JPH04277876A - Method and device for vectorizing linear graphic - Google Patents

Abstract

PURPOSE:To obtain a enough result concerning accuracy as well while suppressing calculation cost low. CONSTITUTION:This device is provided with a main attention point detecting means 5 to sample a contour while moving forward on a contour information train obtained by trace, a sub attention point detecting means 6 to detect a point on the contour closest to the main attention point with a black picture element between as a sub attention point, a discriminating means 8 to discriminate whether a core line is extracted or the position of the main attention point is detected again by comparing a distance between the both attention points with a prescribed threshold value, a core line extracting means 10 to calculate a midpoint between the both attention points as a point expressing the core line based on the discriminated result and to store the point in a core line information storing means 11, and a main attention point estimating means 9 to predict the appearance of branched structure based on the discriminated result, to estimate a branch in a smoothly continuous direction from the extracted core line information and to newly set the main attention point.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a line figure vectorization method and apparatus for converting line figures obtained by quantization using an image scanner or the like, such as raster data of various design drawings, into vector data.

0002

2. Description of the Related Art Conventionally, various methods as described below have been proposed as a process for inputting a drawing input using an image scanner or the like and converting the figure into vector data in which the figure is approximated by a set of short line segments.

The first method is to generate vector data by tracing a thinned image obtained by thinning a figure. The thinning process is a process of repeatedly erasing a figure on a digital image by one pixel from the surrounding black pixels until the line width becomes one pixel. Since a line image with a line width of 1 can be easily traced, approximation processing can be performed during tracing to obtain vector data. For typical algorithms for thinning, see the document “Thin”.
ing algorithms on rectang
ular, hexagonal and tr
angular arrays” (E.S.Deu
tsch, C. ACM, vol. 15, no. 9,
1972, pp. 827-837) and others.

[0004] A second method is to place emphasis on processing speed on software, and after tracing the outline of a figure on a digital image, use data obtained by approximating the outline image with short line segments. In this short line segment data, short vector pairs are detected based on the property that pairs of short vectors that face each other with a black pixel (graphic part) in between have a direction difference of about 180 degrees, and the midpoint position of these short vector pairs is calculated. By doing this, the core lines of the figure are detected. In this case, extremely large calculations are required to search for a pair of short vectors facing each other across a black pixel from a set of approximate short line segments of the contour. A method has been proposed to improve efficiency by avoiding searching in . For details, refer to the document “Drawing processing using multidimensional data structures” (Osawa, Sakauchi, Proceedings of the Institute of Electronics and Communication Engineers (D) J68-
D.No. 4).

Furthermore, as a third method, a method has been reported in which the locus of the center point is detected by moving the largest circular arc that does not extend beyond the black pixel area of the figure within the black pixel area of the figure. Although this method has high accuracy, the calculation cost is extremely high, so a method has been reported in which similar processing is performed using the maximum rectangle instead of the maximum circular arc in order to reduce the calculation cost.

[0006]

[Problems to be Solved by the Invention] In the first conventional method described above, the raster scan of the image must be repeated as many times as the number of layers of pixels to be erased, resulting in an enormous calculation cost and making it impractical to use. In addition to requiring dedicated hardware to process in time, there is a problem in that the corners of the figure become blunt due to the effect of thinning distortion, and the structure of sharp corners is not sufficiently preserved.

Furthermore, the second method has a problem in that the accuracy of the pair vector search based on a huge number of contour line approximation short vectors is not sufficient. In other words, in a search that uses only the distance and mutual directional relationship between contour approximation short vectors as clues, the local unevenness of the contours has a large influence, and pair vectors may not be detected, especially in images with complex structures. This is often the case.

[0008] Compared to the first and second methods, the third method has the following advantages:
Although high accuracy can be obtained, it requires a very large amount of calculation. This is the same even if the largest circle is a rectangle.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above-mentioned problems and provide a line graphic vectorization method and device that generates highly accurate vector data while keeping the amount of calculations relatively small.

