JP2001331764A - Method for recognizing character - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the convenience of the result of character recognition by recognizing the characters of cells in areas in a fixed relation within the same row as one character string. SOLUTION: The table structure of a tabular document is recognized to extract ruled lines and when a ruled line dividing adjacent cells in a row is a dot line, the adjacent cells are integrated to perform the character recognition of them as one cell. It is possible that after integrating the adjacent cells, the dot line dividing the adjacent cells is deleted to character-recognize the integrate cells. It is also possible that after integrating the adjacent cells, the adjacent cells are character-recognized individually to combine the results of the character recognition. When the respective sizes of the adjacent cells are smaller than a fixed threshold and their shapes are similar to each other, the cells can be integrated. Furthermore, it is possible to perform character recognition by integrating plural cells held between the right and left ruled lines of the item area by each row concerning cells in a row lower than the item area of the tabular document.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION In recent years, the demand for character recognition processing as input peripheral devices has been increasing. The present invention relates to a character recognition method capable of making the character recognition result for a cell of a table format document more convenient in the character recognition processing.

[0002]

2. Description of the Related Art Conventionally, in order to recognize cells of a table format document as shown in FIGS. 9 and 10, a ruled line of a table is extracted, and character recognition is performed for each cell separated by the ruled line. The results were kept separately.

[0003]

However, the character groups of a plurality of cells having a fixed relationship in the same row, such as an area A surrounded by a dotted line in FIGS. 9 and 10, are connected to each other. For example, “1234” in the area A in FIG.
56 "is not simply an arrangement of the individual numbers" 1 "," 2 ",...," 6 "but means 123,456. Recognition as a number makes it meaningful as a character string for the first time.In the conventional recognition method, character recognition is performed for each cell separated by ruled lines, and the character recognition result is stored separately, Since the recognition result could not be obtained as a meaningful character string as described above, when the result recognized by the conventional method is processed by, for example, spreadsheet software, the result recognized for each cell is linked to each other. The present invention has been made in view of the above circumstances, and has been made in view of the above circumstances, and has been described above. One sentence of letters By automatically recognized as a column, etc. to be able to directly applied to spreadsheets, an object to a character recognition result and more convenient.

[0004]

FIG. 1 is a diagram for explaining the outline of the present invention. As shown in FIG. 1, in the present invention,
The above problem is solved as follows. (1) As shown in FIG. 1A, the table structure of a tabular document is recognized and ruled lines are extracted. Then, based on the ruled line extraction result, the ruled line that separates adjacent cells in the row is checked. If the ruled line that separates adjacent cells in the row is a dotted line, the adjacent cells are integrated into one Perform recognition. When performing the above-described character recognition, after integrating adjacent cells, a dotted line that separates adjacent cells is deleted from the image, and the integrated cells are subjected to character recognition. It is also possible to perform character recognition on cells individually and obtain a character string obtained by combining the character recognition results. (2) As shown in FIG. 1B, the table structure of the tabular document is recognized and ruled lines are extracted. Then, when the size of each adjacent cell in the row is smaller than a certain threshold value and the shapes are similar, the cells are integrated and character recognition is performed as one cell. When performing the above-described character recognition, after integrating adjacent cells, a ruled line that separates adjacent cells is deleted from the image, and the integrated cells are subjected to character recognition.Also, after integrating adjacent cells, adjacent cells are integrated. It is also possible to perform character recognition on cells individually and obtain a character string obtained by combining the character recognition results. (3) As shown in FIG. 1C, the table structure of the tabular document is recognized and ruled lines are extracted. Then, the item area (for example, FIG. 1)
"Money" in the item area, and the character recognition result in the item area is
If the character string matches a character string registered in advance, character recognition is performed on cells in a row below the item area by integrating a plurality of cells sandwiched between left and right ruled lines in the item area for each row. By performing the character recognition as described above, it is possible to acquire a character group of an adjacent cell of the tabular document as one meaningful character string.

[0005]

FIG. 2 is a processing block diagram of a first embodiment of the present invention. Note that the present invention can be realized by a normal computer system including a CPU, a memory, an external storage device, an input / output device, a scanner for reading an image, a recording medium reading device, a communication interface, and the like. The program for performing the character recognition process of the present invention is stored in the memory. When the program is executed, the program is read into a memory, and the image read by the scanner or the like is subjected to character recognition by the character recognition process of the present invention. Is output from the input / output device.

Next, a first embodiment of the present invention will be described with reference to FIG. This embodiment is applied, for example, to character recognition in an area in which characters are separated by vertical dotted lines, such as the area A in FIG. In the present embodiment, as shown in FIG. 2, first, the table structure recognizing unit 11 extracts a ruled line from an input image, and extracts information such as the length and position of the ruled line, attributes such as a dotted line and a solid line, and the like. Acquire and recognize the structure of the table. As a ruled line extraction method, for example, Japanese Unexamined Patent Application Publication No. 9-5
A known method described in JP-A-0527 can be used. The method of extracting ruled lines and recognizing the table structure is as follows. For ruled lines, attributes such as position, dotted line, solid line, etc., for tables, included cells, rows, etc., information on ruled lines constituting each cell are obtained. Any method can be used.

Next, the cell integration processing unit 12 performs a cell integration process on an area (for example, area A in FIG. 9) separated by a vertical dotted line in the tabular document. As shown in the flowchart of FIG. 4 described later, the cell integration process
Processing is performed for each row, and it is checked whether the ruled line separating adjacent cells is a dotted line. If the vertical ruled line separating adjacent cells is a dotted line, cells are integrated. This process is performed for all cells in a row to integrate the cells. Next, the vertical dotted line between adjacent cells is deleted by the dividing dotted line removing unit 13, and the integrated cells are collectively recognized by the character recognizing unit 14. Thereby, the character recognition result in the integrated cell can be obtained as a character string. In addition, as a character recognition method,
Various methods conventionally proposed can be used.

In FIG. 2, the vertical dotted line between adjacent cells is deleted from the image in the separating dotted line deleting unit 13. However, a character recognition result combining unit is provided, and after performing character recognition for each cell, the recognition is performed. The obtained character strings may be joined to obtain a character string. That is, as shown in FIG. 3, the table structure recognizing unit 11 acquires information such as lengths, positions, dotted lines, solid lines, and other attributes related to ruled lines, recognizes the table structure, and performs cell integration processing. The unit 12 performs a cell integration process. Next, character recognition is performed for each of the cells integrated by the character recognition unit 14, and a character recognition result combining unit 15 is performed.
Then, the characters of the individual character recognition results of the integrated cell are joined to obtain a character string for the integrated cell.

Next, the cell integration processing of the present embodiment will be described with reference to the flowchart shown in FIG. First, cells existing in a row are sorted in order from the left. That is, i = 1
(Step S1), it is checked whether i ≦ [number of lines] (step S2). If i ≦ [number of lines], the process ends. If i ≦ [number of rows], the cells in the i-th row are sorted in order from the left (step S3). Next, the cells are sequentially taken out, and it is checked whether or not the ruled line separating the next cell (adjacent cell) is a dotted line. That is, assuming that j = 1 (step S4), it is checked whether j <[the number of cells in the i-th row] (step S5).
Return to step S2 as i + 1 (step S6). If j <[the number of cells in the i-th row], k = j + 1 (step S7), and it is checked whether k ≦ [the number of cells in the i-th row] (step S8). If k ≦ [the number of cells in the i-th row] is not satisfied, the process proceeds to step S6, where i = i +
As 1 (step S6), the process returns to step S2.

If k.ltoreq. [The number of cells in the i-th row], the process goes to step S9, where the rightmost cell integrated with the j-th cell and the k-th cell are adjacent to each other, and the ruled line that separates them is a vertical dotted line. Find out if If this condition is satisfied, in step S11, these cells (the j-th cell and the k-th cell) are integrated, and k = k + 1 (step S1).
2) Return to step S8. If the above condition is not satisfied, j = k (step S10), and the process returns to step S5. The process of integrating cells as described above is repeated for all cells in a row, and the integration of cells in a row is repeated for all rows.

Next, a second embodiment of the present invention will be described. FIG. 5 shows a processing block diagram of the second embodiment of the present invention. This embodiment is applied to character recognition of an area in which characters are separated by a vertical ruled line, the character size is equal to or smaller than a threshold value, and the characters have similar shapes, as in the area A in FIG. . In the present embodiment, as shown in FIG. 5, first, the table structure recognizing unit 21 extracts a ruled line from an input image and obtains information such as the length and position of the ruled line, attributes such as a dotted line and a solid line, and the like. And recognize the structure of the table. As a ruled line extracting method, as described above, for example, Japanese Patent Laid-Open No. 9-5052
A known method described in Japanese Patent Publication No. 7 can be used. The method of extracting ruled lines and recognizing the table structure is as follows. For ruled lines, attributes such as position, dotted line, solid line, etc., for tables, included cells, rows, etc., information on ruled lines constituting each cell are obtained. Then, as in the first embodiment, the method does not matter.

Next, the cell integration processing unit 22 performs cell integration processing. The cell integration processing is performed for each row as shown in a flowchart of FIG. 7 described later. That is,
For each row, cells are sequentially taken out, and if the taken out cell (cell 1) is smaller than the threshold, the next cell (cell 2) is taken out, and cell 2 is smaller than the threshold, and is similar to cell 1 In the case of a shape, both cells are integrated. Here, whether or not the cells have the same shape is determined, for example, because the cells in the same row have the same height. Can be similar. Next, a vertical ruled line between adjacent cells is deleted by a dividing ruled line deleting unit 23, and a character recognizing unit 24 collectively recognizes characters of the integrated cells. Thereby, the character recognition result in the integrated cell can be obtained as a character string.

In FIG. 5, a vertical ruled line between adjacent cells is deleted from the image in the delimiter ruled line deleter 23. However, as in the first embodiment, a character recognition result combining unit is provided and each cell is deleted. After character recognition is performed, the recognized character strings may be joined to obtain a character string. That is, as shown in FIG. 6, the table structure recognizing unit 21 obtains information such as the length and position of the ruled line, the attributes such as the dotted line and the solid line, and recognizes the table structure as described above. The unit 22 performs cell integration processing. Then, the character recognition unit 24
The character recognition is performed on the individual cells of the integrated cells in step (1), and the character recognition result combining unit 25 connects the characters of the individual character recognition results of the integrated cells to obtain a character string for the integrated cells.

Next, the cell integration processing of the present embodiment will be described with reference to the flowchart shown in FIG. First, cells existing in a row are sorted in order from the left. That is, i = 1
(Step S1), it is checked whether i ≦ [number of lines] (step S2). If i ≦ [number of lines], the process ends. If i ≦ [number of rows], the cells in the i-th row are sorted in order from the left (step S3). Next, cells are sequentially taken out, and then cells are taken out sequentially (cell 1). When the extracted cell 1 is smaller than the threshold, the next cell (adjacent cell: cell 2) is extracted, and when the cell 2 is smaller than the threshold and has a similar shape to the cell 1, the two cells are integrated. I do. That is, j = 1 (step S4), j <
It is checked whether it is [the number of cells in the i-th row] (step S5), j
If not [the number of cells in the i-th row], i = i + 1 is set (step S6), and the process returns to step S2. If j <[the number of cells in the i-th row], it is checked whether or not the j-th cell has a size equal to or smaller than the threshold (step S7). S8), k ≦
It is checked whether it is [the number of cells in the i-th row] (step S9).
If k ≦ [the number of cells in the i-th row] is not satisfied, the process goes from step S5 to step S6, where i = i + 1 as described above.
(Step S6) and return to Step S2. If the j-th cell has a size equal to or smaller than the threshold value in step S7, the process proceeds to step S10, and returns to step S5 with j = j + 1.

In step S9, if k ≦ [the number of cells in the i-th row], the process goes to step S11, where the k-th cell is smaller than the threshold value and is adjacent to the j-th rightmost cell; Check if the shapes are similar. If this condition is satisfied, in step S13, these cells (the j-th cell and the k-th cell) are integrated, and k = k + 1
(Step S14), and returns to Step S9. If the above condition is not satisfied, j = k is set (step S12), and the process returns to step S5. The process of integrating cells as described above is repeated for all cells in a row,
Process all rows in the table. Next, the integrated cells are taken out and individual character recognition is performed.

Next, a third embodiment of the present invention will be described. FIG. 8 shows a processing block diagram of the third embodiment of the present invention. This embodiment is applied to character recognition of a tabular document in which a character string indicating the attribute of a cell in the following row is written in the first row of the table, for example, in the areas B1, B2, and B3 in FIG. . In this embodiment, as shown in FIG. 7, the table structure recognition unit 31 first recognizes the table structure of the input image, and acquires cell information and ruled line information of the table. As the table structure recognition method, for example, a known method described in JP-A-9-50527 can be used as in the first and second embodiments. The method of extracting ruled lines and recognizing the table structure is such that information such as the position and length of the ruled lines, the structure of the cells and rows included in the table, and the information of the ruled lines constituting each cell can be obtained. Any method can be used.

Next, an item area of the table is extracted by the item area extracting unit 32. Here, the item area is, for example, B1,
Cells in the first row of the table, like the areas B2 and B3,
This is a cell in which a character string indicating the attribute of the cell in the second and lower rows of the table is written. For example, in FIG. 9, a cell following the character string “money” in the area B3 indicates that a number meaning the money is entered. Note that the item area is not limited to the cell in the first row of the table, but may be any cell that is extracted according to preset conditions and is a cell in which a character string indicating the attribute of another related cell group is entered. Good. Next, the item region character recognition unit 33 performs character recognition of the extracted item region, and the item region character recognition result comparison unit 34 performs comparison with a character string registered in advance. For example, in the case of FIG. 9, when the character "money" in the item area is recognized, the character is collated with a character string registered in advance.

If the collation results match, the cell integrating unit 35 integrates a plurality of cells sandwiched between the left and right vertical ruled lines of the item area in a row below the item area. As a result, FIG.
In the line below the item "Amount", for example, a numeric string "
1 ”,“ 2 ”,...,“ 6 ”are integrated and“ 123456 ”
Becomes When the cells are integrated as described above, the character recognition unit 36 performs character recognition of the integrated cells. In this character recognition, as shown in FIGS. 2 and 4, the vertical lines separating the individual cells of the integrated cells are deleted from the image, and the integrated cells are collectively subjected to character recognition. The character recognition result in can be obtained as a character string. In addition, FIG.
As shown in FIG. 5, the integrated cells can be subjected to character recognition for each cell, and the character recognition results can be combined to obtain a character string. Note that the character recognition methods shown in the first to third embodiments may be used alone for a tabular document to be subjected to character recognition, or may be used in the first to third embodiments. May be used in combination.

[0019]

As described above, according to the present invention,
Individual characters in cells separated by a dotted line in the same row are automatically obtained as a meaningful character string, or individual characters in adjacent cells in the same row that are smaller than the threshold and have similar shapes Is automatically obtained as a meaningful character string, or when the character string in the item area matches the character string registered in advance, the characters of a plurality of cells in the row below the item area are automatically It can be obtained as a meaningful character string. Therefore, for example, when the character recognition result is used by spreadsheet software, it is not necessary to perform a process of integrating a plurality of cells, and the character recognition result can be made more convenient.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an outline of the present invention.

FIG. 2 is a processing block diagram showing a first embodiment of the present invention.

FIG. 3 is a processing block diagram in a case where the character string recognition method of the first embodiment is replaced with a method of individually recognizing individual cells of integrated cells and combining the recognized characters.

FIG. 4 is a flowchart of a cell integration process according to the first embodiment of the present invention.

FIG. 5 is a processing block diagram showing a second embodiment of the present invention.

FIG. 6 is a processing block diagram in the case where the character recognition method of the second embodiment is replaced with a method of individually character-recognizing individual cells of integrated cells and combining the recognized characters.

FIG. 7 is a flowchart of a cell integration process according to a second embodiment of the present invention.

FIG. 8 is a processing block diagram showing a third embodiment of the present invention.

FIG. 9 is a diagram (1) illustrating an example of a tabular document to be recognized by the present invention;

FIG. 10 is a diagram (2) illustrating an example of a tabular document to be recognized by the present invention;

[Explanation of symbols]

 11, 21, 31 Table structure recognition unit 12, 22 Cell integration processing unit 13 Separation dotted line deletion unit 14, 24, 36 Character recognition unit 15, 25 Character recognition result combining unit 23 Separator ruled line deletion unit 32 Item area extraction unit 33 Item area Character recognition unit 34 Item area character recognition result collation unit 35 Cell integration unit

 ──────────────────────────────────────────────────続 き Continued on the front page F-term (reference) 5B029 AA01 BB02 CC18 CC27 CC30 EE12 5B064 AA01 AB09 AB13 BA01 CA08

Claims (9)

[Claims] 1. A character recognition method for performing character recognition on cells in a tabular document separated by ruled lines, wherein a ruled line separating adjacent cells in a row of the tabular document is a dotted line. A character recognition method, comprising: integrating adjacent cells to perform character recognition as one cell. 2. A character recognition method for performing character recognition on cells in a tabular document separated by ruled lines, wherein a size of an adjacent cell in a row of the tabular document is smaller than a certain threshold value. And a character recognition method wherein, when shapes are similar, the cells are integrated to perform character recognition as one cell. 3. A character recognition method for performing character recognition on cells in a table format document delimited by ruled lines, comprising recognizing the table structure of the table format document, and retrieving table items from the table structure recognition result. Extract the area, and if the character recognition result of the item area matches a pre-registered character string, for the line below the item area,
A character recognition method characterized in that a plurality of cells sandwiched between left and right ruled lines of an item area are integrated for each row and character recognition is performed as one cell. 4. A character recognition program for performing character recognition on cells in a table format document delimited by ruled lines, the character recognition program comprising: a ruled line separating adjacent cells in a row of the table format document. A character recognition program for causing a computer to execute a process of character recognition of the adjacent cell group as one cell when is a dotted line. 5. A character recognition program for performing character recognition on cells in a table-format document delimited by ruled lines, wherein the character recognition program determines the size of an adjacent cell in a row of the table-format document. A character recognition program for causing a computer to execute a process of integrating characters and recognizing characters as one cell when the shapes are smaller than a certain threshold value and the shapes are similar. 6. A character recognition program for performing character recognition on cells in a table format document delimited by ruled lines, wherein the character recognition program recognizes a table structure of the table format document, and Extracting the item area of the table from the recognition result, and if the character recognition result of the item area matches a pre-registered character string, for a row below the item area,
A character recognition program for causing a computer to execute a process of integrating a plurality of cells sandwiched between left and right ruled lines of an item area for each row and performing character recognition as one cell. 7. A recording medium storing a character recognition program for performing character recognition on cells in a table format document delimited by ruled lines, wherein the character recognition program includes an adjacent line in a row of the table format document. A character recognition program for recognizing a character as a single cell when the ruled line that separates a cell is a dotted line. 8. A recording medium for recording a character recognition program for performing character recognition on cells in a table format document delimited by ruled lines, wherein the character recognition program includes an adjacent line in a row of the table format document. A recording medium storing a character recognition program characterized in that, when the size of a cell to be processed is smaller than a predetermined threshold value and the shapes are similar, the cells are integrated and the character is recognized as one cell. 9. A recording medium recording a character recognition program for performing character recognition on cells in a table format document delimited by ruled lines, wherein the character recognition program recognizes a table structure of the table format document. Is performed, and an item area of the table is extracted from the table structure recognition result. If the character recognition result of the item area matches a character string registered in advance, for a row below the item area,
A recording medium on which a character recognition program is recorded, wherein a plurality of cells sandwiched between ruled lines on the left and right of an item area are integrated for each row to perform character recognition as one cell.