JPH01147786A - Reading device for document containing table - Google Patents

Abstract

PURPOSE:To read a document containing the table of optional construction by making a document picture containing the table into a vector, extracting all ruled lines, dividing a table area one after another and making a column area into table data containing table construction. CONSTITUTION:Picture data PD from an image scanner 2 is directly written in an image memory 11. A vectorizing operation part 12 performs contour tracking for the document picture PD', and makes it into a vector by attaching a prescribed code classified by a tracing direction, or the like, and gives vector data VD to a table construction analyzing part 13. The analyzing part 13 extracts the ruled line from the data VD, and divides the table area one after another, and calculates table construction descriptive data Ds. A table data reading part 14 cuts out and reads a character of every individual column area from the data Ds and the document picture PD'' given from the memory 11, and sends it as table data Dd containing the table construction to a main processor 15. The processor 15 displays the data Dd on a CRT display 3 through an input/output interface.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a reading device capable of reading documents including tables.

[Conventional technology]

2. Description of the Related Art Conventionally, as a method for reading a document including a table, a method is known in which only the data portion of the table is specified and read using an OCR (optical reader).

[Problem that the invention seeks to solve]

However, such reading devices not only require troublesome processing such as specifying only the data part, but also can only be applied to tables with a fixed format, and can read arbitrary table data (table data). The problem is that it cannot be read.

Therefore, it is an object of the present invention to make documents containing tables of arbitrary structure readable.

[Means for solving problems]

An input means for inputting a document including a table; a vectorization means for vectorizing a document image; A table structure analysis means that sequentially divides the area using short ruled lines and labels each of them to extract individual column areas, and extracts characters from the image of each column area and reads the individual characters. Table data reading means is provided.

[Effect]

A document containing a table is input through an image scanner, etc., and the binary image is vectorized using a method such as contour tracking. Ruled lines are extracted from the vector data obtained, and the largest horizontal and vertical ruled lines are formed. Tables with various structures can be read by sequentially dividing a column area using shorter ruled lines, labeling each area, and extracting individual column areas. Furthermore, by cutting out characters from each column area and reading the characters, the contents of each column of the table can be analyzed.

〔Example〕

FIG. 1 is a block diagram showing an embodiment of the invention. The reading processor 1 used in the knee is composed of an image memory 11, a vectorization calculation section 12, a back structure analysis section 13, a table data reading section 14, a main processor 15, an input/output ink face 16, etc. The image scanner 2 and CRT display 3 together constitute a document information processing system.

Image data PD from the image scanner 2 is directly written into the image memory 11. Vectorization calculation unit 12
performs known vectorization on the document image PD' by, for example, tracing the outline and attaching a predetermined code for each tracing direction, and provides vector data VD to the back structure analysis section 13. The back structure analysis unit 13 extracts ruled lines from this vector data VD, sequentially divides the table area, and generates back structure description data D.
Calculate 5. The front data reading unit 14 cuts out characters for each column area from this back structure description data Ds and the document image PD″ given from the image memory 11, reads them, and reads them as main front data D4 including the back structure. The main processor 15 sends the table data D to the processor 15.
d to the CRT shown in FIG. 2 via the input/output ink face 16.
Display on Display 2 or edit by entering necessary data from the keyboard. In addition, this table data can be saved on a floppy disk or stored on another 0A (Office).
ce Automation) processor.

FIG. 3 shows an example of a document including a table. In this case, Table 4A with ruled lines is included in document 4, but character patterns etc. are generally shorter than the vectors of the binarized image of the table, so this is used to separate the character patterns. be able to.

FIG. 4 shows an example of an image of the table. The thick line in the figure is the underlying structure vector data. Here, for one vector ■, which consists of, the vector ■ on the opposite side of the thick line. I+v. 2 is a conjugate vector, and vector V0.2 is a conjugate vector, and vector V0. Let V and □ be defined as adjacent vectors. Note that vectors are indicated with a ``•'' mark, but are omitted unless necessary.

FIG. 5 shows an example of descriptive information for vector .

As is clear from the figure, vector 0. is the starting point P14
．． It is described by the end point Ps□, the length j2i, the slope θi, and the adjacent vectors V, , V, etc.

The contents of the vector information memory are shown in FIG. 6A.

In the vector information memory (not shown), vector number i
(i = 1......, Nv), the starting point (position vector) xl, the ending point (position vector) xEi, and the length of the vector! 21. Slope θi, vector value ■3. Previous adjacent vector number N. , rear adjacent vector number N4. .

Number of ruled lines Fi, ruled line number NAz' (j=1...
....

Fi) etc. are stored. Immediately after vectorization, the number of ruled lines Fi and the ruled line number N A i j are zero (initial values), and after the ruled line is extracted, the number and number of ruled lines of which the vector is a component are determined. Note that the vector ■ is set as Vi =XEi XSi.

FIG. 6B shows the contents of the ruled line information memory after the ruled lines have been extracted.

This is created by extracting all vectors that can be combined, including vectors that overlap or are connected by extension lines, from the data as shown in FIG. 6A. For example, in the example of FIG. 4, the vectors V,,V. l+
■. 2 is one ruled line. Note that the same vector may be included in multiple ruled lines. thus,
Each ruled line has the number of vectors N□ and the vector number N! l,'' (k=
1~Not).