0010

[Means for Solving the Problems] The present invention traces all boundaries between white pixels and black pixels of binary raster data obtained by quantizing line figures using an image scanner or the like clockwise or vice versa. Regarding the extracted contour information, when an arbitrary contour pixel is set as the main point of interest and a contour pixel located at the shortest distance from the main point of interest across the figure is set as the sub-point of interest, the corresponding sub-point of interest is moved while moving the main point of interest. In a line figure vectorization method that searches for a point of interest and detects a curve connecting the midpoint of both points of interest as a core line, when the distance between the main point of interest and the sub-point of interest is within a predetermined value, the main point of interest is Both points of interest are detected by moving in the data order of the contour information and moving the sub-point of interest in the reverse order of the data order of the contour information, and if the distance between the two points of interest exceeds a predetermined value, , the position of the next main point of interest is estimated from the approximate polynomial of the detected core line, and a new main point of interest is detected based on the estimation result.

The present invention also provides contour information extracted by tracing all the boundaries between white pixels and black pixels of binary raster data obtained by quantizing a line figure using an image scanner or the like in a clockwise direction or in the opposite direction. , when an arbitrary contour pixel is set as the main point of interest and a contour pixel located at the shortest distance across the figure from the main point of interest is set as the sub-point of interest, the corresponding sub-point of interest is searched for while moving the main point of interest. , a line figure vectorization device configured to detect the midpoint between both points of interest as a core line, comprising an image input means for inputting an image to obtain binary raster data, and an image storage means for storing the raster data. , a contour information detection means for detecting contour information from raster data, a contour information storage means for storing the detected contour information, and a main point of interest for extracting the main points of interest from the contour information at regular intervals in the order in which the data of the contour information is arranged. In order to determine the extraction means and the corresponding sub-point of interest, data is extracted from the contour information storage means in the reverse order of the data arrangement of the contour information, starting from the previous sub-point of interest, and using the distance calculation means. A sub-point of interest detection means detects a point where the distance between both points of interest is the shortest, and a distance between the obtained sub-point of interest and the corresponding main point of interest is compared with a predetermined threshold value. If within, the process is transferred to skeleton extraction means, and if it exceeds a predetermined threshold,
a determining means for transferring the processing to the main point of interest position estimation means; a main point of interest estimating means for estimating the main point of interest from the detected core line according to the result of the determining means; The present invention is characterized by comprising skeleton extraction means for extracting midpoints as points on the skeleton, and skeleton information storage means for storing detected skeleton information.

0012

[Example] Regarding the line figure vectorization method according to claim 1,
This will be explained using figures. FIGS. 2, 3, and 4 are diagrams for explaining the operation of the present invention.

The raster-vector conversion of the present invention uses contour information obtained by tracking all contour black pixels in either a clockwise or counterclockwise direction. This will be explained as follows. The contour information mentioned here is array data in which the coordinate values of contour pixels and the label values of objects to which the pixels belong are arranged in a line in the order in which they are detected by tracking. Furthermore, two pixels on the contour that are the shortest distance apart with a black pixel in between are called a main point of interest and a sub-point of interest. These two points of interest can be detected by sequentially searching for sub-points of interest corresponding to each main point of interest while tracing the contour and detecting the main points of interest at regular intervals. In this method, a method for efficiently performing this process will be described. If both points of interest can be detected, skeleton data can be extracted by calculating the coordinates of their midpoint. Hereinafter, a processing method for a portion where a branch point does not appear and a processing method for a portion where a branch point exists will be explained in order. (Creating skeleton lines in areas with no branching) The process of searching for sub-points of interest is generally very computationally expensive; however, for areas without branching structures, it is possible to
Limit the search range by using the data arrangement order of contour information,
Reduce computational costs. The main point of interest is detected by reading out the contents of the array data of the contour information at regular intervals in the forward direction. In addition, sub-point of interest detection is performed by reading the coordinate value data within a certain range in reverse order from the previous sub-point of interest, calculating the distance to the corresponding main point of interest, and detecting the data that takes the minimum value. conduct.

