branch point P _12, and finds this branch point P ₁₂ If there is a contour element E in the same direction as the previous traveling direction or in the opposite direction (that is, a non-intersecting direction, that is, rightward or leftward in the example of FIG. 11), the tracing is continued. For example, in the case of FIG. ₁₁ , there are six branch points P ₁₂ between the tracking start end point P _11A and the other end point P _11B , but these branch points P ₁₂
With the presence of, the tracing is not divided and the horizontal line portion is extracted as one continuous single figure D.

That is, this FIG. 11 is an example assuming the processing of the central portion of the graphic as illustrated in FIG. 17, and originally, one line image is converted into a core vector without being divided at the branch point P _12. .

(F) Loop figure extracting method and skeletonizing method First, as shown in FIG. 12 (a), a certain end point is set as a tracking start end point P _11A, and the data of the contour element E is sequentially traced from this end point P _11A to 1 Vector groups forming two closed loops,
An outer closed loop or an inner closed loop parallel to the vector group within a certain range (within the allowable line width) is detected. Of these closed loops, the part divided at the branch point P ₁₂ is connected to form a complete double closed loop (to be an extracted loop figure) R ₁ and R ₂ as shown in FIG. 12 (b). After forming, a parallel vector is detected from these double closed loops R ₁ and R ₂ and a vector passing through the center is defined as a core vector V ₄ . This is executed for one round of the closed loop to convert the loop figure into a core line vector.

That is, also in the extraction of the loop figure, the processing is performed without being divided at the branch point P ₁ .

(G) Method of dividing a figure A figure that is neither a single figure nor a loop figure (see Fig. 2 (c)) is divided into several single figures or a single figure and a loop figure. The procedure is to first find a figure with an end point on one side, follow its contour, and reach the branch point,
Therefore, the figure is divided, and a single figure is created by using that as the other end point. If this is executed until there is no figure having an end point on one side, the figure shown in FIG. 13 (a) can be divided into the loop figure of FIG. 13 (b) + the single figure of FIG. 13 (c). . Further, by holding information on such divided portions and connecting them after core-forming based on the information, the connection processing can be facilitated.

Next, processing examples 1 and 2 of practical core line vectorization based on such a configuration and method will be described. Figures 14 and 15
In the figure, the arrow indicating the direction of the vector and the branch point are omitted.

A. Processing Example 1 First, based on the data of the contour element E and the data of the end point P ₁₁ and the branch point P ₁₂ as shown in FIG. To extract a single figure. D shows the extracted single figure. This is the core line vector
As shown by V ₃ , the core line is vectorized. Next, the contour element correction process shown in FIG.
The contour element E connected by P ₁₂ is the core vector V ₃
Extend to the line and use it as the end point P ₁₁ . Then, as shown in FIG. 7D, a single figure is extracted again from this portion to be a core line vector. In the subsequent modification of the contour element, no branch point was found in the extracted single figure, so no modification is performed in this case. Then, the single figure is searched again, but since there is no single figure in this processing example, the loop figure extracting means 24 searches for the loop figure this time. Therefore, as shown in FIG. 14 (e), the core line vector is converted into the core line vector V ₄ based on the extracted loop figures R ₁ and R ₃ . In this case as well, in this processing example, since there is no branch point in the extracted loop graphic in the subsequent contour element correction, the correction is not performed.
With the processing up to (e) in the figure, since there are no unprocessed contour elements in this processing example, shaping processing is performed on the core line vectors V ₃ and V ₄ as shown in (f) in the figure, and only the vector results are obtained. Is obtained, and the process ends.

B. Processing example 2 In the case of a figure composed of the contour element E as shown in FIG.
Even if the single figure extracting process and the loop figure extracting process are sequentially executed, neither the single figure nor the loop figure is found. Therefore, the figure dividing means 26 performs figure dividing processing to divide into two single figures and one loop figure as shown in FIG. P ₄ is the end point. Next, the core vectorization process is performed on the divided single figure as shown in FIG. In this case, since the single figure has no branch point, the contour element is not corrected as post-processing. After that, as shown in FIG. 3D, a core vectorization process is performed on the divided loop figure. In this case, there is a branch point P _{12 in the} loop figure, so the same figure (e)
As shown in, the contour element E of that portion is extended to the core vector V ₄ and a new end point P ₁₂ is obtained. This allows
As shown in (f) of the same figure, a new single figure is newly extracted, and a core line vectorization process is performed on this single figure. This processing eliminates the unprocessed contour element E, so that FIG.
As shown in, the core line vector shaping process such as the connection between the core line vectors V ₃ and V ₄ is performed, and the series of processes is completed.

In the present embodiment, a pair of the single figure extracting means 21 and the single figure skeletonizing means 22 and a pair of the loop figure extracting means 24 and the loop figure skeletonizing means 25 are provided as an apparatus.
These are installed in separate devices to make a single figure core line dedicated machine and a loop figure core line dedicated machine, and each single figure is extracted from any line figure, or only loop figure is extracted and each core line is made. The processing may be performed, and the composition processing may be performed later.

