JPH0679350B2 - Line figure processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to a line graphic processing device, and particularly to a line graphic such as a drawing, and a fine graphic such as a character symbol is included in the line graphic. Accurately and faithfully extract the line figure from a binary image that touches / intersects at a straight line portion or a bending point portion, and then use this to automatically input topographic maps / housing maps in a geographic database, or in various drawing databases. The present invention relates to a line figure processing device capable of efficiently performing automatic input of drawings.

[Conventional technology]

The conventional methods for realizing this type of device are (a) a method of linearly approximating a black pixel row obtained by thinning the image data on the image memory, and (b) a slender black line whose boundary lines are parallel to each other. A method of tracing and extracting a line figure using the boundary line pair, which is regarded as a pixel area, and (c) tracing the boundary line of the figure and the distance information along the boundary line to obtain the surrounding relation of the boundary line and the circumscribed rectangle, Method of extracting a graphic, (d) Extracting a linear graphic by determining the length of a black pixel area that is linearly continuous in the tracking direction while tracing the black pixel area and the number of black pixels in the direction perpendicular to the tracking direction. The method to do, etc. have been announced.

Among these, in (a), the thinning process is applied to each figure data which is a component of the image data on the image memory. Therefore, the image quality is deteriorated due to the change of the intersection and the occurrence of the beard, and the line width is reduced. The lack of information and image information makes it difficult to distinguish between line figures and fine figures, and is not suitable for faithfully extracting line figures from the original image.

On the other hand, the methods (b), (c), and (d) are methods that do not perform thinning processing. And a fine figure cannot be separated. Therefore, it is not suitable for extracting a line figure from an image in which a fine figure touches or intersects with a line figure.

Since the method of (C) uses distance information, it can be dealt with when a fine figure is in contact with a line figure. However, also in this case, it is not possible to deal with the case where the fine figure intersects the line figure.

On the other hand, in the method (d), since a straight line is fitted on the black pixel area representing a figure, it is possible to deal with a case where a fine figure directly intersects with a line figure. Is only in the case of contact / intersection in the straight line part of the line figure, and cannot be extracted in the case of contact / intersection at the bending point part.

Further, in general image data, the same relatively small figure often appears as a line figure or a fine figure depending on the type of drawing. To deal with this, the meaning content of the figure Higher-order processing in consideration of is required. However, none of the above methods can deal with this problem as they are.

[Object of the Invention]

Generally, as conditions for extracting a line figure from a given binary image (hereinafter referred to as extraction processing conditions), (a) line figure can be faithfully extracted to the original image, and (b) fine figure is a line figure. It is possible to perform extraction processing regardless of whether it touches or intersects with either the straight line part or the bending point part of (3). (C) The shape of the fine pattern remaining after extracting and separating the line pattern from the original image is damaged. There are none, (d) It is possible to deal with higher-level processing based on the meaning content of the graphic element as necessary. Since the above method does not satisfy all of these extraction processing conditions, the application area naturally has a limit.

The present invention aims to satisfy all of these conditions. Specifically, each graphic element in a given binary image data is defined as the largest square inscribed in the graphic element and adjacent to each other. It consists of a square data string (called a maximum square. It may also be simply called a square below) (called a chain) and a maximum square (called a connector) that connects chains and inscribes in the figure. By expressing the binary image data as a single graphic file by expressing it as a structured structure, the data description of the graphic is performed based on the file, and the straight line tracing process is performed on the data description. The purpose is to solve the above problems. It should be noted that the chain and connector of the figure are collectively referred to as the core of the figure. Therefore, the graphic file can be regarded as the core structure of the graphic. That is, the graphic file expresses the geometric structure of the original image data as a connection relationship between the chain and the connector,
The chain data is the line width at each position of the line figure, or the information of the surface of the fine figure is the length of the side of the square (called the side length).
In addition, since the original image is completely reproduced, it is possible to easily cope with the configuration of the method that satisfies the extraction processing conditions (a) to (d).

〔Example〕

FIG. 1 shows an embodiment of the device configuration according to the present invention. A given binary image data 1 such as a drawing is a graphic file 2 in which the core line structure of the graphic is described in advance by the graphic file operation unit 3.
The process relating to line figure extraction will be performed while sequentially reading the chain data from this figure file. Here, the outline of the overall flow of the process will be described first. The chain data selected and read by the chain data selection unit 4 from the graphic file 2 via the graphic file operation unit 3 is once stored in the chain data storage unit 5, and then one or more chains are processed by the thickness division processing unit 6. Are divided into partial chain data, and the square address of each division point (each square constituting the chain is addressed in the order of 1, 2, ... In accordance with the adjacency relation) is stored in the square address storage unit 7. A partial chain (divided cormorant chain) determined by the division point square address on the storage unit 7 and the chain data on the chain data storage unit 5 is polygonal-line-approximated by the polygonal-line approximation processing unit 8, and its bending point square address is the same as that of 7. Are stored in the square address storage unit 12 of. Next, each divided part chain is described / registered in the core line structure description table 14 by the core line structure description processing unit 13 together with the bending point square address and the connector related thereto. When this series of processing is completed for all of the chain data in the graphic file 2, the straight line tracking processing unit 15 is activated, and the core line data in the core line structure description table 14 is linearly traced, and the straight line is recorded on the table. Extract parts. Upon completion of this processing, the line figure output processing unit 16 is activated to search for straight line data on the table and output the straight line data to a fax output device, a graphic display device or the like as necessary.

Hereinafter, the outline of the function of each processing unit and the implementation method will be described.

(A) Graphic file operation unit-This operation unit is MSM proposed as a method for extracting the core structure of digitized binary figures.
Law (Wakayama, T, “Core-Line Tracing Algortithm Bas
ed on Maximal Squara Moving, "IEEE Trans.PAMI, Vol.PA
MI-4, No.1 (1982)), using the representation format of the core data, the binary image is a single file (figure file)
As a file, and search the graphic data from the file as needed. FIG. 2 shows an example of core data extraction.
In Fig. 2, the Y-axis is oriented downward (the same applies below). In FIG. 2, a black circle represents (the center point of) the largest square inscribed in the figure, and its connection represents the core line of the figure. The data structure of the graphic file is shown in FIG. In FIG. 3, (a) represents the core line portion (● is the largest square) in FIG. 2. Two largest squares adjacent to each other in the center line of the figure are said to be MSM adjacent to each other. As shown in Fig. 3, the core line is composed of the largest square ordered sequences (C1, C2, etc.) that are MSM adjacent to each other and called J1, J2, etc. that connect the chains by the geometrical shape of the branch points. It can be decomposed into the largest squares (which we call connectors). The core data in the graphic file is linearly represented by the connection relationship between the chain and the connector as shown in FIG. in this case,
Each chain and connector in the file is an address (chain address, connector address) on the file
And these data are read by giving the address. The figure file operation unit is realized by applying the method described in the document “Wakayama,“ Figure operations using MSM core data, ”Institute of Information Science, CV27-2 (1983)”.

(B) Chain data selection / reading unit-This processing unit sequentially reads the data in the graphic file described in (A) above from the beginning of the file by giving the chain address. Initialize the chain address and read the first chain data. After that, the chain address is advanced by receiving the end signal of the core structure description storage unit 13, and the chain data is read at the address. The read data is stored in the chain data storage unit 5.

In general, the chain C can be expressed as a column C = [S1, S2, ..., Sn] (i) of maximum squares Si (i = 1, 2, ..., N) adjacent to each other by MSM, by definition. Where n is the chain C
I is called the address of the square Si in C (square address). In addition, the square Si constituting C is Si = [(xi, yi), li] (ii) (xi, yi): coordinates of the upper left corner of the square li: side length of the square (= 0,1,2 ...) Can be described. At this time, assuming that the maximum possible chain length is M, the chain data storage unit 5 stores FIFO-type registers 5-1 and 5-2 of length M for respectively storing xi, yi, and li. , 5-3.

(C) Thickness division processing unit-This processing unit divides the chain accumulated in the chain data accumulating unit 5 into a thin portion and a thick portion according to its thickness.

Now, the side length li for the square address i on the chain C
Let L (l) = [l1, l2, ... ln] (iii) be the sequence of the above and call it the side length sequence of the chain C.
The side length sequence gives the line width of C. At this time, with respect to the line figure shown in FIG. 4 (a) (assumed to be represented as one chain C), the horizontal axis is the square address, the vertical axis is the side length, and the side length sequence is If plotted, it would look like Fig. 4 (b). If C is divided into a thin portion and a thick portion by the side length sequence, the division point is appropriate to be the square (center) of the addresses a3, a4, a5 and a6. Dividing the chain into one or more chains according to the side length sequence is called thickness division of the chain.

Therefore, the main division processing unit compares the values of the previous data while sequentially reading the values in the FIFO format from the side length sequence register 5-3 in the chain storage unit 5, and the embodiment of the main processing unit described below. When the division point is found by the method of 1, the square address is output and stored in the square address storage unit 7 (FIFO type register). The present processing unit is activated by receiving the read end signal from the processing unit 4.

An embodiment of this processing unit is shown in FIG. Controller 6-1
Initializes the address register 6-2 for sequentially reading the value of the square side length from the side length sequence register 5-3,
After that, it is incremented by 1 by receiving the end signals of the processing circuits 6-16 to 6-21 described later. Further register 6-2
Is sent to the FIFO type register 5-3, and the square side length value designated by it is read out to the register 6-3. The same side length value is subtracted from the side length value held by the delay circuit 6-6 before one operation, and is output to the difference register 6-5. The same difference value and the difference value before one operation held in the delay circuit 6-7 are determined by the determination circuits 6-9 and 6-10 as positive, 0, and negative values, respectively, and the processing circuit 6 is determined by the combination thereof. -5
~ 6-21 starts up. 6- as a register related to this
11 to 6-14 are prepared. The register 6-11 is a change point (address) of the side length value in the side length sequence register 5-3.
Is a register with. The register 6-12 is a counter that counts the number of consecutive side lengths having the same value. The register 6-13 is a constant to be compared with the contents of the register 6-12, and the register 6-14 is a side length value 6-before one operation.
6 and the absolute value of the difference between the long axis of the side read from 5-3 and the constant to be compared are respectively held as needed. Processing circuit 6-15 increments the contents of register 6-12 by one. The processing circuit 6-16 resets the register 6-12 (to the value 1). The processing circuit 6-17 is a register 6-12.
And 6-13 are compared, and if the former is large, register 6
-2 is transferred to register 6-11. After that, the registers 6-12 are reset. The processing circuit 6-18 obtains the absolute value of the difference between the value of the register 5-3 read with the content of the register 6-11 as an address and the side length value 6-6 before one operation, and the absolute value is the register 6 If it is less than the value in -14, the contents of register 6-3 are transferred to register 6-11. If not, the registers 6-12 and 6-13 are compared, and if the former is large, the value obtained by subtracting the content of 6-12 from the content of the register 6-2 is used. If the latter is large, the register 6-2 is used. The contents of
The thickness division point square address is transferred to the square address storage unit 7. Similarly, the processing circuits 6-19 to 6-21 compare and transfer values among these registers, and the change point address of the side length value is square address accumulating section 7 as necessary.
Is output to.

(D) Polygonal line approximation processing unit-This processing unit includes a chain center line calculation unit 9 and a polylined line approximation unit 11 as its subordinate processing units.
The center line calculation unit 9 calculates the center line of each partial chain (hereinafter, simply referred to as a chain) divided by the thickness division processing unit 6, and stores the center point sequence in the coordinate point sequence storage unit 10. Next, the center line data is approximated to a polygonal line by the processing unit 11, and a square address corresponding to the bending point is stored in the square address storage unit 12.

Now, with respect to the chain C represented by the above equations (i) and (ii), the center point Pi of the square Si on C Pi Pi = (xi + li / 2, yi + li / 2) (iv) Example L (C ) Consider L (C) = [P1, P2, ... Pn] (v), and call this the center line of the chain C. Further, with respect to the center line (v) of the chain C, points Qi (i = 1, 2, ..., N−
When 1) is an intermediate point between two points Pi and Pi + 1, the point sequence L '(C) = [Q1, Q2, ... Qm] (vi) is called a smoothed curve of the curve L (C).

The center line L (C) of the chain C is smoothed, each line is approximated by a polygonal line, and the square address of the bending point that receives both ends is a0 (= 1), a1, ..., ak (= n) (K
Is the number of approximate line segments) (FIG. 6), this is called the line approximation of the chain C. In the case of FIG. 4 (a), one of the thickness-divided partial chains C1 is the bending point address a0.
It is expressed as a polygonal line approximation with (= 1), a1, a2, a3, a4 (= n1).

FIG. 7 shows an embodiment of the polygonal line approximation processing unit. Adjacent address pairs on the square address accumulating unit 7 are read from 7, square data in the section are sequentially read from the chain data accumulating unit 5, and each square data is x register 9-1, y register 9-2, l. Store in register 9-3. From these data, the center point of the square is obtained by the center point calculation circuit and stored in the register 9-5. Next, the center point data and the center point data before one operation held by the delay circuit 9-6 are averaged by the averaging circuit 9-7, and the intermediate point between them is stored in the coordinate point sequence accumulating unit 10. 5
When all the corresponding squares in 10 have been processed, the point sequence data in 10 is subjected to polygonal line approximation by the polygonal line approximation circuit 11, and the bending point square address is stored in the square address accumulating unit 12. Specifically, the circuit 11 uses the distance between the straight line formed by the end points of the point sequence and other points on the point sequence,
Easy to implement.

(E) Core line structure description processing unit-This processing unit uses the core line structure data on the graphic file to create one core line structure data based on the connection relationship of each core line, the bending point square address by the broken line approximation process, the average value of the chain thickness, etc. It is expressed as a table (called a core structure description table).

Now, consider the figure shown in FIG. This is 6 chains J1 and J of C1, C2, ..., C6 on the graphic file 2.
It consists of 2 connectors. Here, if the individual chains are divided into thicknesses, for example, the chain C1 becomes C11, C12.
The thickness is divided into three partial chains of C13 and C13, and the others are the same, but if the connection relation (core structure) between them is formally expressed explicitly, it becomes as shown in Fig. 8 (b). . In FIG. 8, the width of the line figure is also represented for convenience. In this case, for example, the chains C13, C2, C3 are said to be adjacent to each other via the connector J1, and the chains C11 and C12 are adjacent to each other by dividing the thickness. Further, the two may be collectively referred to simply as chains being adjacent to each other. The core structure description table represents these chains and connectors and the relationships between them as one data structure. This description table describes the geometric structure of the original image, which allows complicated processing.

FIG. 9 shows an embodiment of the core structure description table for FIG. 8 (b). In FIG. 9, 14-1 relates to chain data, and 14-2 relates to connector data. In 14-1, column T shows the average thickness of the chain, NP T and NP B represents a pointer to a chain or a connector that connects at the beginning and end of the chain (in the squares forming the chain, the end part on the chain where the square with the smallest square address exists is the beginning part of the chain, The other end is called the end of the chain). 0 indicates that there is nothing to connect, negative value indicates connection to the chain, positive value indicates connection to the connector, and absolute value other than 0 indicates the value (pointer to the entry in the corresponding table). )
Is shown. The column FP is a pointer to each core element on the graphic file. This enables more detailed processing. Column TL T and TL B is link data written by a straight line tracking processing unit described later. Column SA is a bending point square address based on the polygonal line approximation. The data in each column is a description example regarding the graphic in FIG.

(F) Straight line tracing processing unit-This processing unit is activated when the above series of processing is completed for all the chain data in the graphic file 2, and performs straight line tracing processing on the core structure description table. Specifically, by sequentially checking the continuity (straight line continuity) as a straight line at an adjacent point (connector or thickness division point) between chains that are adjacent via a connector or by thickness division Done. Hereinafter, such a tracking process will be referred to as a straight line tracking process.

Generally, a chain on the core structure description table is expressed as a connection (polygonal line) of one or more line segments. In this case, each line segment on the chain can be regarded as a vector (tracking vector) in consideration of its tracking direction. Therefore, tracing a straight line on a data structure means that vectors that are on different chains (ends) and that are adjacent to each other (start point and end point) via connectors or by thickness division The continuity as a straight line is checked, and if both tracking vectors form a straight line, a link is established on the table between the corresponding chain data. In addition, when the chain C is traced in a certain direction, the k-th polygonal line constituent vector is V (C, k)
Shall be written.

It is now assumed that the figure is traced from the left on the core structure description table in FIG. 8 and reaches the chain C11. In this case, the linear continuity between the vector V (C11,4) and the vector V (C12,1) is examined by the method of checking the linear continuity of the tracking vector described in the next paragraph. Similarly, the linear continuity between the vectors V (C12,1) and V (C13,1) is examined. It is assumed that the straight line tracking further advances and reaches the chain C13. In this case, a chain having a vector that is linearly continuous with the vector V (C13,1) via the connector J1 is selected from the vicinity of J1. In this case C3 is selected. In the following, straight line tracking is performed in this manner. The description of the linear continuity between chains is made by linking the chains as described above.
It will be described with reference to the drawings.

Column TL on table 14-1 in FIG. 9 T and TL B Related to this, TL T is the link information at the beginning of the chain, TL B is used to represent the link information at the end of the chain. If there is a linear continuity between chain C1 and chain C2, TL of C1 T or TL
In column B (depending on which column the linear continuity of the chain holds (beginning / end)), if the linear continuity holds on the thickness dividing point, a value of 1 is established via the connector. If so, enter the number of chain C2 in the NBR column for that connector. When two tracking vectors pass on each connector, the intersection is calculated and registered as an attribute (connector passing point).

Next, a method for checking the linear continuity of the tracking vector will be described. In straight line tracking, a vector for starting tracking is initially set. This vector (straight line tracking vector)
As the line continuity check described in this section progresses, the head of the line continuously extends in the line tracing direction. Now, the straight line tracking vector V at a certain point (Pt and Ph are called the end and the beginning of the tracking vector, respectively). In this case, the vector V'as the next vector to be tracked. , If the two vectors are almost aligned and the partial chain C'representing V'is not thicker than the tracking start chain C, the straight line tracking vector V grows. (Fig. 10 (a)). Next, consider the case where C'is thick. In this case, when C'is long, the straight line tracking is finished, but when it is short as shown in FIG. 10 (b), the tracking vector V is left as it is, and the chain (vector) is further tracked. Move. Here the tracking vector V ″ Exists, and is thin as shown in FIG. 10 (b), the straight line tracing vector V grows. Becomes If there is no next adjacent chain (vector) as shown in FIG. 10 (c), the line tracing vector V grows. Becomes Here, Po is the foot of the perpendicular line drawn from the point Ph 'to the vector V.

(G) Line figure output processing section-This processing section represents a chain described on the core structure description table 14 and a chain linearly continuous with the chain as one line figure (a coordinate point sequence representing a polygonal line approximation). The line drawing extracted in this way is integrated and output to an external storage device, a fax output device, a graphic display device, or the like. Called by a signal.

〔The invention's effect〕

As described above, according to the present invention, the above-described various conditions (conditions for extracting a line figure from a binary image) that cannot be satisfied by the conventional technique, that is, (a) The line figure can be faithfully extracted, (b) the extraction process can be performed even if the fine figure is in contact with or intersects with either the straight line portion or the bending point portion of the line figure, (c) the line figure can be extracted from the original image. It is possible to satisfy all that the shape of the remaining fine figure after extraction / separation is not impaired, and (d) it is possible to deal with higher-level processing based on the semantic content having a figure element if necessary. Become.

FIG. 11 shows an example in which line graphics are extracted from a binary image by the present apparatus. FIG. 11 (a) shows the original image and FIG. 11 (b) shows the result of straight line extraction. As can be seen from FIG. 11, straight lines are faithfully extracted even when characters / symbols, etc., contact and intersect the line figure at the bending points. Since the connection relationships of these line segments are all described in the core structure description table, it is possible to configure a higher-level data processing unit and connect the core structure description table to this if necessary. Becomes

[Brief description of drawings]

FIG. 1 shows an embodiment (apparatus configuration diagram) of the present invention. FIG. 2 shows an example of core line extraction. FIG. 3 is a conceptual diagram of the configuration of a graphic file. (A) is a grouping based on the connection relation of the largest squares, and (b) is a linearized representation of core data. FIG. 4 explains the division by the thickness of the chain,
(A) shows an example of a line figure (it is expressed as one chain), and (b) shows a graph display of the side length sequence of the chain (a). FIG. 5 shows an embodiment of the thickness division processing section. FIG. 6 is an explanatory diagram of the broken line approximation of the chain. FIG. 7 shows an embodiment of the polygonal line approximation processing unit. FIG. 8 is an explanatory diagram of a structural representation of a line figure, (a) is an example of the figure, and (b) is a conceptual diagram of the core line structure representation of the figure (a). FIG. 9 shows an example of a core structure description table (FIG. 8 (b)).
It corresponds to)). FIG. 10 is an explanatory diagram of a method for inspecting the linear continuity of tracking vectors. FIG. 11 is an example of line figure extraction processing by this device,
(A) shows an original image and (b) shows a straight line extraction processing result. 1 ... Drawings (binary image data), 2 ... Graphic file, 3 ... Graphic file operation unit, 4 ... Chain data selection / reading unit, 5 ... Chain data storage unit, 5-1
...... FIFO type register that stores coordinate point sequence data (x coordinate), 5-2 FI that stores coordinate point sequence data (y coordinate)
FO type register, 5-3 ... FIFO type register for storing square side length sequence data, 6 ... Thickness division processing unit, 6-1 ...
… Controller, 6-2 …… Address counter, 6-3
... edge length value register, 6-4 ... adder, 6-5 ... difference register, 6-6 ... delay circuit, 6-7 ... difference register, 6-9 ... judgment circuit, 6-10 ... Judgment circuit, 6-
11 …… Change point register, 6-12 …… Equivalent number counter, 6
-13 ... Constant value register, 6-14 ... Constant value register,
6-15 to 6-21 ... Processing circuit, 7 ... Square address storage unit, 8 ... Broken line approximation processing unit, 9 ... Chain center line calculation unit, 9-1, 9-2, 9-3. Registers for storing x, y, l, 9-4 ... Center point calculation circuit, 9-5 ... Coordinate point storage register, 9-6 ... Delay circuit, 9-7 ... Intermediate point calculation circuit, 10 ... … Coordinate point sequence accumulation part, 11 …… Polygonal line approximation part, 12
…… Square address storage unit, 13 …… Core line structure description processing unit, 14 …… Core line structure description table, 15 …… Line tracing processing unit, 16 …… Line figure output processing unit.

Claims (4)

[Claims] 1. A line figure processing apparatus which traces and extracts the line figure from binary image data composed of line figures and characters / symbols (referred to as fine figures) to approximate a polygonal line. The core data of each graphic element included in the value image (the column of the largest square (called the maximum square) inscribed in the graphic element (called a chain) and the maximum connecting the chain to the chain and inscribed in the graphic) By describing squares (referred to as connectors) and structures (referred to as core structure) that represent the connection between them, the binary image is filed in advance as a single graphic file, and A graphic file operation unit for reading core data from the graphic file as necessary, and (b) a chain for selecting and reading chain data from the graphic file via the graphic file operation unit. A data selection / reading unit, and (c) the chain data fetched by the selection / reading unit is divided into partial data for each thickness based on the side length data of the squares forming the chain data. A thickness division processing unit; (d) a polygonal line approximation processing unit that approximates the data (referred to as partial chain data) of each part of the chain divided by the thickness division processing unit to a polygonal line; and (e) the thickness A core line structure description processing unit that describes the core line structure related to each partial chain data obtained by the division processing unit as a table (called a core line structure description table) based on the data obtained by the polygonal line approximation processing unit. (F) The chain data on the core structure description table described by the core structure description processing unit is traced by tracing the pointer on the table, and is adjacent to the chain data. Whether the chain data is continuous as one straight line (which is called a straight line), and if the straight chain is continuous, a straight line tracking processing unit that links the two chain data on the table; g) Utilizing that the chain data linked on the table is continuous as a straight line from the core structure description table traced by the straight line processing unit,
A line figure output processing unit that outputs line figure data as a sequence of points representing a polygonal line approximation, and a line figure processing apparatus. 2. The thickness division processing unit divides the chain data of each graphic element into a thick portion and a thin portion by a data string of the side length of each square forming the chain. The line figure processing device according to the first section. 3. A polygonal line approximation processing unit calculates a center line of a chain of each graphic element as a data string of center points of squares forming the chain, and the data string (coordinate point string).
The line figure processing apparatus according to claim 1, wherein the chain data is approximated to a broken line by approximating as a line segment connection. 4. The straight line tracking processing unit determines whether or not two chain data on the core structure description table are continuous in a straight line by using chain thickness information. The line graphic processing device according to item 1.