[0014] In order to specifically show what kind of data is extracted in the above processing, a description will be given using FIG. 2. Here, {a} is a set of contour pixels, and takes a coordinate value a(x,y) as an element. Further, in FIG. 2, the pixels indicated by the symbol ○ are contour pixels, the pixels indicated by the symbol ● are the main points of interest among the contour pixels, and the pixels indicated by the symbol ◎ are the sub-points of interest. The points indicated by the symbol x are a sequence of points indicating the detected skeleton passing positions. ai is the coordinate value of the main point of interest, and aj is the coordinate value of the corresponding sub-point of interest. In the above contour information, ai is the i-th data, and aj is the j-th data. In the figure, contour information is arranged at regular intervals T (
In the figure, the positions of the main points of interest obtained by reading at T=4) are indicated as ai, ai+T, and ai+2T. Here, the value of T is a value determined depending on the resolution of the input figure and image data. Set a value that is approximately slightly longer than the average line width.

Data is read out in order from the i-th data of the contour information at intervals of T, and the coordinate values of the main point of interest ai, ai+T,
When ai+2T is detected, the corresponding sub-points of interest are searched in the reverse order of the outline information data arrangement under the following conditions. <Search Conditions> The kth sub-point of interest from ai is a point that takes the minimum distance dk within the range shown by the following formula.

[0016]

[Math 1]

[0017]

However, if the coordinate values of ai and aj are respectively (
xai, yai), (xaj, yaj),

[Math 2]

[0019]

[0020] aj-mk is contour information and is data that is traced back mk times from aj.

In the section where structures such as branches do not appear, that is, in FIG. 2, three sets of main points of interest and sub-points of interest are sequentially detected.

[0022] The sequence of points representing the core line is calculated using Equation 3 below.

[0023]

[Math 3]

[0024]

However, in FIG. 2, k=0, 1, and 2.

By the above method, distance calculation (Equation 1
) and midpoint position calculation (Equation 3), skeleton lines can be extracted only. (Creating a core line in a part with a branch) Next, processing for a part with a branch will be explained. When the value dk in Equation 1 exceeds a predetermined threshold value D, it is possible to obtain an approximate curve from the sequence of points representing the core line extracted so far and determine the next search direction based on this. be. Thereby, even if there is a branch structure, skeleton extraction can be continued in the direction in which the skeletons extracted so far are smoothly connected. This will be explained in detail using FIGS. 2, 3, and 4.

Assume now that a sequence of points f0, f1, f2 representing skeleton lines has been extracted. If D is a predetermined threshold value, then equation 1 becomes

[Math 4]

[0028]

If the following is satisfied, it can be determined that the kth midpoint approaches a branch point. In FIG. 2, it is assumed that Equation 4 is satisfied when k=3. Extracted point sequence fk, k={0,1,
2} approximate curve

[Math 5]

[0030]

##EQU1## is determined by least squares approximation, and this is defined as the estimated core line F (see FIG. 3).

Next, the point where the approximate curve F intersects with the contour pixel at the end of f2 is detected, the position of the detected contour pixel in the contour information data is determined, and according to the core line processing in the part where there is no branch,
Two points representing the core line

[Math 6]

[0033]

Calculate . This is a temporary core wire,
For verification purposes, the line segment connecting these two points is regarded as a vector, and the following processing is performed.

On the estimated core line F in FIG.

[Math 7]

[0036]

Take [0037],

[Math. 8]

[0038]

[0039] was extracted from the tentative main point of interest.

[Math. 9]

[0040]

It is checked whether [0041] satisfies the following conditional expression (Equation 10).

[0042]

[Math. 10]

[0043]

However, θT is a threshold value of about 0.5. If the conditional expression (Equation 10) is not satisfied, a point that satisfies the condition is searched for by exchanging the main point of interest with the sub-point of interest, temporarily moving the main point of interest to another nearby contour pixel, etc. If the conditional expression is satisfied, the temporary main point of interest and skeleton information are regarded as formal data, and skeleton extraction processing continues. In the example of FIG. 3, skeleton pixel arrays can be detected as shown in FIG.

[0045] With the above method, in a part of a line figure that does not have a branching structure, core lines can be extracted by using the data arrangement order of the contour information data obtained in the contour extraction process and only by calculating distances for a small number of candidates. be. Furthermore, in the branch structure portion, it is possible to extract lines in a direction that continues smoothly beyond the intersecting line segments. Therefore, in terms of processing speed and core line extraction accuracy, which are problems in conventional methods, it is possible to provide a line figure vectorization method and an apparatus therefor that have a good balance between the two.

