JP3409992B2 - Black and white reversal area identification device and table recognition device using the device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a black-and-white reversal area identifying device used when inputting tabular image data such as a form with an input device such as a scanner and performing table recognition processing on the image data. A table recognition device using the device.

[0002]

2. Description of the Related Art In recent years, an input device such as a scanner has been widely used as a device for supplying characters on a document or a form form to a computer as image data. Further, image data including a table supplied from such an input device is recognized as a character or ruled line by the image recognition device,
For example, it is stored in the computer as a corresponding character code or vector data.

In the image recognition apparatus as described above, in order to recognize a table having a relatively thin line such as a ruled line and having long lines in the vertical and horizontal directions, a table recognition apparatus using run in particular (Japanese Patent Laid-Open No. 6-33790). ) Is proposed. This device
As shown in FIG. 34, from the image input unit 101 to the image area 10
The table image is input to 2, and the solid line run extraction unit 103 uses runs in the vertical and horizontal directions with respect to the input image to extract line segments of a certain length or more and connect them to form a ruled line. To extract. Also, among the extracted ruled lines,
A portion surrounded by a ruled line having a width equal to or larger than a threshold value is output to the black / white inversion cell candidate extraction unit 104 as a black / white inversion candidate. Then, in the black / white inversion cell determination unit 105, if the density of black pixels inside the extracted black / white inversion cell candidate is, for example, 50% or more, the corresponding area (cell) is recognized as a black / white inversion cell,
The pixel inversion unit 106 performs black and white inversion processing on the cell.

After the black-and-white inversion processing described above, character cutting processing is performed on the data written in the black-and-white inversion area, and recognition processing is performed on the cut out characters.

[0005]

The above-described conventional table recognition device has the following problems. (A) First, in the table recognition device of the conventional example, as shown in FIG. 35 (a), long horizontal runs and vertical runs are used, and therefore, as shown in FIG. If it is composed of images, there is no problem, but for example, FIG.
As shown in (1), there is a problem when the black and white reversal area and the normal cell area are mixed in one cell. That is, FIG.
When the number of black pixels is counted using the horizontal run and the vertical run of the length shown in FIG. 9A, the number of white pixels is extremely large, and the number of black pixels cannot be accurately counted. Therefore, the black-and-white inversion area is erroneously identified. (B) In the conventional table recognition device, as shown in FIG. 35 (a), a long horizontal run and a vertical run (for example, a length equal to the length of the black-and-white reversal area shown in FIG. Lateral run,
And vertical run), it is not possible to accurately count the number of black pixels when recognizing black-and-white inversion areas in which the horizontal and vertical lengths of the cells to be recognized are different. For example, when the length of the horizontal run is L1 and the length of the vertical run is L2 as shown in FIG. 8A, the horizontal width of the black-and-white reversal cell to be recognized is longer than the above L1 and the vertical width is the above. When the length is longer than L2, the number of white pixels is large and the black-and-white inversion area cannot be accurately identified.

The same thing occurs when the character size in the black-and-white inversion cell area is large. That is, in this case, the character size of the blank characters of the character string “company name” shown in FIG. 7B is large, and when the number of black pixels is counted according to the horizontal run or the vertical run, the blank character part is large,
Since the count value of the number of black pixels becomes small, it is not possible to accurately recognize the black and white inversion area (black and white inversion cell). (C) Further, when the reading process of the front image such as the form form by the scanner is performed in a tilted state, for example, the black-and-white reversal region has a tilt, and a long horizontal run without tilt and a vertical run are used. Counting the number of black pixels by counting the wrong (smaller) number of black pixels results in erroneous recognition of the black-and-white inversion cell.

In such a case, for example, it is necessary to perform the counting process by inclining the horizontal run and the vertical run by the same angle.
However, such tilt correction requires complicated processing. In order to solve the above problems, the present invention uses a short line segment extraction mask in the horizontal direction and a short line segment extraction mask in the vertical direction, counts the number of black pixels included in the line segment extraction mask, and performs integrated processing. , If the line width of the integrated line segment exceeds a predetermined threshold value, the black-and-white inversion candidate region is created, and the black-and-white inversion candidate region is created in the horizontal direction and the vertical direction. By doing,
(EN) A black and white reversal area identifying device that solves the above problems (A) to (C), and a table recognition device that uses the device.

[0008]

In order to solve the above-mentioned problems, the invention described in claim 1 scans the binary image data in the table format in the horizontal direction using a line segment extraction mask long in the horizontal direction, A region extracting unit that scans in the vertical direction using a line segment extraction mask that is long in the vertical direction, and extracts regions each having a black pixel density higher than a predetermined threshold value; and a region having a high black pixel density extracted by the region extracting unit. And a first unifying means for unifying the areas in a direction orthogonal to the respective mask scanning directions, and if the line width of the area integrated by the first unifying means is equal to or larger than a predetermined threshold value, it is regarded as a black and white reversal candidate area. Second integration means for integrating the black-and-white inversion candidate areas in the respective mask scanning directions, and line segment segments in the horizontal and vertical directions are created for the contours of the black-and-white inversion candidate areas integrated by the second integration means. Line segment And a black and white reversal area setting means for setting an area surrounded by the line segment segment row in the horizontal direction and the line segment segment row in the vertical direction created by the line segment segment row creating means as a black and white inversion area. It can be achieved by providing a black-and-white reversal area identification device with.

That is, in this example, the horizontal length is long, and vertical × horizontal (m × n) is, for example, 1 × 6, 1 × 15, 1 × 30, etc.)
The line segment extraction mask is used to scan in the horizontal direction at regular intervals, and when the rectangular region cut out by the line segment extraction mask contains black pixels of a predetermined threshold value or more, the region is The black and white inversion candidate area is integrated if it is integrated in a direction orthogonal to the horizontal direction and the line width is equal to or larger than a predetermined threshold value as a result of the integration, and the area (line segment) of the black and white inversion candidate area is made horizontal by the second integration means. Extend in the direction.

On the other hand, the same processing is performed also in the vertical direction, and m × n (for example, 6 × 1, 15 ×) that is long in the vertical direction is performed.
1, 30 × 1, etc.) line segment extraction mask is used, and scanning is sequentially performed in the vertical direction for every fixed dot, and black pixels of a predetermined threshold value or more are included in the region cut by the line segment extraction mask. At this time, the regions (line segments) are integrated in a direction orthogonal to the vertical direction, and if the line width obtained as a result of the integration is a line width equal to or larger than a predetermined threshold value, it is set as a black-and-white reversal candidate region, and the second integrating means is further included. The line segment of the black-and-white inversion candidate area is extended in the vertical direction.

Then, with respect to the black-and-white reversal candidate regions created in the horizontal direction and the vertical direction, line segment segment lines showing respective contours are created in the horizontal direction and the vertical direction, and the line segment segments created in this way The area surrounded by columns is the black-and-white inversion area.

As described above, the invention according to claim 1 uses the line segment extraction mask which is extremely thin in the horizontal direction (for example, 1 dot) and is relatively short (for example, 6 dots, 15 dots, 30 dots), and Very thin in the vertical direction (eg 1
(Dot) and a relatively short line segment extraction mask (for example, 6 dots, 15 dots, 30 dots) is used to identify the black and white reversal area by integrating the areas with a large number of black pixels. Even if a black-and-white reversal area exists in a part of the area, the area can be accurately identified (solution to the above-mentioned problem (a)). Even if a large-sized inverted character exists in the black-and-white inversion area, the line segment extraction mask used in this example uses an extremely thin and relatively short line segment extraction mask as described above, so Can be identified (solution of the above problem (b)). Further, even if the front image is read with an inclination, the black-and-white inversion area is identified by the integration process while recognizing the number of black pixels in a narrow range, so that it can be recognized as an inclined line segment, and the black-and-white inversion in an inclined state can be recognized. It can be recognized as an area (solution of the above-mentioned problem (C)).

In order to solve the above-mentioned problems, the invention according to claim 2 scans the binary image data in the table format in the horizontal direction using a line segment extraction mask which is long in the horizontal direction and long line segments in the vertical direction. A region extracting unit that vertically scans by using the extraction mask and extracts regions in which the black pixel density is higher than a predetermined threshold, and the regions with high black pixel density extracted by the region extracting unit in respective mask scanning directions. If the line width of the area integrated by the first integration means and the area integrated by the first integration means is greater than or equal to a predetermined threshold value, it is determined as a black and white inversion candidate area, and each of the black and white inversion candidate areas is Second unifying means for unifying in the mask scanning direction, and a black and white reversing area is set according to the overlapping state of the horizontally reversing black and white reversing candidate area and the vertical black and white reversing candidate area integrated by the second unifying means. It can be achieved by providing a rolling tone reversal region identification apparatus provided with an area setting means.

That is, in this embodiment, the invention according to claim 1 creates a line segment segment sequence for the contour of the black / white inversion candidate region and identifies the region surrounded by the line segment segment sequence as the black / white inversion region. On the other hand, the invention of the present embodiment has a configuration in which the black-and-white inversion area is identified by the overlapping state of the horizontal black-and-white inversion candidate area and the vertical direction black-and-white inversion candidate area.

That is, also in this example, a line segment extraction mask of m × n which is long in the horizontal direction is used, and scanning is sequentially performed in the horizontal direction for every fixed dot, and a predetermined threshold value is set in a rectangular region cut by the line segment extraction mask. When the above black pixels are included, the rectangular areas are integrated in the direction orthogonal to the horizontal direction, and if the result of the integration is that the line width is equal to or greater than a predetermined threshold value, the black and white inversion candidate area is set, and the vertical direction is set. The same process is performed on the other hand, a line segment extraction mask that is long in the vertical direction is used, and scanning is performed in the vertical direction sequentially for every fixed dot, and a black value equal to or larger than a predetermined threshold value is drawn in the rectangular area cut by the line segment extraction mask. When pixels are included, the rectangular areas are integrated in the direction orthogonal to the vertical direction, and if the result of the integration is that the line width is equal to or greater than a predetermined threshold value, the black and white inversion candidate area is selected.

Then, with respect to the black-and-white inversion candidate areas created in the horizontal direction and the vertical direction, the overlapping state is judged, and the black-and-white inversion area is identified. For example, in this overlapping state, one area (either the horizontal direction or the vertical direction) is defined as the area area where the black and white inverted areas in both the horizontal and vertical directions overlap.
When the value divided by the area of the black-and-white inversion area in the direction is equal to or larger than the predetermined threshold value, the area surrounded by the black-and-white inversion areas in both the horizontal and vertical directions is set as the black-and-white inversion area. With this configuration, the black-and-white reversal area can be accurately identified, and the problems (a) to (c) can be solved.

The identification of the black and white reversal area is not limited to the above example, and the overlapping sum of the black and white reversal candidate areas in the horizontal direction and the vertical direction is obtained, and when the sum is a predetermined value or more, the black and white reversal candidates in both the horizontal and vertical directions. The area surrounded by the areas may be the black-and-white inverted area, and various other methods are possible.

The third aspect of the present invention is to embody the invention of the first or second aspect. For example, it is determined that the line width of the area integrated by the first integrating means is equal to or less than a predetermined threshold value. In doing so, it has a ruled line judging means for judging the area integrated by the first integrating means as a ruled line.

That is, the ruled line has a relatively thin line width,
When it is determined that the line width of the area integrated by the first integrating means is equal to or larger than a predetermined threshold value, it is considered as the black-and-white reversal candidate area as described above, but when it is determined that the line width is equal to or smaller than the predetermined threshold value, the area is determined to be the same. Judge as a ruled line.

With this configuration, according to the present example, the black-and-white inversion area can be identified and the ruled line can be extracted at the same time, so that the black-and-white inversion area identifying apparatus can be made extremely efficient.

The description of claim 4 further makes the description of claim 3 more concrete. For example, ruled line creation for creating ruled lines by integrating the ruled lines determined by the ruled line determination means in the respective mask scanning directions. It is a configuration having means.

With this configuration, when recognizing the tabular image data such as a form, the ruled lines can be recognized together with the black-and-white inversion area (black-and-white inversion cell), and the black-and-white inversion area identification processing ends. At the same time, the ruled lines of the form form are also completed at the same time, which makes it possible to provide an extremely efficient black and white reversal area identification device.

According to a fifth aspect of the present invention, which is more specific than that of the fourth aspect, when the ruled line creating means creates vertical ruled lines, for example, the line segment segment string creating means creates the line segments. Horizontal line segment row,
The area surrounded by the ruled lines in the vertical direction is a black and white reversal area.

The line segment sequence is the outline of the black-and-white inversion candidate area in the horizontal direction. For example, the area surrounded by the horizontally-inverted black-and-white inversion candidate area and the vertical ruled line is the black-and-white inversion area. As described above, in this example, the area surrounded by the line segment segment sequence (black and white inversion candidate area) and the ruled line is identified as the black and white inversion area.

[0025] Further, the sixth aspect of the invention is a concrete example of the invention of the fourth aspect, and when the ruled line setting means sets a ruled line in the horizontal direction, for example, the line segment segment sequence creating means. The black and white inversion area is the area surrounded by the line segment segments in the vertical direction created in 1 above and the ruled lines in the horizontal direction.

Contrary to the above-mentioned claim 5, the description of the present example is such that the line segment segment sequence is the outline of the black-and-white inversion candidate region in the vertical direction, and is surrounded by this line segment segment sequence and the horizontal ruled line. The area that is marked is identified as the black-and-white reversal area.

The description of claim 7 is specific to the description of claim 1, 5 or 6, and a margin for absorbing the irregularities of the contour of the black-and-white reversal region is added to the line segment segment row. It is a configured structure.

That is, in this example, when line segment extraction processing is performed using the above line segment extraction mask, the individually extracted regions are
For example, it has an unevenness in the up-down direction or the left-right direction. In this example, a predetermined number of dots are added to the unevenness in the outer direction of the black-and-white inversion area (outer direction of the normal vector of the line segment) to obtain the outline of the black-and-white inversion area. This is a structure that absorbs the unevenness of.

The eighth aspect of the present invention embodies the invention of the first or second aspect. The black-and-white reversal area setting means sets horizontal when the number of connected components is equal to or less than a predetermined threshold value. In this configuration, the area surrounded by the line segment segment row in the vertical direction and the line segment segment row in the vertical direction is determined as the black and white reversal area.

Therefore, in the case of the present embodiment, even if the black-and-white inversion area is identified by the invention according to claim 1, the black-and-white inversion area is not determined unless the number of connected components in the black-and-white inversion area is less than a predetermined threshold value. is there. Here, the number of connected components indicates how many black print portions or white print portions are connected in the black-and-white inversion candidate area. For example, a large number of black dots such as halftone dots in the black-and-white inversion candidate area. When dots are included, the influence of halftone dots and the like is eliminated by limiting the number of connected components.

The description of claim 9 makes the description of claim 8 more concrete, and, for example, the calculating means for calculating the number of connected components and the number of connected components calculated by the calculating means are predetermined. And a discriminating means for discriminating that it is equal to or less than the threshold value.

Here, the calculation of the number of connected components by the calculation means is, for example, a calculation for calculating an independent black print portion in the black-and-white reversal region, and, for example, when a halftone dot or the like is provided, the number of black dots is calculated. This is the configuration. On the other hand, the identifying means is means for identifying whether the number of connected components calculated by the calculating means is less than or equal to a preset threshold value. For example, when a pattern such as a halftone dot is added to the black-and-white inversion area, the identifying means The cell to which the pattern is added and the black-and-white inverted cell are surely identified.

According to a tenth aspect of the present invention, the invention according to the fourth aspect is more concretely described, in which the area extracting means extracts an isolated ruled line candidate, and the black-and-white is provided around the isolated ruled line candidate. When there is a reversal candidate area, the isolated ruled line candidate also has an isolated ruled line unifying means for unifying it with the black and white reversal candidate area.

That is, as a result of integrating the areas having a high black pixel density in the direction orthogonal to the mask scanning direction, if the line width of the areas is equal to or larger than a predetermined threshold value, it is extracted as a white / black inversion candidate area. However, when the black and white inversion candidate area exists around the ruled line candidate, it is originally a black and white inversion candidate area.
For example, it may be erroneously extracted as an isolated ruled line candidate due to "blurring" or the like. Therefore, in order to relieve such a case, the isolated ruled line integrating means integrates the black and white inversion candidate area.

With this configuration, even if there is thin black print such as "blurring" in the black-and-white reversal area, it can be surely detected, corrected and identified as the black-and-white reversal area.

The description of claim 11 is more specific than that of claim 10, wherein the extraction of the isolated ruled line candidates is provided with a means for distinguishing between the character area and the ruled line area, and the ruled line extraction processing is performed. The configuration is such that the black-and-white inversion area is extracted later.

That is, in the ruled lines extracted by the ruled line extraction, a character string or the like may be erroneously recognized as a ruled line in addition to the original ruled line. In such a case, the character area and the ruled line area are distinguished from each other to ensure the ruled line. The ruled lines are extracted.

The twelfth aspect of the invention is a more specific version of the invention of the first aspect, wherein the line segment segment string has an attribute indicating that the line segment segment string is a black / white inverted area. Is added.

The addition of an attribute to the line segment segment sequence is as follows.
For example, it is realized by setting a dedicated flag in the memory for storing the line segment segment sequence and storing the segment sequence indicating the black / white inversion area.

According to a thirteenth aspect of the present invention, the statement of the twelfth aspect of the present invention is made more concrete. In the line segment segment row, the position of the line segment segment row with respect to the black-and-white inversion region is vertical or horizontal. It is a configuration added as position information of the side of.

For example, in the case of two line segment segment rows in the horizontal direction, the position information above or below is added to the line segment segment row depending on the position with respect to the black and white reversal area. Also, in the case of two line segment segments in the vertical direction, right or left position information is added depending on the position with respect to the black and white reversal area.

With this configuration, the position of the created line segment segment sequence with respect to the black-and-white inversion area becomes clear, and it can be easily performed, for example, when distinguishing the black-and-white inversion area from the normal cell area.

The statement of claim 14 makes the statement of claim 13 more concrete. When analyzing the cell structure of the table, the position information of the cell is added to the line segment sequence. If it is the same as the position information of the upper and lower sides or the left and right sides, it is determined to be a black and white reversal area, and if it is different, it is determined to be a normal cell area.

This example is a specific application example of the thirteenth aspect, and with the above configuration, it is possible to easily distinguish between the black / white inversion area and the normal cell area. In addition to the above, the vertical and horizontal position information added to the line segment sequence can be used, for example, for calculating the number of black and white reversal areas included in the table, the size of the black and white reversal areas, and the like.

The description of claim 15 relates to claim 1 or 2 above.
When the black-and-white inversion area is composed of a plurality of cell areas, the ruled line connecting to the black-and-white inversion area is extended to extract the cell area in the black-and-white inversion area. It is a composition.

For example, when the black-and-white inversion area is formed in the horizontal direction, the ruled line in the vertical direction is extended into the black-and-white inversion area, and the black-and-white inversion area is divided into, for example, strips and separated by the vertical ruled line. The cell area is defined as the cell area.

With this structure, the black and white reversal area can be easily divided into cells. The description of claim 16 makes the description of claim 15 more specific, and when dividing the cells in the black-and-white inversion area, the area in which the ruled line connected to the black-and-white inversion area is extended is subdivided. This is a configuration having means for obtaining the black pixel density.

That is, when dividing the black-and-white inversion area into cells according to the above-mentioned claim 15, for example, when extending the vertical ruled lines into the black-and-white inversion area and dividing the black-and-white inversion area into strips, the ruled lines are If a black-and-white inverted character exists at the position of the extended black-and-white inverted area, this position cannot be separated by a ruled line, so this is a process performed to confirm the existence of the black-and-white inverted character.

Therefore, for example, a line segment extraction mask that is long in the vertical direction is used, and the black pixel number is extracted in the target black-and-white inversion region by the line segment extraction mask. When the number is large, it can be seen that a black-and-white inverted character (white character) exists at the position and the position cannot be divided into cells.

With this configuration, the black and white inversion area can be accurately divided into cells. Claim 17
In order to solve the above-mentioned problems, the described invention scans binary image data in a table format in the horizontal direction using a line segment extraction mask long in the horizontal direction, and vertically scans it using a line segment extraction mask long in the vertical direction. Area extraction means that scans each direction and extracts each area having a black pixel density higher than a predetermined threshold value, and the area having a high black pixel density extracted by the area extraction means is integrated in a direction orthogonal to each mask scanning direction. If the line widths of the first integrating means and the area integrated by the first integrating means are equal to or larger than a predetermined threshold value, it is determined that the area is a black / white inversion candidate area.
Second integration means for integrating the black-and-white inversion candidate areas in the respective mask scanning directions, and line segment segments in the horizontal and vertical directions with respect to the outline of the black-and-white inversion candidate areas integrated by the second integration means. A line segment segment string creating means to be created, and a region surrounded by the line segment segment string in the horizontal direction and the line segment segment string in the vertical direction created by the line segment segment string creating means is set as a black and white reversal area. A black and white reversal area identifying section including setting means, a black and white reversal processing section for performing a black and white reversal process on the black and white reversal area identified by the black and white reversal area recognizing section, and a reversal process by the black and white reversal processing section. This can be achieved by providing a table recognition device having a recognition means for performing recognition processing on a character string located in a region.

The present invention is a table recognition device using the black-and-white reversal area identification device according to claim 1. That is, using a line segment extraction mask that is long in the horizontal direction and a line segment extraction mask that is long in the vertical direction, respectively extract regions where the black pixel density is higher than a predetermined threshold value, and for the region where the black pixel density is high,
Integration in the direction orthogonal to the mask scanning direction, and further integration in the mask scanning direction, for example, forming a pair of line segment segments in the outline of the black / white inversion candidate area, and black / white inversion of the area surrounded by the line segment segments A black-white reversal processing unit performs black-and-white reversal processing on the area set in the area, and a recognition unit that further performs recognition processing on a character string located in the reversal-processed area by the black-and-white reversal processing unit. It is a table recognition device having.

With this configuration, it is possible to provide a table recognition device that performs black-and-white inversion processing on a correctly identified black-and-white inversion area and performs character recognition processing on a character string in the black-and-white inversion area. In this case, the character string subjected to the black-and-white reversal processing is accurately printed in black, so that reliable table recognition processing can be performed.

In order to solve the above-mentioned problems, the eighteenth aspect of the present invention scans the binary image data in the table format in the horizontal direction using a line segment extraction mask that is long in the horizontal direction to obtain long line segments in the vertical direction. Scan vertically using the extraction mask,
Area extracting means for extracting areas each having a black pixel density higher than a predetermined threshold, and first integration for integrating the areas having a high black pixel density extracted by the area extracting means in a direction orthogonal to each mask scanning direction. Means, and if the line width of the area integrated by the first integrating means is equal to or larger than a predetermined threshold value, it is determined to be a black / white inversion candidate area, and the black / white inversion candidate area is integrated in each mask scanning direction. Means and the second
And a black-and-white inversion area identifying unit having a black-and-white inversion area setting means for setting the black-and-white inversion area according to the overlapping state of the horizontal black-and-white inversion candidate area integrated by the integration means of A black-and-white inversion processing means for performing a black-and-white inversion processing on the black-and-white inversion area identified by the area recognition section; and a recognition means for performing a recognition processing on a character string located in the area inverted by the black-and-white inversion processing means. This can be achieved by providing a table recognition device having

The present invention is a table recognition device using the black-and-white reversal area identification device according to claim 2, wherein
In contrast to the table recognition device described in No. 7, a black-and-white inversion area identification device is used which sets a black-and-white inversion area according to the overlapping state of the horizontal black-and-white inversion candidate area integrated by the second integrating means and the vertical black-and-white inversion candidate area. It is a table recognition device.

With this configuration, it is possible to provide a table recognition device that performs black-and-white inversion processing on a correctly identified black-and-white inversion area and performs character recognition processing on a character string in the black-and-white inversion area. In this case, the character string subjected to the black-and-white reversal processing is accurately printed in black, so that reliable recognition processing can be performed.

A nineteenth aspect of the invention embodies the invention of the seventeenth aspect of the invention, wherein the black and white reversal processing means fills out the contour segments of the black and white reversal area, and the contour segment string. It is configured to include processing means for inverting the black and white inversion areas other than the above.

That is, in this example, the outline portion of the black-and-white reversal area is filled with, for example, black to frame the outline portion of the black-and-white inversion area, and the black-and-white inversion area therein is inverted by the processing means.

A twentieth aspect of the present invention is a concrete description of the nineteenth aspect of the present invention, wherein the black and white reversal processing means has a cell area filling means for filling an area divided by cells.

Also in this case, the dividing line for dividing the cells in the black-and-white inversion area is filled with, for example, black so that the cell dividing line can be seen when the black-and-white inversion area is inverted.

The twenty-first aspect of the present invention is a concrete example of the invention of the seventeenth or eighteenth aspect of the present invention, in which the line width of the character string located in the black-and-white inversion area inverted by the black-and-white inversion processing means is determined. The character string is calculated, and the character string is expanded or contracted according to the calculated line width to shape the character string.

With this configuration, the abnormal line width of the character string that may occur due to the black-and-white reversal process is corrected and the character string is shaped into an accurate character string. A twenty-second aspect of the invention embodies the invention of the seventeenth or eighteenth aspect, wherein the black-and-white inversion area setting means sets the black-and-white inversion area when the number of connected components is less than or equal to a predetermined threshold value. It is a configuration for performing processing.

This example has a configuration corresponding to the above-mentioned claim 8, and the configuration is applied to a table recognition device.
That is, when setting the black-and-white inversion area, the number of connected components is calculated, and when the number of connected components is within a predetermined threshold value, the black-and-white inversion area cell can be surely identified to determine the black-and-white inversion area cell. On the other hand, it is possible to provide a table recognition device capable of performing recognition processing such as character recognition in a state where the inversion processing is accurately performed.

The twenty-third aspect of the present invention is a concrete example of the seventeenth or eighteenth aspect of the invention, in which the area extracting means extracts an isolated ruled line candidate and surrounds the isolated ruled line candidate with the isolated ruled line candidate. When there is a black-and-white reversal candidate area, the isolated ruled line candidate is also integrated with the black-and-white reversal candidate area.

This example has a configuration corresponding to the above-mentioned claim 10, and the configuration is applied to a table recognition device. That is, as a result of integrating the regions with high black pixel density in the direction orthogonal to the mask scanning direction, if the line width of the regions is less than or equal to a predetermined threshold value, it is extracted as a ruled line candidate, and the black-and-white inversion candidate is extracted around the ruled line candidate. When an area exists, it is set as a black and white inversion candidate area. With this configuration, for example, even when it is determined as an isolated ruled line candidate due to “blurring”, the isolated ruled line integration unit integrates the ruled line candidate into the black and white inversion candidate area, and the inversion processing is performed on the accurate area of the black and white inversion area. It is possible to provide a table recognition device that can perform recognition processing such as character recognition in the applied state.

The statement of claim 24 is a concrete example of the invention of claim 17, wherein the line segment segment string has, for example, an attribute indicating that the line segment segment string is a black / white inversion area. Is added.

This example has a configuration corresponding to the above-mentioned claim 12, and the configuration is applied to a table recognition device. That is, the line segment segment sequence corresponds to the black / white inversion region, and the position of the line segment segment sequence is added by adding an attribute such as up / down or left / right as a position attribute to the line / segment inversion region. Therefore, it is possible to provide a table recognizing device that facilitates the table recognizing process because the distinction between the black-and-white inverted cell and the normal cell is simplified.

The twenty-fifth aspect of the present invention is a concrete example of the invention of the seventeenth or eighteenth aspect, and when the inside of the black-and-white inversion area is composed of, for example, a plurality of cell areas, it is connected to the black-and-white inversion area. The ruled line is extended to extract the cell area in the black-and-white inverted area.

This example has a configuration corresponding to the above-mentioned claim 15, and the configuration is applied to a table recognition device. That is, when a plurality of cell areas are present in the black-and-white inversion area, the black-and-white inversion area is divided into a plurality of cells by using ruled lines whose extension lines connect to the black-and-white inversion area in a strip shape.

With this configuration, even if there are a plurality of cell areas in one black-and-white inversion area, the cell area can be surely divided, and the correctly divided cell area can be recognized. A table recognition device capable of performing processing is provided.

A twenty-sixth aspect is a concrete embodiment of the invention according to the twenty-fifth aspect, and when dividing cells in the black-and-white inversion area, for example, an area in which a ruled line connected to the black-and-white inversion area is extended. Is subdivided to obtain the black pixel density.

This example describes one method for dividing cells in the black-and-white inversion area. According to this method, a table recognition device that prevents a character string in the black-and-white inversion area from being mistakenly cell-divided. Can be provided. The method of dividing the black and white inversion area into cells is not limited to the above example.

In order to solve the above-mentioned problem, the invention according to claim 27 scans the binary image data in the table format in the horizontal direction by using a line segment extraction mask long in the horizontal direction, and long line segments in the vertical direction. Scan vertically using the extraction mask,
A function of extracting each area having a black pixel density higher than a predetermined threshold value, a function of integrating the areas having a high black pixel density extracted in the function in a direction orthogonal to each mask scanning direction, and an integration of the function If the line width of the area is greater than or equal to a predetermined threshold value, it is determined to be a black-and-white inversion candidate area, and the function of integrating the black-and-white inversion candidate area in each mask scanning direction and the outline of the black-and-white inversion candidate area integrated by the function , The ability to create horizontal and vertical line segment segments,
The computer storing a program including instructions for causing the computer to execute a function of setting a region surrounded by a horizontal line segment segment sequence and a vertical line segment segment sequence created by the function as a black and white inversion region, This can be achieved by providing a readable storage medium.

The present invention has a structure in which the program for performing the above processing is stored in a storage medium such as a hard disk, an IC memory card, a magnetic tape, a floppy disk, or an optical disk. When the program is executed, the program is read from the storage medium. To execute.

In order to solve the above-mentioned problem, the twenty-eighth aspect of the present invention scans the binary image data in the table format in the horizontal direction by using a line segment extraction mask long in the horizontal direction, and long line segments in the vertical direction. Scan vertically using the extraction mask,
A function of extracting each area having a black pixel density higher than a predetermined threshold value, a function of integrating the areas having a high black pixel density extracted in the function in a direction orthogonal to each mask scanning direction, and an integration of the function If the line width of the area is greater than or equal to a predetermined threshold value, it is determined to be a black-and-white inversion candidate area, and a function of integrating the black-and-white inversion candidate area in each mask scanning direction and a horizontal black-and-white inversion candidate area integrated by the function This can be achieved by providing the computer-readable storage medium that stores a program including an instruction for causing a computer to execute a function of setting a black-and-white inversion region depending on an overlapping state of black-and-white inversion candidate regions in the vertical direction.

The present invention also has a configuration in which the program for performing the above processing is stored in a storage medium such as a hard disk, an IC memory card, a magnetic tape, a floppy disk, or an optical disk. When the program is executed, the program is read from the storage medium. To execute.

Further, the description of claim 29 specifically shows the meaning of the black-and-white inversion area of the invention of claim 1 or 2, and the black-and-white of the black-and-white inversion area is created by three elements of color. It is possible to consider various combinations such as white and red, white and blue, yellow and red, yellow and green, and the like.

Further, similarly, the statement of claim 30 concretely shows the meaning of the black-and-white inversion area of the invention of claim 17 or 18, and the black-and-white of the black-and-white inversion area is defined by three elements of color. It includes any two colors to be created, and various combinations such as white and red, white and blue, yellow and red, and yellow and green are conceivable.

[0078]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a system configuration diagram of a table recognition device including the black-and-white inverted region recognition device of the present embodiment. This example extracts m × n in order to extract the black and white reversal area in the table used for spreadsheet software or form form, for example.
The line segment extraction mask of is used to locally obtain a region having a high black pixel density, and by integrating them, it is possible to extract a locally existing black and white inversion region.

In the figure, 1 is a central processing unit (CPU) for performing various processes, 2 is a program memory for storing programs executed by the CPU 1, and 3 is an image memory for storing image data in a bitmap format. Four
Is a work memory used for image processing, 5 is a scanner for optically reading an image, 6 is a memory for temporarily storing the information read by the scanner 5, 7 is a dictionary file storing the characteristics of each character image, 8 Is a display for displaying the recognition result, 9 is a printer for printing the recognition result, 10 is an interface circuit of the display 8 and the printer 9, 11 is the CPU 1, program memory 2, image memory 3, work memory 4, memory 6, dictionary file 7 , A bus connecting the interface circuit 10 and the driver 12, 13 is a hard disk, 14 is an IC memory card, 15 is a magnetic tape, 16 is a floppy disk, 1
Reference numeral 7 is an optical disk such as a CD-ROM.

The system of the present apparatus temporarily stores the image data read by the scanner 5 in the memory 6, and expands the image data in the image memory 3 in the bitmap format. Then, pattern extraction processing is performed on the binary image data copied from the image memory 3 to the work memory 4. Based on the result, a character image is cut out from the image data read by the scanner 5, the characteristics of the cut out character image and the characteristic data stored in the dictionary file 7 are compared, and the recognition result is displayed. Alternatively, it is output to the printer 9.

In this system, the extraction of the black-and-white reversal area is realized as a function of the CPU 1 which performs processing in accordance with the program stored in the program memory 2. Here, the program for performing the black-and-white reversal area extraction processing is stored in advance in the ROM of the program memory 2. In addition, after loading the program for performing the black-and-white inversion area extraction processing from the storage medium such as the hard disk 13, the IC memory card 14, the magnetic tape 15, the floppy disk 16, or the optical disk 17 into the RAM of the program memory 2,
This program may be executed by the CPU 1.

Further, the program for performing the black and white reversal area extraction processing can be taken out from the communication network via the communication interface. For example, as a communication network connected to the communication interface 19, for example, a LAN (Local Area Network)
k), WAN (Wide Area Network)
k), Internet, analog telephone network, digital telephone network (ISDN: Integral Service D)
It is possible to use a wireless communication network such as digital network), PHS (personal handy system), satellite communication, or the like.

Next, the processing shown in FIG. 2 is based on the program stored in the program memory 2 as described above.
This is a recognition process performed by the CPU 1. In the figure, the CPU
1. First, the image input 20 is processed, and the image data (image data) stored in the image memory 3 is input as described above. Next, the process of line segment extraction 21 for sequentially extracting the image data using the line segment extraction mask is performed. For example, this line segment extraction mask is a dot extraction mask of m × n in the vertical and horizontal directions, and is a mask for extracting the line segment. Therefore, when extracting the line segment in the horizontal direction, it is long in the horizontal direction (for example, 1 × When a line segment in the vertical direction is extracted using a mask of 6, 1 × 30, etc., it is long in the vertical direction (for example, 6 × 1, 3).
0x1 mask is used. Then, using the line segment extraction mask that is long in the horizontal direction or the line segment extraction mask that is long in the vertical direction as described above, the image data is sequentially cut out, and the number of black pixels included in the cut out data is a constant threshold value (threshold value ) When it is more than the above, all are line segments of the black pixel component.

Next, the black-and-white inversion area extraction 22 processing is performed. In this process, line segments that are long in the horizontal direction or line segments that are long in the vertical direction are integrated by the process of the above-described line segment extraction 21, and it is determined whether the integrated region is a ruled line or a black / white inverted region, Extract the black and white reversal area. A predetermined threshold value is used for this determination, and an integrated line segment having a line width equal to or larger than the predetermined threshold value is set as a target line segment of the black-and-white inversion area.

Next, in the cell area extraction 23, among the line segments extracted in the processing of the line segment extraction 21, the line segment which is also subjected to the black-and-white inversion area extraction 22, that is, a predetermined integrated line segment, is determined. An integrated line segment having a line width equal to or smaller than the threshold is set as a ruled line candidate, and a cell area is extracted.

Next, the in-cell item extraction 24 is performed by inverting the items in the cell area extracted by the cell area extraction 23 and the black-and-white inversion area extracted by the black-and-white inversion area extraction 22 in black and white. Processing is performed, and in-cell item extraction processing is performed on, for example, the character string after the inversion processing.

Finally, the character recognition 25 is performed on the item data in the cell. In this processing, for example, the character (bitmap data) in the extracted item is compared with the characteristic data of the character registered in the dictionary 7 in advance, so that the character data having the same characteristic, or the most characteristic is approximated. Character data to be recognized as a recognition character.