Starting point)' Sj + Ending point ytj + Length 4. The slope θj and the ruled line vector K are determined and stored as shown in FIG. 6B. In addition, for the ruled line vector, KJ=yEJ−ys
Let it be J.

is the column area, and the subscript at the shoulder indicates the row number, and the subscript at the bottom indicates the column number. In addition, Ki in the figure is the starting point ys
i, the end point is yti, and the length is Ki. This is an example in which the back structure can be described in one operation using the longest four horizontal ruled lines and three vertical ruled lines. Furthermore, by cutting out characters from each of the column areas and reading the characters, the contents of the table can be known.

FIG. 8 is a flowchart showing the processing operation of the back structure analysis section. The operation will be explained below using the table of FIG. 9 as an example.

First, calculate the length of the vector data as shown in Figure 6A. .

Sort in descending order of size (see ■). Next, in order from the longest vector, extract all the ruled lines that can be combined from vectors that are conjugate to this vector or adjacent vectors with the same direction, and store them in a part of the vector information memory and the ruled line information memory (see ■) . The ruled line information is also sorted in descending order of length L i (see ■).

Up to this point, all of the vertical ruled lines and horizontal ruled lines in Table 4B shown in FIG. 9 have been extracted. Next, extract the longest vertical and horizontal ruled lines, extract the column area, and label it (■
, see ■). As a result, from the table in Figure 9, only the long ruled lines, . is attached. Note that some tables do not have vertical ruled lines on both sides, but in that case, the vertical ruled lines on both sides are estimated and supplemented. This will make subsequent column division easier.

Thereafter, each column is searched for longer ruled lines and the column area is divided into smaller pieces (see ■).

Repeat this process until the column area can no longer be divided (
■Reference). As a result of the second processing, the ruled line L as shown in Figure 9B.
x3 to LXe are extracted, the column area is divided, and each is given a label as shown.

At this time, since a ruled line having both ends close to each other is extracted with respect to the ruled line surrounding each column area, in the second processing, for example, LxS and Lxb are extracted first from the column area T2. Furthermore, in the third processing, as shown in FIG. 9C, ruled lines 7 to L, 9 are extracted and the column area is divided, and in the fourth processing, as shown in FIG.
. is extracted and the column area is divided.

Finally, the back structure shown in FIG. 9 can be described in a tree format as shown in FIG.

FIG. 11 is a flowchart showing the processing operation of the table data reading section.

The front data reading unit 14 shown in FIG.
: is extracted as shown in Figure 12. From the extracted data,
Lines are cut out by determining the projection value in the horizontal direction (see ■, ■), and then individual characters are cut out by determining the projection value in the vertical direction. Identify characters using the method (■
reference). The above process is performed on the data in all columns, and the data in the table is read.

As described above, the back structure analysis section analyzes the table and describes the structure, and the table data reading section reads the characters in each column area and extracts the table data, thereby allowing tables with various structures to be created. It can read both table data and create a database.

Note that the present invention can be applied not only to cases where the ruled lines are diagonal, but also to cases where figures described by horizontal and vertical line segments are read, as in tables.

〔Effect of the invention〕

According to this invention, a document image including a table is converted into a hector,
All the ruled lines are extracted and the table area is divided in descending order of length, and the column areas are extracted and the table is written hierarchically, making it possible to efficiently read the entire back side. That is, it is possible to input a table with a free structure rather than a fixed format, and since the table structure can be uniquely described, it is possible to obtain table data in a form in which the read results can be reused as a database.

the result (1) Free format tables can be read.

(2) Tables with the same format can be described using the same tree structure (unique representation format).

(3) Tabular data input and statistical processing can be performed in a unified manner.

Effects such as this can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the invention, FIG. 2 is a schematic diagram showing a document information processing system including a reading device according to the invention, and FIG. 3 is an explanation for explaining an example of a document including a table. 4 is an explanatory diagram for explaining an example of a table image, FIG. 5 is an explanatory diagram for explaining vector description information, and FIG. 6A is an explanatory diagram for explaining the contents of vector information memory. , FIG. 6B is an explanatory diagram for explaining the contents of the ruled line information memory, FIG. 7 is an explanatory diagram for explaining the relationship between the ruled lines and area a, and FIG. 8 is a flowchart showing the processing operation of the back structure analysis section. , FIG. 9 is an explanatory diagram for explaining an example of a rather complicated table, FIGS. 98 to 9B are explanatory diagrams for explaining the process of dividing the 69i area and labeling, and FIG. 9 is a tree diagram showing the table structure of FIG. 9, FIG. 11 is a flowchart showing the processing operation of the table data reading section, and FIG. 12 is a tree diagram showing the table structure and table data. Code explanation 1...Reading processor, 2...Image scanner,
3...CR''T display, 4...Document including table, 4A, 4B...Table, 11...Image memory, 1
2... Vectorization calculation unit, 13... Back structure analysis unit,
14... Table data reading unit, 15... Main processor, 16... Input/output ink face. Agent Patent Attorney Kiyoshi Matsuzaki M4Il! ! $5! J lEe Person Chart Figure 6B It! ! 1 @911 Japanese M8 figure 9 person figure 9BI1 go figure figure 9D figure 11

Claims (1)

[Claims] An input means for inputting a document including a table, a vectorization means for vectorizing a document image, and a vectorization means for extracting all ruled lines from the vector data and forming the largest horizontal ruled line and vertical ruled line. Table structure analysis means that sequentially divides a column area using shorter ruled lines and labels each area to extract individual column areas; A reading device for a document including a table, comprising: table data reading means for reading individual characters;