Next, an embodiment of the line graphic vectorization apparatus according to the second aspect will be described with reference to the drawings. FIG. 1 is a diagram showing an embodiment of a line graphic vectorization device.

This line figure vectorization device includes an image input means 1, an image storage means 2, a contour information detection means 3, a contour information storage means 4, a main point of interest extraction means 5, and a sub point of interest detection means 6.
, a distance calculation means 7, a determination means 8, a main point of interest position estimation means 9, a skeleton extraction means 10, and a skeleton information storage means 11.

The image input means 1 reads line figures on paper and converts them into binary raster data. Shine light onto the paper,
The reflected light is photoelectrically converted by a one-dimensional CCD array element, and A/D
It is configured to obtain digital data by converting it with a converter. The digital data is further binarized using a comparator or the like using an appropriate threshold. The configuration is such that two-dimensional raster data can be obtained by mechanical scanning of the CCD array element. The obtained data is sent to the image storage means 2 and stored.

The image storage means 2 stores the binary raster data obtained by the image input means 1. The image storage means is constituted by a memory.

The contour information detection means 3 reads the raster data from the image storage means 2, scans it in the horizontal and vertical directions, and sets the black pixel at the position where the white pixel changes from a white pixel to a black pixel or from a black pixel to a white pixel as a starting point. , black pixels are tracked clockwise or counterclockwise to detect contour information. Here, the contour information is a sequence of coordinate values of contour pixels labeled for each object. It can be configured by a CPU, a program memory having the processing procedure shown in FIG. 5, a work memory, and the like. It is also possible to configure the processing procedure of FIG. 5 using logic elements. The contour information is written into the contour information storage means 4.

The contour information storage means 4 stores the contour information obtained by the contour information detection means 3. The contour information storage means is constituted by a memory.

The main point of interest extraction means 5 extracts the main point of interest from the contour information. It has an access circuit to the contour information storage means 4, and has a function of detecting the main point of interest therein and outputting the address where the data is stored. The details of the processing are as in the embodiment of claim 1. It is configured as a circuit centered around a CPU, a program memory that stores processing procedures, and a work memory.

The sub-point-of-interest detection means 6 detects the closest contour pixel between the main point of interest and the figure. It has a function of receiving the data address of the main point of interest from the main point of interest detection means 5, and based on this, referring to the coordinate data in the contour information to find the sub-point of interest. A data address where the coordinate value of the pixel having the minimum distance is stored and a detected distance value are output. Details of the processing procedure are based on the embodiment of claim 1. It consists of a CPU, a program memory that stores the above processing procedures, a work memory, etc.

The distance calculating means 7 calculates the distance between the main point of interest and the sub-point of interest. Comparison from sub-point of interest detection means 6 2
It has the function of receiving the coordinate values of two pixels, calculating the distance, and returning the value. It is composed of ALU, etc.

The determining means 8 compares the distance value between the secondary point of interest obtained by the secondary point of interest detection means 6 and the corresponding main point of interest with a threshold value, and determines whether or not they are correct corresponding points. . It has a register that receives and stores the distance value between the main point of interest and the sub-point of interest from the sub-point of interest detection means 6. Also, claim 1
It has a register for storing the threshold D value mentioned in the embodiment, a comparator for comparing both register values, and a process is performed on either the main point of interest position estimating means 9 or the skeleton extracting means 10 according to the output of the comparator. It has a control circuit that moves. This is a necessary means for extracting core lines in a smoothly continuous straight line direction without being affected even if there are line segments that intersect.

The main point of interest position estimating means 9 is activated by the determining means 8 and has a function of accessing the main point of interest extraction means 5 and estimating the next position of the main point of interest. The detailed processing content of this means is as described in the embodiment of claim 1, and is composed of a CPU, a program memory storing the above processing procedure, a work memory, etc.

The skeleton extraction means 10 is also activated by the determination means 8 and calculates the coordinate values of the points through which the skeleton passes. This process calculates the midpoint between the main point of interest and the sub-point of interest. The details are as described in claim 1. It consists of a register that stores coordinate values and an ALU.

