JPH08287275A - Table recognition device and method for interpolating break of table frame line - Google Patents

Abstract

PURPOSE: To provide the device and method which can correctly interpolate a break of a table frame even when the break is large by measuring the widths of table items, checking their frequencies, and interpolating a break of the table frame line present in the table. CONSTITUTION: This device is equipped with a means 1 which detects the external outline of the table frame line, a means 2 which extracts one or plural table internal outlines present in the extracted closed curve, a table frame width measuring means 3 which measures the distance between two pixels that belong to the same table internal outline and are present in the same row or the same column as the table frame width as to each extracted table internal outline, a histogram measuring means 4 which finds a histogram of the table frame width, a table frame line break estimating means 5 which regards a value of high frequency in the histogram as the standard table frame width and judges that there is a table frame line break at a place corresponding to a value of low frequency, and an interpolating means 6 which interpolates the estimated broken table frame life.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a table recognition apparatus and method, and more particularly to a table recognition apparatus and method for reading a document including a table format and recognizing a table structure.

[0002]

2. Description of the Related Art Recently, there is a great need to read a document with an optical device to digitize character information, and many such documents include a tabular format.

In the case of digitizing a table document, conventionally, as described in, for example, Toshiba's Express Reader manual or Media Drive's Win Reader manual, ruled lines are removed and only characters are written. There is a way to recognize.

However, the position of each item frame (referred to as "cell") is one of the important information in the table document. Therefore, it is possible to correctly extract each cell and read the characters written in the cell. is important.

Generally, in order to extract a cell, a straight line may be detected using Hough transform or the like, or a closed area may be found by tracing black pixels corresponding to a table frame line.

However, the former method has a problem that it takes a lot of processing time. On the other hand, the latter method requires less calculation than the former method, but if some of the table frame lines in the original are originally broken or thinly drawn, they will be displayed in the optically read image. It is difficult to extract the table structure correctly because the frame breaks. Therefore, a technique for interpolating the broken frame line in the image is required.

As a conventional device for recognizing a table format including a broken table frame line, for example, in Japanese Patent Laid-Open No. 5-73721, only pixels corresponding to the table frame line are first selected from black pixels forming characters and tables. Image is generated, and this image is
The image is divided into N in the horizontal direction to form an M × N area, and if even one pixel has a black pixel in each area, a reduced image of size M × N that is one dot of black pixel is created, and the black pixel in the reduced image is By extracting the table frame line from the line and creating a reduced image, there is no break in the line due to broken lines or blurring,
There has been proposed a table recognition device capable of accurately recognizing a table structure.

[0008]

However, in the apparatus described in the above-mentioned Japanese Patent Laid-Open No. 5-73721, the fact that the broken table frame line is not broken in the reduced image is used, so the method of determining the reduction ratio is It becomes a problem.

This is because the optimum value of the reduction ratio depends on the height / width of the table in the original image and the length of breaks in the frame line, and it is difficult to set this automatically.

Further, in the device disclosed in the above-mentioned Japanese Patent Laid-Open No. 5-73721, there is a high possibility that the cells cannot be correctly extracted when the discontinuity of the table frame is large.

Therefore, the present invention solves the above-mentioned problems, measures the width of a table item, and checks the frequency to interpolate the breaks in the table frame lines existing inside the table, thus enabling the interpolation of large breaks. It is an object of the present invention to provide a table recognizing device and method.

[0012]

In order to achieve the above object, the present invention provides a table recognition method for recognizing a table structure from an image obtained by scanning an original document including a table format. Front-outer contour line extracting means for detecting the outer contour of the front-side frame line, and front-side contour line extracting means for extracting one or more front-side contour lines existing in the closed curve extracted by the front-outer contour line extracting means. For each of the extracted inside contour lines, table frame width measuring means for measuring the distance between two pixels belonging to the same inside contour line and existing in the same row or the same column as a table frame width, In the histogram measuring means for obtaining the histogram of the table frame width, and in the histogram obtained by the histogram measuring means, a value having a high frequency is set as a standard table frame width, and a portion corresponding to a value having a low frequency is a table frame line. Providing an estimation unit interruption table border shall cut occurs, and table frame line interpolation means for interpolating the estimated interrupted table border, a table recognition apparatus characterized by comprising a.

In the present invention, the table frame width extracting means measures the distance between two table frame lines in the same column in the vertical direction, and the two table frame widths in the same row. Horizontal table frame width measuring means for measuring the distance between the lines, and estimating the horizontal table frame line break from the histogram obtained by measuring the vertical table frame width, and then interpolating A vertical table frame line break may be estimated and interpolated from a histogram obtained by measuring the table frame width in the direction.

Further, according to the present invention, (a) a step of extracting an outer contour of a table frame line from an image signal obtained by scanning a document including a table, and (b) existence of the outer contour in the extracted outer contour. A step of extracting one or a plurality of front-side inner contours, and (c) between two pixels that belong to the ends on the front-side inner contour line of each of the extracted front-side inner contour lines and face each other through the inner side of the table. A step of measuring the distance of the table frame width in a predetermined direction in a range where the table frame line exists, (d) a step of obtaining a histogram of the table frame width, and (e) in the histogram,
The frame width with the highest frequency is set as the standard table frame width, and the process of estimating that the position corresponding to the frame width with the low frequency is the table frame line break, and (f) the broken table frame line And a step of interpolating, which provides an interruption interpolation method for a table frame line.

[0015]

The principle and operation of the present invention will be described below. In the present invention, the table frame width is measured for one closed area surrounded by the table frame line, the histogram is taken, and the relationship between the table frame width occupied in one closed area and its frequency is obtained. If the table format to be recognized is a collection of rectangular items, the table frame width in one item (cell) is the same regardless of the location. Therefore, one peak appears in the histogram.

However, the table frame line is broken, and a plurality of cells are 1
If you are in one closed region, you will get a histogram with multiple peaks. For example, assuming that the break length of the table frame line is shorter than half the length of the side of the cell, in the measured histogram, the high frequency value is the original table frame width and the low frequency value is the frame line. It is possible to determine that the width of the table frame is measured by the interruption, and the interruption point can be extracted.

[0017]

Embodiments of the present invention will be described below with reference to the drawings.

[0018]

[Embodiment 1] FIG. 1 is a diagram for explaining the configuration of an embodiment of the present invention.

Referring to FIG. 1, front and outer contour line extracting means 1
Detects the outer contour line of the table frame line. It is assumed that the table format handled by the outer-outside contour line extraction means 1 is formed only of the table frame lines in the horizontal / vertical direction or a direction close to the horizontal / vertical lines, and that the outermost table frame lines have no breaks. Further, it is assumed that another cell is not included in the cell and that the cell and the character in the cell do not contact each other.

The outer-surface contour line extracting means 1 will be described in detail with reference to FIGS.

The image data formed by reading a document including a table such as the table format 21 with a scanner or the like is raster-scanned with the upper left pixel in the drawing as a starting point to search for a portion where a white pixel and a black pixel are adjacent to each other. .

When a place where the white pixel changes to the black pixel is found, the position of the black pixel and the fact that the pixel is located at the left end of the table frame line are stored.

Then, starting from this point, the outer contour of the table frame line is traced, and the position coordinates and the relative position of the pixel with respect to the table frame line are sequentially stored. However, the position information stored here may be only when the pixel is located at the end of the table frame line, for example, at the right end or the left end.

With reference to FIG. 3, for convenience of explanation, the table frame line 3
2 has an “L” flag for the pixel group 31 located at the left end and an “R” flag for the pixel group located at the right end (not shown).
Are stored as relative position information.

As a result of tracing, when the position of the starting point is returned, the tracing is ended, the height and width of the area are examined, and if it is sufficiently larger than the maximum character size to be recognized, it is judged to be a table and the next character is recognized. Go to processing.

If it is smaller than the maximum character size, it is determined that it is not a table, and scanning is performed for pixels that have not been scanned.

Now, when the outer-outer contour line is detected by the outer-outer contour line extracting means 1, there is always one or more inner contours of the front-frame lines inside.

The inside-outside contour extracting means 2 stores the position of a black pixel forming the inside-outside contour and the relative position of the pixel with respect to the outside border of the frame in the same manner as tracing the outside-side contour. That is, a predetermined identification flag is set for the pixel group forming the end of the front inside contour.

Next, referring to FIG. 4, the table frame width measuring means 3
Will be described. The processing after this processing is applied to each of the extracted front inside contours.

The front frame width measuring means 3 obtains the horizontal distance of the front inner contour, that is, the "front frame width". In each horizontal pixel row, the rightmost pixel 41 among the pixels of the inside contour to which the “R” flag is added, and the “L” that is located on the same horizontal pixel as the pixel 41 and is located on the leftmost side. '
The distance to the pixel 42 (which comes to a position facing the pixel 41) of the front inside contour to which the flag is added is defined as a table frame width 43, and this table frame width 43 is set to each horizontal pixel column in the range where the table frame line exists. Ask for. That is, the table frame width 43 is sequentially obtained in the direction of water pouring of the table contour.

The histogram measuring means 4 takes a histogram (frequency distribution) for the table frame width extracted by the table frame width measuring means 3.

FIG. 5 shows an example of a histogram of the table frame width corresponding to the table format shown in FIG.

Referring to FIG. 5, the table frame line break estimation means 5
Is a high frequency part 5 from the extracted histogram.
1 is the standard table frame width, and it is determined that the infrequent portion 52 having a larger frame width than the standard table frame width is the frame width generated by the frame line break.

Further, as shown in FIG. 5, when the high frequency portion 51 appears on the left side of the low frequency portion 52, the table in the direction orthogonal to the table frame width measurement direction (horizontal direction in FIG. 4) is displayed. It is presumed that the border is broken.

FIG. 6 shows an example of a table format in which the horizontal table frame line is broken. When the table frame width of the tabular form 61 is measured in the horizontal direction, the histogram shows that the maximum frequency is 7 as shown in FIG.
A value 72 smaller than the standard table frame width of 1 is detected from the break position. In FIG. 7, when the high frequency portion 71 appears on the right side of the low frequency portion 72, it is estimated that the table frame line is broken in a direction parallel to the table frame width measurement direction (horizontal direction in FIG. 6). To do.

Therefore, in the case where the table frame line is interrupted in the direction orthogonal to the table frame width measuring direction, the table frame line interpolating means 6 has other portions of the table frame width in the horizontal pixel row corresponding to the frame width which is less frequent. Write black pixels in the original image so that

In the case where the table frame line in the table frame width measurement direction is interrupted, in the horizontal pixel row having a table frame width that is infrequent, the pixel from the pixel to which the'R 'flag is added to the pixel to the'L' flag is changed. The table frame line can be interpolated by filling the added pixels with black pixels.

In the present embodiment, the method of measuring the width of the table frame in the horizontal direction has been described. However, a flag (that is, a flag for identifying the upper end and the lower end) is set on the inside contours located in the vertical direction of the table frame line. Even if the table frame width is measured in the vertical direction, the table frame line break can be interpolated by the same method.

In this embodiment, the amount of calculation is kept low on the assumption that the table is not tilted. However, if the table in the document is tilted, first, a predetermined tilt correction in a table format is performed. ,
Then, this embodiment is applied.

[0040]

Second Embodiment Another embodiment of the present invention will be described with reference to FIG.

When the width of the table frame is measured in the horizontal direction, there is a possibility that the histogram for the table format 81, which has a break in the table frame line in the same direction as the table frame and is slightly tilted, does not show a clear peak.

In such a case, the table frame width is measured in the horizontal direction and the vertical table frame line break is interpolated, and then the table frame width is measured in the vertical direction to extract the horizontal table frame line break. .

By repeating vertical interpolation and horizontal interpolation until there is no interruption, it is possible to cope with interruptions such as the intersections of the table frame lines.

Although the present invention has been described with reference to the above embodiments, it is needless to say that the present invention is not limited to the above embodiments and includes various embodiments according to the principles of the present invention.

[0045]

As described above, according to the present invention,
By measuring the width of the table item and checking the frequency, it is possible to detect the break of the table frame line existing inside the table and interpolate the break of the frame line existing inside the table.

In particular, if there is a slight break in the length, it is possible to fill the break even with a local process such as creating a reduced image as in the prior art example. There is a high possibility that it cannot be extracted. On the other hand, according to the present invention, it is possible to interpolate a larger discontinuity by using the feature that the table item is basically rectangular.

Further, according to the present invention (claim 2), the vertical interpolation and the horizontal interpolation are repeated until the discontinuity disappears, so that the discontinuity at the intersection of the table frame lines can be interpolated.

Further, the present invention has an effect that the discontinuity can be interpolated even if the table format is slightly inclined.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an exemplary embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a table in which a table frame line has a break.

FIG. 3 is an enlarged view of an upper left corner portion of the table shown in FIG.

FIG. 4 is a diagram for explaining a method of measuring a table frame width according to an embodiment of the present invention.

5 is a histogram showing a relationship between a table frame line and a table frame width obtained by processing the table shown in FIG. 4 in one embodiment of the present invention.

FIG. 6 is a diagram showing an example of a table in which a horizontal table frame line is interrupted.

FIG. 7 is a histogram showing a relationship between a table frame line and a table frame width obtained by processing the table shown in FIG. 6 in one example of the present invention.

FIG. 8 is a diagram showing an example of a case where the table format is slightly tilted.

[Explanation of symbols]

 21 Table frame line 31 Pixel located on the left side of the table frame line 32 Table frame line 33 Inside table 41 Pixel on the right side of the contour line of the inside contour line 42 Pixel on the left side of the contour line of the inside contour line 43 Table frame width 51 High frequency histogram area 52 Low frequency histogram area 61 Table frame line 81 Table frame line

Claims (4)

[Claims] 1. In a table recognition method for recognizing a table structure from an image obtained by scanning an original document including a table format, an outer-outer contour line extracting means for detecting an outer contour of a table frame line. A front-inside contour extracting unit that extracts one or a plurality of front-inner contours existing in the closed curve extracted by the front-outer contour extracting unit; and the same table for each of the extracted front-inner contour lines. Table frame width measuring means for measuring a distance between two pixels belonging to the inner contour line and existing in the same row or the same column as a table frame width; histogram measuring means for obtaining a histogram of the table frame width; In the histogram obtained by the means, a table frame line breakage estimating means that the frequency is a standard table frame width, and a portion corresponding to a low frequency value is a table frame line breakage, Table recognition apparatus, wherein the table frame line interpolation means for interpolating a constant has been interrupted table border, further comprising a. 2. The vertical frame width measuring means for measuring the distance between two table frame lines existing in the same column, and the table frame width extracting means between the two table frame lines existing in the same row. The horizontal table frame width measuring means for measuring the distance is provided, and the horizontal table frame line break is estimated and interpolated from the histogram obtained by measuring the vertical table frame width. The table recognition device according to claim 1, wherein a vertical table frame line break is estimated and interpolated from a histogram obtained by measuring a frame width. 3. A step of: (a) extracting an outer contour of a table frame line with respect to an image signal obtained by scanning a document including a table; and (b) one or more existing within the extracted outer contour. And (c) for each of the extracted front-side inner contour lines, displaying a distance between two pixels that belong to an end portion on the front-side inner contour line and face each other through the front-side inside. As a frame width, a step of measuring in a predetermined direction within a range in which a table frame line exists, (d) a step of obtaining a histogram of the table frame width, and (e) in the histogram, the frame width with the highest frequency is the standard. It is characterized by including a step of estimating a portion corresponding to a frame width having a low frequency as a frame width with a low frequency, and (f) interpolating the broken table frame line. Interpolation interpolation method for table frame lines. 4. The table frame line according to claim 3, wherein a predetermined flag for identifying an end portion of the pixel forming the end portion of the extracted table frame line is provided. Interruption interpolation method.