JPH05159100A - Table recognition device - Google Patents

Abstract

PURPOSE:To provide a table detector which does not mistake a character in a cell surrounded by a ruled line for a broken line. CONSTITUTION:A ruled line extracting part 24 extracts the ruled line from the combination of a run (combined black picture elements) detected by a run extracting part 23 and a labeling part 25 extracts a combined black picture element where the height/width of circumscribe rectangular are less than a prescribed value. A histogram part 26 investigates an arranging state by preparing the histograms in vertical/horizontal directions of them and a table construction detecting part 27 judges a circumscribed rectangular of the accumulating number larger than the prescribed value and of the arrangement of intervals less than the prescribed value to be a broken line. As a line segment is detected based on a characteristic which should be rovided for a broken line, a character is not mistaken for a broken line.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a table recognition device for reading a printed table and recognizing a table structure and characters in cells.

[0002]

2. Description of the Related Art In recent years, a table recognition device has come to be used as an input device for a computer or the like. However, considering the use of the table recognition device as a document clearing system, the ruled line position, line type, and cell You need to recognize the characters.

A conventional table recognition device will be described below with reference to the drawings. FIG. 5 is a block diagram showing the configuration of a conventional table recognition device. In the figure, a recognition command unit 11 issues a command for table recognition by an instruction from a keyboard, a mouse, etc., an image memory 12 stores the input image data, and a solid line run extraction unit 13 scans the image data. An array of black pixels having a certain length or more (hereinafter, referred to as a run) is extracted, while the broken line run extraction unit 14 scans the image data to extract an array of runs of a specific pattern. The table structure extraction unit 15 includes a solid line run extraction unit 13 and a broken line run extraction unit 1.
4, the ruled lines are extracted by checking the connectivity of the runs extracted in step 4, and the cells are extracted by observing the mutual relation of the extracted ruled line positions, and the character recognition unit 17 determines the pattern in each cell of the character pattern dictionary 16. If the similarity is closer than a predetermined value by comparing with the sample character, the character corresponding to the pattern is output, and the table output unit 18 outputs the recognized table information.

[0004]

In such a conventional table recognition apparatus, when the characters are dense in the table, the broken line run extraction unit erroneously recognizes the character portion as a ruled line, and the table structure is improved. There was a problem that could not be recognized correctly.

The present invention has been made to solve the above problems, and an object of the present invention is to provide a table recognition device capable of accurately recognizing a table structure even when characters are densely arranged in the table.

[0006]

In order to achieve the above-mentioned object, the present invention provides a table recognition device for recognizing a table structure based on original figure information obtained by scanning a given input figure, A run extracting means for extracting a black pixel run having a predetermined length or more among two or more black pixel runs in the vertical and horizontal directions, a ruled line extracting means for extracting a solid ruled line from the connected state of the extracted runs, and the extracted ruled line In the cell surrounded by, the labeling means for extracting the circumscribed rectangle of the connected black pixels and the extracted circumscribed rectangle whose width or height is less than or equal to the threshold value are selected, and the histograms in the vertical and horizontal directions are selected. Histogram extracting means to be created, and a dashed line extracting hand for judging a circumscribing rectangle showing a cumulative number of a predetermined value or more in the vertical direction or the horizontal direction in the histogram and positioned at intervals of a predetermined value or less as a broken line A table recognition device equipped with and.

[0007]

According to the present invention, the ruled line extracting unit extracts ruled lines from the connectivity of the runs (connected black pixels) extracted by the run extracting unit, and the labeling unit is provided inside the cell surrounded by the ruled lines. The height or width of the circumscribed rectangle of connected black pixels is extracted below a predetermined value, the histogram section creates histograms in the vertical and horizontal directions to check the alignment state, and the table structure extraction section determines the predetermined value. A circumscribing rectangle having a larger cumulative number than the above and having an array with an interval of a predetermined value or less is determined as a broken line. A line segment is extracted based on the characteristic that the broken line should have.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A table recognition device according to an embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the functional arrangement of a table recognition device according to an embodiment of the present invention. In the figure, 21 is a recognition command section for issuing a command for table recognition by an instruction from a keyboard or mouse, 22 is an image memory for storing input image data, and 23 is a predetermined length by scanning the image data. A run extraction unit that extracts the above runs, 24 is a ruled line extraction unit that checks the connectivity of the extracted runs and extracts ruled lines,
25 is a labeling unit for extracting the circumscribed rectangle of the connected black pixels in the area surrounded by the extracted ruled lines, and 26 is the extracted circumscribed rectangle whose width or height is less than or equal to a threshold value. A histogram extraction unit that creates vertical and horizontal histograms, 27 is a table structure extraction unit that extracts ruled lines other than solid lines from the features of the histogram and the positional relationship of circumscribing rectangles, and decomposes them into cells. 28 stores sample characters. The character pattern dictionary 29 that compares the pattern in each cell with the sample character pattern of the character pattern dictionary 28, and if the similarity is closer than a predetermined value, the character recognition unit 30 that outputs the character recognizes it. It is a table output unit that outputs table information.

FIG. 2 is a block diagram showing a circuit configuration of a table recognition device according to an embodiment of the present invention. In the figure, 31 is a scanner that reads a front image and outputs image data, 32
Is a RAM, an image memory area 33 for storing image data output from the scanner 31, a run area 34 for storing an array of black pixels in the vertical and horizontal directions of the image data,
A circumscribing rectangular area 35 that stores the circumscribing rectangle of connected black pixels of image data, a histogram area 36 that stores a histogram of the frequency of the circumscribing rectangle in the vertical and horizontal directions, a table structure obtained as a result of table recognition, and characters in cells. Has a table data area 37 for storing. Reference numeral 38 denotes a ROM, which has a feature dictionary area 39 storing a dictionary used for character recognition and a program storage area 40. Reference numeral 41 is a processing circuit that executes processing according to the control program, 42 is a keyboard that issues a recognition command, and 43 is a display unit that displays a table stored in the table data area 37.

The mutual relationship and operation of the above components will be described below. FIG. 3 is a flow chart showing the operation of the table recognition apparatus according to the embodiment of the present invention. In step s1, the image is input from the scanner and stored in the memory. Step s2
The inclination of the input image is corrected and the run is extracted from the arrangement of black pixels in the vertical and horizontal directions of the image in step s3. Among the runs extracted in step s4, a run having a predetermined length or more is searched for, and in step s5, the searched runs are connected to extract a ruled line. In step s6, first, in step s5
The presence / absence of the portion surrounded by the solid line extracted in step S1 is checked, and if there is, the process proceeds to step s7, and if not, the process proceeds to step s8. In step s7, ruled lines other than solid lines are extracted (details of this step will be described later). In step s8, the part surrounded by the ruled line extracted in step s5 and the broken line extracted in step s7 is extracted as a cell. In step s9, it is checked whether or not there is a character in the cell for which character recognition has not been completed. If there is, character recognition in the cell is performed in step s11. If not, the process proceeds to step s12 and the recognized table is output. .. With the above processing, the table recognition processing ends.

FIG. 4 is a flow chart showing details of the broken line extracting operation of step s7 in FIG. Step s71
Then, a circumscribed rectangle of connected black pixels in the area is extracted to create a list (hereinafter, this processing is referred to as labeling). In step s72, the presence or absence of unprocessed labeling data is checked. If yes, the process goes to step s73, and if not, the process goes to step s77.
Move to. In step s73, it is checked whether the height of the target labeling data is below a threshold value. If it is less than the threshold value, the process proceeds to step s74, and the histogram of the portion where the labeling data is projected in the horizontal direction is incremented by one. If it exceeds the threshold value in step s73, the process proceeds to step s75 to check whether the width of the target labeling data is less than or equal to the threshold value. If it is less than the threshold value, the process proceeds to step s76, and the histogram of the portion where the labeling data is projected in the vertical direction is incremented by one. If the threshold value is exceeded in step s75, the process returns to step s72. By repeating the above processing, labeling of all circumscribing rectangles is completed, and a histogram of a circumscribing rectangle whose height is less than or equal to a threshold value and a histogram of a circumscribing rectangle whose width is less than or equal to a threshold value are created. In step s77, it is checked whether or not the histogram level is higher than the threshold value. If yes, the processes in and after step s78 are repeated, and if none, the process ends. In step s78, the labeling data passing through the portion found in step s77 is searched, and it is checked whether the maximum interval between the labeling data is less than or equal to a predetermined interval value. If it is less than the threshold value, it is output as a ruled line other than the solid line in step s79. ..