The skeleton information storage means 11 includes the skeleton extraction means 1
It is composed of a memory that stores skeleton information calculated as 0.

A system can be realized with the above configuration.

[0060]

According to the present invention, in a part of a line figure that does not have a branching structure, core lines can be extracted by only calculating distances for a small number of candidates by using the data arrangement order of contour information data obtained in contour extraction processing. is possible. Furthermore, in the branch structure portion, core lines can be extracted in a smoothly continuous direction beyond the intersecting line segments. Therefore, with regard to the problems of processing cost and core line extraction accuracy that are problems with conventional methods, software processing can maintain sufficient accuracy for practical use within the computation cost, making it possible to convert line figures into vectors with a good balance between the two. A method and an apparatus thereof can be provided.

[Brief explanation of the drawing]

FIG. 1 is a functional block diagram showing an embodiment of a line graphic vectorization device of the present invention.

FIG. 2 is a diagram for explaining a line graphic vectorization method.

FIG. 3 is a diagram for explaining a line graphic vectorization method.

FIG. 4 is a diagram for explaining a line graphic vectorization method.

FIG. 5 is a flowchart showing a processing procedure for detecting contour information.

[Explanation of symbols]

1 Image input means 2 Image storage means 3 Contour information detection means 4 Contour information storage means 5 Main point of interest extraction means 6 Sub-point of interest detection means 7 Distance calculation means 8 Judgment means 9 Main point of interest position estimation means 10 Core wire extraction means 11 Core wire information storage means

Claims (2)

[Claims] Claim 1: Concerning contour information extracted by tracing all the boundaries between white pixels and black pixels of binary raster data obtained by quantizing a line figure with an image scanner or the like in a clockwise direction or in the opposite direction, arbitrary information can be obtained. When the outline pixel is the main point of interest and the outline pixel located at the shortest distance across the figure from the main point of interest is the sub-point of interest, the corresponding sub-point of interest is searched while moving the main point of interest, and both points of interest are In a line figure vectorization method that detects a curve connecting midpoints of points as a core line, when the distance between the main point of interest and the sub-point of interest is within a predetermined value, the main point of interest is moved in the data arrangement order of the contour information. Then, both points of interest are detected by moving the sub-point of interest in the reverse order of the data arrangement of the contour information, and if the distance between the two points of interest exceeds a predetermined value, the next main point of interest is detected. A line figure vectorization method characterized by estimating a position from an approximate polynomial of a detected core line and detecting a new main point of interest based on the estimation result. Claim 2: Any contour information extracted by tracing the boundaries between white pixels and black pixels of binary raster data obtained by quantizing line figures using an image scanner or the like in a clockwise or reverse direction. When the outline pixel is the main point of interest and the outline pixel located at the shortest distance across the figure from the main point of interest is the sub-point of interest, the corresponding sub-point of interest is searched while moving the main point of interest, and both points of interest are A line figure vectorization device configured to detect midpoints of points as skeleton lines includes an image input means for inputting an image to obtain binary raster data, an image storage means for storing the raster data, and a raster A contour information detection means for detecting contour information from data, a contour information storage means for storing the detected contour information, and a main point of interest extraction means for extracting the main points of interest from the contour information at regular intervals in the order in which the data of the contour information is arranged. In order to determine the corresponding sub-point of interest, data is retrieved from the contour information storage means in the reverse order of the data arrangement of the contour information, starting from the previous sub-point of interest, and the distance calculation means is used to calculate both points of interest. A sub-point of interest detecting means detects a point where the distance between them is the shortest, and a distance between the obtained sub-point of interest and the corresponding main point of interest is compared with a predetermined threshold, and if the distance is within the predetermined threshold. For example, a determination means that transfers the processing to the skeleton extraction means, and if it exceeds a predetermined threshold value, transfers the processing to the main point of interest position estimation means, and a determination means that estimates the main point of interest from the detected skeleton according to the result of the determination means. a main point of interest estimating means for estimating a main point of interest; a skeleton extracting means for extracting a midpoint between the two points of interest as a point on a skeleton according to the result of the determining means; and a skeleton information storage means for storing detected skeleton information. A line figure vectorization device featuring: