JPH0362285A - Document reading system - Google Patents

Abstract

PURPOSE:To improve an automatic reading function of a document reading system by extracting the underline information concurrently with separation of characters and adding the underline information to the result of re cognition. CONSTITUTION:The characters are separated from a character line S including no underline by means of the projection onto a horizontal axis for example. Then the character separating positions are obtained and an underline area T is also separated via the character separating positions. The presence or absence of underlines is decided for each of separated underline areas. Thus the underline characters are decided. The characters A, B and C are underlined in a diagram. Then a character pattern including no underline is recognized, and the result of recognition is outputted together with the character information including the presence or absence of underlines. Thus not only the document character codes but the underline information can be automatically inputted to an information processing system.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a document reading method, and particularly to a document reading method useful for automatic document reading and for own WJllI collection.

(Prior Art) With the diversification of information processing systems, various data input methods are required, and character recognition technology is also being put into practical use as a powerful data input method. However, current character recognition technology is far inferior to humans in terms of character reading performance, and various improvements are being made. One direction of improvement is to improve the recognition performance of individual characters, and another direction is to improve the reading ability of individual documents.

In recent years, with the spread of Japanese word processors, a variety of printed documents have been created in offices.9 People can easily determine and read text parts, figures, tables, etc. from the various documents.

However, the current character recognition device (hereinafter referred to as OCR: 0ptic)
(abbreviation for character reader)), the format of the document is determined in advance, or the operator specifies the text part, reads only the specified text part, and encodes only the characters to create an electronic document. limited to things.

Therefore, in order to input underlines, which are important information in a document, into an information processing system, the coded electronic document must be edited and the underline information added, or the underlined parts must be manually edited by the operator. I need to enter it.

(Problems to be Solved by the Invention l1l) The above-described conventional document reading method requires cumbersome operations for the operator.

Therefore, an object of the present invention is to extract underline information at the same time as character separation and add the underline information to the recognition result, so that an operator can edit or input part of the information in order to add an underline to an electronic document. 5. Solve the problem that you have to do. An object of the present invention is to provide a document reading method that reduces the burden on an operator and improves automatic document reading functions.

(Means for Solving the Problems) In order to solve the above-mentioned problems, a document reading method according to the present invention includes a document image storage unit that stores document images provided as a two-dimensional grid array, and a document image storage unit that stores document images provided as a two-dimensional grid array. a character line processing unit that reads the document image signal from the character line and separates character lines including underlines; a character line image storage unit that stores each character line image separated by the character line processing unit; and a character line image storage unit that stores each character line image separated by the character line processing unit; an underline detection unit that detects a character line area and an underlined area from an image, and performs character separation from the character line image to extract individual character patterns, and detects whether each character pattern is underlined; The apparatus includes a character separation section that outputs character information including a character pattern and the presence or absence of an underline, and a character recognition section that performs recognition processing on each of the character patterns and outputs a recognition result and character information.

(Operation) The principle and operation of the present invention will be explained below with reference to the drawings.

FIG. 2 shows a printed document image that includes an underline. Conventionally, this underline was considered unnecessary information and was deleted by preprocessing of image processing. An example of this is shown in Figure 3 (
b).

As shown in FIG. 3(a), by obtaining horizontal projection information from the document image, clues for determining the position of character lines in the document can be obtained. Large peaks in the projection information indicate character lines, and narrow, relatively low peaks indicate the presence of an underline.

Conventionally, by detecting the width of the peak, the method shown in Figure 3 (
As shown in b), character lines are extracted without underlines. In contrast, in the present invention, character lines are extracted with underlines included, as shown in FIG. 3(c).

The method for detecting the underlined area is as shown in Fig. 4.
This can be realized by separating the underlined region T from the underlined region T.

FIG. 5 shows the relationship between each individual character and the underline. From a character line S that does not contain an underline, for example, perform character separation using projection onto the horizontal axis, find the separation position, and use that separation position to also separate the underline area T. The presence or absence of an underline is determined for each area to determine which character is underlined. In Figure 5, A,
The three characters B and C are underlined. After this,
Recognizes character patterns that do not include underlines, and outputs recognition results and character information including the presence or absence of underlines.

As a result, not only the character code of the document but also the underlined information can be automatically input into the information processing system.

(Example> Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention.

The document image storage unit 1 stores document images converted into electrical signals by an image input means or the like.

The character line processing unit 2 reads the document image signal as the signal 11 from the document image storage unit 1, extracts the character line area including the underline as explained in the operation section, and converts it into the character line image signal 12.
Output as . The details will be described later.

The character line image storage 13 stores the character line image signal 12.

The underline detection unit 4 reads a character line image from the character line image storage unit 3 as a signal 13, checks whether there is an underline, and when an underline is present, separates the character line and the underline as explained in the section of the operation, and detects the underline. It detects the position of the underline, outputs the position coordinates of the start and end of the underline as a signal 14, and when the underline does not exist, outputs a flag signal indicating that the underline does not exist, the details of which will be described later.

The character separation unit 5 reads the character line image signal 13 from the character line image storage M3, reads the underline start and end position signals 14 from the underline detection unit 4, performs character separation as explained in the operation section, and separates the characters. The position rI obtained as a result of separation
Separate the underlined areas using the l coordinate, determine whether or not each separated area contains an underline, and output each individual character pattern, information on the presence or absence of an underline, position information, etc. as a signal 15. The details will be explained later.

The character recognition unit 6 performs character recognition processing on each individual character pattern sent as a signal 15, information on the presence or absence of underlining, position information, etc., and generates character recognition results or candidates. Information on the character type and the presence or absence of the underline,
It outputs position information and the like as a signal 16, and can be realized using known technology. Note that various individual character recognition processing methods have been proposed, and the present invention is not limited to any particular method.

FIG. 6 is a diagram for explaining in detail an example of the configuration of the character line processing section 2. As shown in FIG.

The horizontal projection extraction unit 21 inputs the document image signal 11 and outputs a horizontal projection histogram obtained by scanning in the horizontal direction as a signal 121.

The projection storage unit 22 stores the projection histogram output as the signal 121.

The local maximum/minimum detection unit 23 receives the signal 12 from the projection storage unit 22.
2, read the horizontal projection histogram, detect local maximum values, local minimum values, and their position coordinates within a global range, and detect the local maximum value, local minimum value, and their position coordinates from top to bottom as a signal 123. It outputs sequentially and can be realized using known technology.

The peak width detection unit 24 reads the signal 123, detects a peak from the series of maximum values and minimum values, calculates the width of the peak from the coordinates of the upper end and the coordinate of the lower end of each peak, and calculates the coordinates of the upper end and lower end of each peak. and the width are outputted as a signal 124 in order from the top, and can be realized by known technology.

The character line detection unit 25 reads the coordinates of the upper and lower ends of each peak and the width of the peak as a signal 124, and when a small peak is adjacent to a large peak, the character line detection unit 25 reads the upper end coordinates of the larger peak and the lower end coordinates of the smaller peak. are determined as the upper and lower coordinates of the character line and output as a signal 125, and when a large peak is adjacent to a large peak, the coordinates of the upper and lower ends of the upper large peak are determined as the upper and lower coordinates of the character line.
JI(, lower end Jijl[) is determined and output as a signal 125, and this process is performed sequentially for all peaks from top to bottom. Note that threshold processing may be used to determine whether a peak is a large peak or a small peak.

The character line extraction unit 2'6 reads the document image signal 11 and the signal 125 indicating the upper and lower coordinates of each character line, and extracts a rectangular area defined by the upper and lower coordinates of each character line as the character line image signal 12. This can be achieved by sequential output.

FIG. 7 is a diagram for explaining in detail an example of the configuration of the underline detection section 4. As shown in FIG.

The horizontal projection extraction section 41 reads the character line image signal 13, scans it in the horizontal direction, and outputs a horizontal projection histogram obtained as a signal 141, and has the same function as the horizontal extraction section 21.

The projection storage unit 42 stores the projection histogram output as the signal 141.

