JPH01149184A - Character area extracting system - Google Patents

Abstract

PURPOSE:To correctly discriminate a character area even when a character contacts with other character and graphic by using the property of the character that the character locally has a more complicated stroke than that of the graphic. CONSTITUTION:A drawing to be inputted is inputted as a binary picture by using an AI scanner 2 and stored in a picture memory 5 through a scanner interface 3. A processor 1 makes complicated areas into the character area by obtaining areas in which short line segments, which are equal to or shorter than a threshold value determined by the size of the character to be extracted which is previously determined based on the line information of the image data of the binary picture, are locally crowded after directly approximating the line information of the image data of the binary picture, integrating those areas based on a neighbourhood relation, and defining the complexity in the integrated area. Thus, the extracted character area can be displayed on a CRT 7.

Description

DETAILED DESCRIPTION OF THE INVENTION [Objective of the Invention] (Industrial Application Field) The present invention relates to drawings in which figures and characters mainly composed of line segments are mixedly written on a sheet of paper.

This invention relates to a character area extraction method for recognizing characters in drawings.

(Prior Art) Many drawings include figures and characters mixed together. In such automatic drawing recognition technology, a technology for separating graphic areas and character areas (1) is indispensable.The cut out character areas are sent to a character recognition unit and processed separately from one graphic.

At Yodorai, as a method for cutting out 5-character areas, we focused on the fact that the r size of characters is generally smaller than that of figures, and we labeled each connected component in the drawing and determined the size of the two-dimensional spread of each component. Discrimination was made between character areas and graphic areas. However, with this method, the 15 letters are taken one step closer to τ, and the 5 characters are combined into 5 characters, and the 91 character area is boldly cut out. There wasn't.

(Problems to be Solved by the Invention) As described above, the conventional method has a drawback in that a character area cannot be correctly cut out when characters are in contact with each other due to figures or characters.

It is an object of the present invention to provide one of the character area determination methods that can correctly determine the character area even when the characters described above are in contact with other characters or figures.

[Structure of the invention]

(Means for Solving the Problems) The present invention provides an image memory for storing drawing information including character strings, a means for obtaining line information from the drawing information in the image memory, and a means for obtaining line information from the drawing information in the image memory. means for linearly approximating information; means for determining a dense region having a length equal to or less than a certain threshold determined by the character size to be extracted for the linearly approximated line segment; and means for combining the determined dense regions; The present invention is characterized by comprising means for determining whether or not the combined area is a character area based on the characteristics of the combined area.

(Function) The character area extraction method of the present invention inputs a drawing in which a single figure and characters are mixed, quantizes the drawing by means of scanning and quantizing it, stores the binary code as an image, and stores the image data as a line. After linearly approximating the information (boundary line or thinned figure)? Find regions where short line segments smaller than a threshold value determined in advance based on the size of characters to be extracted are locally concentrated, and integrate these regions based on neighborhood relationships. [By defining the degree of complexity of the integrated area pi], a complex area is set as a character area.

(Example) FIG. 1 is a diagram for explaining an example of the present invention. The control of the entire device is carried out by the processor. The drawing to be input is input as a binary image using the AI scanner (2), and is sent to the image memory controller (4) via the scanner interface (3). Image memory (
5). Address control of the image memory is performed by an address controller (6).

Drawing input instructions are given by the user selecting menus and knees on the sheet using the tablet (8). Images stored in image memory are processed. (!The character area is extracted by processing 13U as described later. The extraction result is displayed on the CRT (7) for confirmation.

The tablet and CRT are controlled by the tablet interface (9) and water entry controller 0, respectively, upon activation of the processor.

FIG. 2 is a diagram for explaining the processing performed by the processor (1) in FIG.

First, the AI scanner 2 scans the drawing shown in Fig. 3, which contains a mixture of figures and characters, performs a photoelectric effect, and converts the black area into to1#.
, is quantized into a binary image with a white background as 0#. The processing of the present invention for this binary drawing information will be described for each step in FIG.

(2) In boundary line tracing, a binary image is used to trace the boundary line of a black area that is in contact with a white bamboo field based on the well-known 4-contact or 8-contact rule.

FIG. 4 shows the traced trajectory of the boundary line.

■In boundary line approximation (C, Williams, Bou
ndedstraight-1ine approx
imation of digitized plana
r cures and 1ines,Compu
, Graphics Jmageprocess 16,
1981.370-381, etc.) The boundary line is approximated by a straight line using a known method. The result of the approximation is shown in FIG. 5, where the black dots represent the nodes of the line segment where the approximation was performed. As for the line segment information, consecutive line segments are registered in the table shown in FIG. 6 for each boundary line of a book. The registered information includes the coordinates and length of the starting point and ending point of the line segment. In addition, the length that some line segments of characters do not exceed, which is calculated from a predetermined character size, is set as a threshold value, and if the length is longer than the threshold value, the flag is set to 0, and if it is shorter, the flag is set to 0. Set flag to 1.

(2) Line segment set extraction focuses on the line segment information in Figure 8g6 that is set to flag 75j 1, and registers continuous line segments with flag 1 in the table shown in Figure 7. Acceptable information is the boundary number that the line segment is included in, and the start and end numbers of the line segment number. Also, find a circumscribed rectangle that includes those line segments. In this embodiment, each side of the rectangle is restricted to be horizontal or perpendicular to the X and y axes. Text is limited to horizontal and vertical writing.

This is enough for Iruhiko. Register the upper left point, lower right point, and center of gravity of the obtained rectangle. This will be called a line segment set.

■In line segment set integration, the line segment sets described above are integrated. As a condition for integration, line segment sets whose rectangular centers of gravity are in the vicinity are grouped together. The conditions are that each side of the circumscribed rectangle containing the integrated line segments is larger than the size of the character, and the shorter side is approximately the size of the character.

In zero line feature extraction, line segment sets unified under one or more conditions are registered in the table shown in FIG. The information to be registered is the integrated line segment set number, the coordinates of the upper left point and lower right point of the circumscribed rectangle that includes all of these line segments, the total circumference of the rectangle, and the length of the included line segments. Information on whether the determination of the character area is correct or incorrect is later written in the flag as Tl#, *0''.

■In character area determination, the judgment condition for determining whether or not it is a character area using the information in the table in Figure 8 is that the character area is the sum of the area and perimeter of a rectangle and the length of the line segment. 8, and write 1 in the flag shown in FIG. 8. Otherwise, write 0.

(However, S is the area of the rectangle, L is the perimeter of the rectangle, and LS is the sum of the lengths of line segments.) ■For character cutting, information on the character area is obtained and the character area is extracted from the 2-step λ image storage unit. .

The character image cut out into the memory αυ of FIG. 1 is recognized as a character using a character recognition technique which is known as a conventional technique.

Next, comprehensive processing of dense portions of short line segments will be explained in detail using a specific example.

The drawing in Figure 9 is approximated as a straight line by the boundary line as described above,
FIG. 10 shows a set of short continuous line segments and a circumscribed rectangle of the set of line segments. The character 12# consists of two sets of line segments. The character "c'Aru013", which has a shape in which the circumscribed rectangle a of one line segment set includes the circumscribed rectangle a of the other -1 line segment set, has two separate line segments because it is cut vertically.・For the characters 4" and "5", there are two tangent line segment sets that include the characters @4# and 5#. The obtained line segment set has the center of gravity of the circumscribed rectangle. In this example, since the distance A between the Oiteha circumscribed rectangles 12a and 13b is short, the circumscribed rectangles 12a and 13b are integrated.At this time, the new circumscribed rectangle 19A containing the two line segment sets has a large size as shown in FIG. The integration will be successful because no changes will be made.

Next, since the distance 1111i18B between the circumscribed rectangles 14c and 15d is short, they are integrated. At this time, although the size of the new circumscribed rectangle 20B including the two line segment sets changes as shown in FIG. fff12, the thickness of the new circumscribed rectangle mentioned above satisfies the conditions, so the process is successful. Although the circumscribed rectangle 16e in FIG. 10 does not have a line segment set for integration, it satisfies the conditions as a region complement and is therefore registered in a table as shown in the ggs diagram. The number of sets is necessarily 1
becomes.

If the three character area candidates cannot be determined by the above-mentioned method in character area determination and are determined to be character areas, 1 is written to the flag in the table in Figure 8, and the circumscribed rectangle is determined to be a character area. becomes.

In the example above, assuming that one character is written horizontally and vertically, the circumscribing rectangle is restricted so that its sides are horizontal or perpendicular to the X and y axes, but the direction of one character is If it is arbitrary, it can be modified and applied by finding a rectangle with the minimum area or a polygon.

〔Effect of the invention〕

As explained above, according to the present invention, characters utilize the property that characters have locally more complex strokes than shapes, so even if characters are in contact with shapes or other characters, it is possible to extract them. It is effective for both character recognition and remaining figure recognition.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of an embodiment of the present invention. FIG. 2 is a processing flowchart of an embodiment of the present invention. Figure 3 is a diagram of the original image of a drawing with a mixture of figures and text, Figure 4 is the opening of the boundary line of a drawing with a mixture of figures and text, and Figure 5 is a diagram of the boundary line of a drawing with a mixture of figures and text. ■Illustration of line approximation.
The figure is a diagram of a table showing short continuous line segments. FIG. 8 is a diagram of a table of integrated line segment sets. FIG. 9 is a diagram of the original image of the specific example, FIG. 10 is a diagram explaining the processing of the specific example, and No. 11111.
FIG. 12 is a diagram showing the relationship between characters and circumscribed rectangles. l...processor, 2...AI scanner. 3... Scanner interface, 4... Image memory controller, 5... Image memory, 6... Address controller, 7... CRT, 8... Tablet,
9...Tough'let interface, 10...R indication=r7 troller, 11...Memory. Agent Patent Attorney Nori Chika ■ Yudo Mitsuyuki Matsuyama I Higashi Bus Figure 1 Figure 2 S3 Figure 6 Figure 7 Figure B It IIK i 12 Figure 9, Figure 1O

Claims (3)

[Claims] (1) an image memory for storing drawing information including character strings, means for obtaining line information from the drawing information in the image memory, and means for linearly approximating the obtained line information;
Means for finding a dense region with a length less than a certain threshold determined by the character size to be extracted for a line segment approximated by a straight line, means for combining the found dense regions, and character extraction from the characteristics of the combined region. A character area extraction method characterized by comprising means for determining whether or not the character area is a region. (2) The character area extraction method according to claim 1, wherein the means for obtaining line information traces boundary lines after binarizing drawing information. (3) The character area extraction method according to claim 2, wherein the means for obtaining line information performs line thinning processing on the binarized drawing information and lines-traces the thinned image.