As described above, according to the table recognition apparatus of the embodiment of the present invention, the labeling section, the histogram extraction section and the broken line extraction section are provided, the labeling section encloses the run in a cell with a circumscribed rectangle, and the histogram section A circumscribing rectangle part with a height or width less than or equal to a threshold value is defined as a dashed line candidate part, and its histocrum is created in correspondence with the projection position, and the broken line extraction unit extracts a circumscribing rectangle arranged in a predetermined number or more in the histogram, and extracts it. The broken lines are extracted based on the characteristics that the broken line structure should have by judging that the circumscribed rectangles with the intervals less than a predetermined value are broken lines, so that the broken lines can be correctly extracted without being confused with the characters, and Can recognize the structure of a table.

[0014]

As is apparent from the above embodiments, the present invention provides a table recognition device for recognizing the structure of a table based on original figure information obtained by scanning a given input figure. Of the runs of black pixels in the two directions, a run extraction unit that extracts a run of a certain length or more, a ruled line extraction unit that extracts a solid ruled line from the connected state of the extracted runs, and a line surrounded by the extracted ruled lines In the cell, the labeling means for extracting the circumscribed rectangle of the connected black pixels, and the selected circumscribed rectangle whose width or height is less than or equal to the threshold value are selected to create the histograms in the vertical and horizontal directions. The extraction means and the broken line extraction means for judging a circumscribed rectangle showing a cumulative number equal to or more than a predetermined value in the vertical direction or the horizontal direction in the histogram and positioned at an interval of the predetermined value or less as a broken line. By a recognition device, without erroneously recognized character in a cell with a dashed line, it is possible to know the structure of the table correctly. .

[Brief description of drawings]

FIG. 1 is a block diagram showing a functional configuration of a table recognition device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a circuit configuration of a table recognition device according to an embodiment of the present invention.

FIG. 3 is a flowchart showing the operation of the table recognition device according to the embodiment of the present invention.

FIG. 4 is a flowchart showing a broken line extracting operation in the table recognition device according to the embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a conventional table recognition device.

[Simple explanation of symbols]

 21 recognition command section 22 image memory 23 run extraction section (run extraction means) 24 ruled line extraction section (ruled line extraction means) 25 labeling section (labeling means) 26 histogram extraction section (histogram extraction means) 27 table structure extraction section (table structure excellence) Outputting means) 28 character pattern dictionary 29 character recognition unit 30 table output unit

Claims (1)

[Claims] 1. A table recognition device for recognizing a table structure on the basis of original figure information obtained by scanning a given input figure, in a vertical and horizontal two-direction black pixel run having a fixed length. A run extraction unit that extracts the above items, a ruled line extraction unit that extracts a ruled line of a solid line from the connection state of the extracted runs, and a circumscribed rectangle of connected black pixels in a cell surrounded by the extracted ruled lines And a histogram extracting means for creating a histogram in the vertical and horizontal directions of the extracted circumscribed rectangles whose width or height is equal to or less than a threshold value, and in the histogram in the vertical or horizontal direction. A table recognition device comprising: a circumscribing rectangle showing a cumulative number equal to or larger than a predetermined value, and a broken line extracting means for judging a rectangle positioned at an interval of a predetermined value or less as a broken line.