JPH0160870B2 - - Google Patents

Description

[Detailed Description of the Invention] (1) Technical Field of the Invention The present invention relates to a character data separation method that enables character groups to be separated and extracted from a drawing containing a mixture of characters, line segments, symbols, etc. and automatically input into a data processing device. In particular, the method estimates the written position of characters based on the recognition results of symbols on drawings, and separates and extracts character groups based on the knowledge obtained from this estimation, thereby extracting characters from image data. This paper relates to a character separation method that improves the accuracy of separating and extracting character data.

(2) Background of the Invention For example, as shown in FIG. 1, when a drawing containing a mixture of line figures and characters handwritten on a sheet of paper with drop-out color grid lines is input into a data processing device such as a computer. , it is necessary for the operator to extract only the linear part and input it using a digitizer or the like. In this case, the operator needs to indicate the end and end of each line and input the information that it is a straight line. Therefore, there is not much of a problem with simple circuit diagrams, but with extremely complex circuit diagrams such as IC circuits, inputting the data becomes extremely complicated and requires considerable patience on the part of the operator. Moreover, it takes a long time to input data, and errors may occur in the input by the operator. In cases where each symbol is assigned a character, the character must be entered separately from a keyboard or from a card.

Therefore, it is possible to automatically extract block pattern symbols and connection information between each symbol from a hand-drawn drawing, such as the one shown in Figure 1, in which characters and figures are mixed, and automatically input them into a data processing device. There was a need for something that could automatically recognize handwritten characters written in or near symbols and provide clean drawings as needed.

(3) Problems with the Prior Art Prior to filing the present application, the applicant proposed a character information extraction method that improved the problem of characters being output with parts missing or line patterns being output. According to this prior art, the graphical structure of the drawing in the vicinity of each grid point on the drawing is represented by a grid point label code, and processing is performed to maintain the continuity of character strokes to avoid missing or broken character strokes. Line patterns mixed in as noise were removed, and text data could be separated and extracted from image data.

However, in the above-mentioned prior art, since there is no means for estimating the area where characters exist, character data extraction processing must be performed on all of the image data, resulting in poor character extraction accuracy and There is also a problem in that it takes a long time to extract characters.

(4) Purpose of the invention In view of the problems in the prior art described above, the purpose of the present invention is to estimate the position where a character should be written from the result of symbol recognition, and to separate character data using this estimated position. The object of the present invention is to improve the accuracy of character separation and extraction and to shorten the time required for character separation and extraction in a character data separation method from image data.

(5) Structure of the invention The gist of the present invention to achieve the above object is as follows:
In a character data separation method that extracts character data from image data obtained by scanning a drawing drawn along a predetermined grid axis and containing a mixture of symbols, characters, and line segments, the type of symbol is character area storage means for storing in advance the relative positions of symbols and areas where characters are to be written according to the grid; and a grid for obtaining grid point code information representing the graphical structure of the drawing in the vicinity of each grid point on the grid axis. A point code information extraction means and a symbol recognition means for recognizing the symbol based on the grid point code information are provided, and an estimated character area is obtained from the character area storage means in accordance with the symbol recognized by the symbol recognition means. , a character data separation method characterized in that character data is extracted based on the estimated character area and the grid point code information.

(6) Embodiments of the invention Examples of the invention will be described below with reference to FIGS. 2 to 4.

FIG. 2 is a diagram showing the relative positions of symbols and character areas. As indicated by diagonal lines in the figure, characters are often written around or inside symbols such as AND gate circuits and flip-flops. In the present invention, the relative position of the character area with respect to the position of the symbol is stored in advance in the form of a dictionary, depending on the type of symbol.

FIG. 3 is a diagram showing a character area on a drawing in which symbols and line segments are mixed. In the figure, the shaded area indicates the character area. As described later, symbol 1, 2, or 3 is first recognized by extracting the grid label code from the image data stored in the image memory, and the character area corresponding to the recognized symbol is stored in the dictionary format described above. Read from. In the estimated character area thus obtained, characters are extracted from the image data by performing processes such as maintaining stroke continuity and removing line patterns.

FIG. 4 is a block circuit diagram showing an apparatus for implementing a character data separation method according to an embodiment of the present invention. In the figure, the first image memory 12 includes:
Image data obtained by scanning a drawing with a mixture of symbols, characters, and line segments drawn along a predetermined grid as shown in FIG. 1 using a facsimile or the like is stored. When a command is input to the control section 10, the address control section 11 sequentially generates addresses corresponding to grid points in the drawing, thereby driving the grid point label code extraction circuit 14. The lattice point label code extraction circuit 14 extracts an address at each lattice point, a character flag indicating the degree of complexity or character-likeness, and 4 indicating the presence or absence of vertical and horizontal strokes.
The bit direction code is extracted from the image data, and this information is held in the grid point label code table 16. For one grid point, the direction code is 1111 if there are strokes in all directions, 0000 if there are no strokes in any direction, and 0000 if there are strokes only in the upward direction. It is determined that it is expressed in 1000. The symbol recognition circuit 18 holds symbols corresponding to direction codes in the form of a dictionary, and receives and holds the direction code for each grid point from the grid point label code table 16 under the control of the address control section 11. Perform matching with the symbol that is
Identify symbols in image data. The identified symbols are maintained in a symbol table. According to the present invention, a character area estimation circuit 22 is provided. The character area estimating circuit 22 stores in advance the relative position of each symbol that can be included in a figure and the area of characters to be written inside or around the symbol, and under the control of the address control unit 22, the character area estimating circuit 22 Perform matching with the symbol data input from the table, add the relative address of the character area to the address of the corresponding symbol to obtain the address of the character area on the image memory, and store this in the character area table 24. .
In this way, a region in which characters are estimated to exist inside or around the symbol is extracted from the image data for each grid point.

Thereafter, only the character data is extracted into the second image memory 30 by performing the character stroke continuity maintenance process and line pattern removal process described in the prior art described above only for the estimated character area. is stored. Conventionally, continuity maintenance processing and line pattern removal processing have been performed on the entire image data, resulting in poor character extraction accuracy and a long time required for character extraction.

Hereinafter, the process of maintaining continuity of character strokes and removing line patterns will be briefly described.

The process of maintaining the continuity of character strokes is to prevent missing or broken character strokes, and to maintain the continuity of character strokes. This phenomenon occurs when a character flag is not set at a grid point where a character stroke exists. In the present invention, under the control of the address control unit 11, the character stroke continuity maintaining circuit 26 receives the address on the image memory of the estimated character area from the character area table 24, and receives the address of the estimated character area on the image memory from the lattice point label code table 16. After receiving the address, only for the area where both addresses match, cut out an image within a rectangular area of N×N mesh centered on the grid point, and check whether the image within the rectangular area touches the edge of the rectangular area. If they touch, check the grid point label code in the tangent direction, and if both the character flag and direction code are "0", change the character flag of this grid point held in the grid point label code from "0" to "1". Update. By sequentially repeating this operation until there is no change in the number of character flags, it is possible to avoid missing or cut character strokes.

The line pattern removal process is a process for removing line patterns mixed in as noise within the rectangular area of the N×N mesh. In the present invention, only for each grid point whose character flag held in the grid point label code table 16 is "1", a line is created by using the eight grid point label codes adjacent to that grid point by the following steps. Perform pattern removal.

Step 1. Under the control of the address control unit 11, the line pattern removal circuit 2 selects the address of the lattice point whose character flag is 1 from the lattice point label code table 16.
8, and the image data in a rectangular area near the grid point is read out from the image memory 12.

Step 2. The direction codes of the eight grid points adjacent to the grid point are checked to see if any of the following four cases apply.

Are the horizontal direction bits of the direction codes of the upper three grid points "1"?

Are the horizontal direction bits of the direction codes of the lower three grid points set to "1"?

Are the vertical direction bits of the direction codes of the three grid points on the right set to 1?

Are the vertical direction bits of the direction codes of the three grid points on the left set to 1?

Step 3. If any of the conditions . . . in step 2 above apply, the next step 4 is performed. If none of the cases in step 2 above apply, it is assumed that there is no line pattern other than characters within this rectangular area, and the image within this rectangular area is separated and extracted.

Step 4. N×1 (1×N) from the line of the rectangular area
Run the window for M meshes and perform the following processing.

Obtain M OR outputs of N×1 (1×N) windows.

If the M outputs obtained in are continuous from the edge and the length L is within (M-1), then N
Image data in an area of ×L (L×N) is removed.

The results of the above processing are sequentially stored in the output image memory 30.
Enter. In this way, only character data can be extracted from the image data stored in the input image memory 12 and stored in the output image memory 12.

(7) Effects of the Invention As explained above, according to the present invention, character areas can be estimated in advance based on symbol recognition.
In the text data separation method from image data,
Character extraction accuracy is improved and the time required for character extraction is shortened.

[Brief explanation of drawings]

Figure 1 shows a drawing handwritten on paper with grid lines, Figure 2 shows the relative positions of symbols and character areas, and Figure 3 shows a drawing with a mixture of symbols and line segments. FIG. 4 is a block diagram showing an apparatus for implementing a character data separation method according to an embodiment of the present invention. 10...Control unit, 11...Address control unit, 1
2... Input image memory, 14... Lattice point label code extraction circuit, 16... Lattice point label code table, 18... Symbol recognition circuit, 20... Symbol table, 22... Character area specification circuit, 24... Character area table, 26...Character stroke continuity maintenance circuit, 28...Line pattern removal circuit, 30...
...Output image memory.

Claims (1)

[Claims] 1 drawn along a predetermined grid axis,
In a character data separation method that extracts character data from image data obtained by scanning a drawing containing a mixture of symbols, characters, and line segments, the relative position of the area where characters are to be written and the symbol according to the type of symbol. character area storage means for pre-storing lattice point code information, lattice point code information extraction means for obtaining lattice point code information expressing the graphical structure of the drawing near each lattice point on the lattice axis, and A symbol recognition means for recognizing the symbol is provided, an estimated character area is obtained from the character area storage means according to the symbol recognized by the symbol recognition means, and a character is generated based on the estimated character area and the grid point code information. A character data separation method characterized by extracting data.