The peak detection unit 43 performs the same processing as the local maximum/minimum detection g23 and the peak width detection unit 24 to check for the presence or absence of a small peak, and when a small peak exists, outputs the upper end coordinate and the lower end coordinate as a signal 143, and detects a small peak. When the peak does not exist, a flag indicating that the peak does not exist is outputted as a signal 14.

FIG. 8 is a diagram for explaining in detail an example of the configuration of the character separating section 5. As shown in FIG.

The area separation unit 51 reads the character line image signal 13 and the signal 14 regarding the presence or absence of an underline or the position coordinates, and when an underline exists, divides it into a character line area and an underline area, and separates the character line area image signal 150 and the underline area. An image signal 154 is output, and when there is no underline, the entire character line image is output as a character line area image signal 150, and an image in which all pixel values are cleared is output as an underline area image signal 154.

The character line area image storage unit 52 stores the character line area image signal 15.
Store 0.

The character cutting unit 53 reads the character line area image signal 151 from the character line area image storage unit 52, performs a process of cutting out individual character patterns, and converts the individual character pattern into a signal 15.
2, the position coordinates of the left end and right end of the individual character pattern are output as a signal 153.

The character pattern storage section 54 stores individual character patterns.

The character pattern coordinate storage section 55 stores the position coordinates of the left end and right end of the individual character pattern represented by the signal 153.

The underlined area image storage unit 56 stores underlined area images.

The underline information detection unit 57 reads the underline area image as a signal 155 from the underline area image storage unit 56, reads the position coordinates of the left end and right end of each individual character pattern as a signal 156 from the character pattern coordinate storage unit 55, and reads the position coordinates of the left end and right end of each individual character pattern as a signal 156. As shown in , it is determined whether or not an underline is drawn at a position corresponding to each individual character pattern, and the presence or absence of an underline is outputted as a signal 157.

The character pattern information integration unit 58 includes the character pattern storage unit 5
4 reads the individual character pattern as signal 158,
The position information of the individual character pattern is read from the character pattern coordinate storage unit 55 as a signal 156, and the underline information detection unit 5
7 reads information on the presence or absence of an underline as a signal 157,
This can be easily realized by outputting these signals together as a signal 15.

(Effects of the Invention) As explained above, according to the present invention, it is possible to extract not only character position information and recognition results, but also information on the presence or absence of underlines, which makes it possible to automatically read documents containing underlines. This greatly helps improve performance.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of a printed document image including underlines, and FIG.
) to (c) are diagrams showing the processing of a document containing underlines, Figure 4 is a diagram showing an example in which the character line area and underline area of a character line image can be separated using a projection histogram, and Figure 5 is character extraction. A diagram showing the process of applying the result to the underlined area to detect the presence or absence of an underline in each individual character pattern.
Block diagram 7 for explaining in detail an example of the configuration of
8 is a block diagram for explaining in detail an example of the configuration of the underline detection unit 4, and FIG. 8 is a block diagram for explaining in detail an example of the configuration of the character separation unit 5. 1... Document image storage section, 2... Character line processing section, 3.
...Character line image storage section, 4...Underline extraction section, 5...
Character separation unit, 6... Character recognition unit, 21... Horizontal projection extraction unit, 22... Projection storage unit, 23... Maximum...
Minimum detection unit, 24... Peak width detection unit, 25... Character line detection unit, 26... Character line extraction unit, 41... Horizontal projection extraction unit, 42... Projection storage unit, 43 . . . Peak detection unit, 51 . . . Region separation unit, 52 .
Character pattern storage section, 55... Character pattern coordinate storage section, 56... Underline area image storage section, 57... Underline information detection section, 58... Character pattern information integration section.

Claims (1)

[Claims] a document image storage unit that stores document images given as a two-dimensional grid array; a character line processing unit that reads the document image signal from the document image storage unit and separates character lines including underlines; a character line image storage unit that stores each character line image separated by the character line processing unit; an underline detection unit that detects a character line area and an underlined area from the character line image;
a character separation unit that performs character separation from the character line image to extract individual character patterns, detects whether or not each character pattern is underlined, and outputs character information including the character pattern and the presence or absence of an underline; , and a character recognition unit that recognizes each of the character patterns and outputs recognition results and character information.