Effect Since the present invention is based on the skeletonization method by the contour line vector as described above, it is possible to solve the drawbacks such as the processing speed of the thinning method being slow and the algorithm being complicated, as well as the conventional contouring method. Also in the comparison with the core line method by the line vector, paying attention to the end point and the branch point in the bending point, the single figure extracting means or the loop figure extracting means which does not interrupt the processing at the branch point is provided, and the result is extracted. Since it is configured to convert the skeleton vector for a single figure or a loop figure, it is highly possible that one continuous line image can be skeletonized without being divided in the middle, and vector combining processing near the branch point is omitted. Since it is possible to reduce vector separation, short vector loss and misalignment of core lines, it does not take much processing time. Even in the case of division, the data of the connection destination can be obtained at the time of division, so that the data can be easily recombined and any line figure can be dealt with.

[Brief description of the drawings]

1 to 15 show an embodiment of the present invention, FIG. 1 is a block diagram, FIG. 2 is an explanatory diagram showing definition of a figure,
FIG. 3 is an explanatory diagram showing a contour line tracking process, FIG. 4 is an explanatory diagram showing a linear approximation process, FIG. 5 is an explanatory diagram showing a bending point detection method, and FIG. 6 is a discrimination process between an end point and a branch point. , FIG. 7 is an explanatory diagram showing an example of the shape of an end point, FIG. 8 is an explanatory diagram showing contour shaping processing, FIG. 9 is an explanatory diagram showing contour shaping and bending point detection processing, and FIG. 10 is Explanatory diagram showing contour shaping and bending point detection processing, FIG. 11 is an explanatory diagram showing single figure extraction and its skeletonization processing, FIG. 12 is an explanatory diagram showing loop figure extraction and its skeletonization processing, FIG. 13 is FIG. 14 is an explanatory diagram showing a graphic division process, FIG. 14 is an explanatory diagram showing a processing example 1 in order, FIG. 15 is an explanatory diagram showing a processing example 2 in order, and FIG. 16 is an explanatory diagram showing a conventional thinning method, The figure is an explanatory view showing a conventional skeletonization method using a contour line vector. 13 ... Bending point detecting means, 14 ... Contour shaping means, 21 ... Single figure extracting means, 22 ... Single figure skeletonizing means, 24 ... Loop figure extracting means, 25 ... Loop figure skeletonizing means, 26 ...... Figure dividing means, 10a ...... binary image data, 11a ...... chain code data, 12a ...... contour line vector data, 13a ......
Endpoint data, 13b ...... branching point data, 14a ...... outline element data, 15a ...... core vector data, P ₁ ...... bending point, P ₁₁
…… End point, P ₁₂ …… Branch point, V ₂ …… Contour line vector, V ₃ ,
V ₄ …… Core line vector

Claims (3)

(57) [Claims] 1. A contour image of binary image data obtained by reading a line figure image with a digital scanner is converted into chain code data, and then linear approximation is performed. Core line vector data is obtained from the linear approximation contour vector data. In the obtained line figure vectorization device, a bending point detecting means for detecting a bending point which is an end point or a branch point based on an angle between the continuous contour line vectors is provided, and between the continuous bending points detected by the bending point detecting means. A contour shaping means for shaping the contour vector by determining whether the contour vector group is one element and determining whether it is a straight line or a circular arc is provided, and a single figure element is formed from the contour element shaped by the contour shaping means based on the bending point data. And a single figure skeletonizing means for converting the single figure extracted by the single figure extracting means into a skeleton vector. Line drawing vectorization and wherein the. 2. A contour line portion of binary image data obtained by reading a line figure image by a digital scanner is converted into chain code data, and then linear approximation is performed. Core line vector data is obtained from the linearly approximated contour line vector data. In the obtained line figure vectorization device, a bending point detecting means for detecting a bending point which is an end point or a branch point based on an angle between the continuous contour line vectors is provided, and between the continuous bending points detected by the bending point detecting means. A contour shaping means for shaping the contour vector by judging whether the contour vector group is one element and determining whether it is a straight line or a circular arc is provided, and a loop graphic element is formed from the contour element shaped by this contour shaping means based on the bending point data. A loop figure extracting means for extracting the loop figure is provided, and the loop figure extracted by the loop figure extracting means is converted into a core vector. Line drawing vectorization apparatus characterized in that a form core means. 3. A core portion of the binary image data obtained by reading a line figure image with a digital scanner is converted into chain code data and then linearly approximated. Core line vector data is obtained from the linearly approximated contour vector data. In the obtained line figure vectorization device, a bending point detecting means for detecting a bending point which is an end point or a branch point based on an angle between the continuous contour line vectors is provided, and between the continuous bending points detected by the bending point detecting means. A contour shaping means for shaping the contour vector by determining whether the contour vector group is one element and determining whether it is a straight line or a circular arc is provided, and a single figure element is formed from the contour element shaped by the contour shaping means based on the bending point data. And a loop figure element based on the bending point data from the contour element shaped by the contour shaping means. A loop shape extracting means is provided for,
A single figure skeletonizing means for skeletonizing the single figure extracted by the single figure slicing means and a loop figure skeletonizing means for skeletonizing the loop figure extracted by the loop figure extracting means are provided. A line graphic vectorization device comprising graphic dividing means for dividing a contour element shaped by a shaping means into a single graphic element and a loop graphic element.