FIG. 3 is a system configuration diagram for specifically explaining the above-mentioned processing. In the figure, the image input unit 30
Is for performing the processing of the image input 20 described above with reference to FIG. 2, and the line segment extraction unit 32 similarly uses the line segment extraction 21 of FIG.
Is to be processed. Further, the line segment integration unit 33, ruled line extraction unit 34, black-and-white inversion region determination unit 35, black-and-white inversion region cell determination unit 36, and pixel inversion unit 37 shown in FIG. The corresponding parts.

The image area storage unit 31 of FIG. 3 is a part of the area of the work memory 4 of FIG. <First Embodiment> The first embodiment will be described below using the system configuration described above.

First, the flow chart shown in FIG. 4 explains the processing of this example. Hereinafter, description will be given with reference to FIG. It is assumed that the image memory 3 has already read a document in the form of a table from the scanner 5 and stores corresponding image data via the bus 11.

First, the CPU 1 reads out the image data stored in the image memory 3 and executes the line segment extraction 22 and the black and white inversion area extraction 23. These processes are separately performed in the horizontal direction or the vertical direction. In the description of this example, the horizontal processing, which is the processing on the left side in FIG. 4, will be described.

First, the horizontal line segment extraction process (step (hereinafter referred to as S) 1) is performed using a horizontal line segment extraction mask. As this line segment extraction mask, m × n
However, for example, 1 × 6 (as a line segment extraction mask, 1 × 6
No. 1 mask, 1 × 15 mask, 1 × 30 mask, etc. can also be used.

FIG. 5 is a flow chart for concretely explaining the above-mentioned line segment extraction processing (S1). As shown in the figure, this process is performed by first using the above-mentioned line segment extraction mask, which corresponds to 1 ×
Area 6 is cut out, and it is determined whether the value obtained by dividing the number of black pixels in the area by the mask size (the number of pattern black pixels in the mask / mask size) is larger than a preset threshold value (S
1-1). As a result, when the divided value is larger than the predetermined threshold value, the line segment cut out by the line segment extraction mask (that is, a thin and relatively short line segment) has many black pixels, and in this case, all line segments in the mask are (Black pixel) (S1-
1 is YES, S1-2). On the other hand, when the value obtained by dividing the number of black pixels in the mask by the mass size is smaller than the threshold, the number of black pixels in the area cut out by the line segment extraction mask is small, and it is determined that the cut area is not a line segment ( S1-1 is N
O).

FIG. 6 is a diagram for specifically explaining the above-described processing. For example, a line segment extraction mask designating a region of pixels (dots) in which m × n (vertical × horizontal) is 1 × 6 is used for horizontal scanning. The case where the direction component is extracted will be described. The same figure (a)
Indicates mask areas “1” to “9” which are overlapped with each other, and the line segment in the horizontal direction is extracted by performing mask processing on the image data of FIG. Indicates that By this processing, a large number of extremely thin and short regions (m × n is a 1 × 6 line segment) are extracted in the horizontal direction and the vertical direction.

Next, the line segment integration process (S2) shown in FIG. 4 is executed. A specific example of this line segment integration processing is the determination (S2-1) and integration processing (S2-2) shown in FIG.
Further, during the integration processing, the black-and-white inverted area candidate extraction processing (S3) shown in FIG. 4 is also executed at the same time (the specific processing of the above-described black-and-white inverted area candidate extraction processing (S3) follows the flowchart of FIG. Done). Further, the line segment integration process is a process performed by the line segment integration unit 33 shown in FIG. That is, according to the process (S1) of FIG. 4 described above, it is determined whether the upper and lower line segments or the left and right line segments are in contact with the extracted line segment (S2-1). (S2
-1 is YES), these are integrated (S2-2), and a rectangular area is created (S2-3).

Here, the line segment integrating section 33 has a function as a first integrating means and a function as a second integrating means. First, as a function as the first integrating means,
The vertical direction (vertical direction) of the above line segments is integrated. That is, when a region (line segment) is cut out in the horizontal direction using the line segment extraction mask as described above, another region (line segment) in a direction (vertical direction) orthogonal to the region is integrated. In this case, the distance between the integrated regions (line segments) must be shorter than a predetermined threshold value (thd) (S3-1), and if the distance between the line segments is longer than the threshold value (thd), it is determined that they are in contact with each other. do not do. This is because, in this case, the line segments are too far apart to be regarded as the same line segment. Further, this integration is performed when the lengths of the line segments are the same (S3-2), and when the lengths are different, partial integration is performed (S3-3).

By the above processing, the 1 × 6 area (line segment) in the horizontal direction extends in the vertical direction if the number of black pixels in the area is equal to or larger than a predetermined threshold value, and the area (line If the width of (min) exceeds a predetermined threshold value, it is determined as a black and white reversal candidate area. Then, if it is determined to be the black and white reversal area,
Next, the integrated line segments are integrated in the mask scanning direction.

This processing is the second function of the line segment integration unit 33. In this integration processing, the black-and-white reversal candidate areas adjacent in the mask scanning direction are sequentially integrated. Then, in this way, the black-and-white reversal candidate area extends in the mask scanning direction, and the above-described rectangular area is formed (S2-3).

8 and 9 are diagrams showing the results of the above-mentioned line segment integration processing. 8 shows the case where there is no "blurring" in the integrated area, and FIG. 9 shows the case where there is "blurring" in the integrated area.

First, the example of FIG. 8 will be described. In FIG. 8, a line segment is a line segment in which regions having a line width of a threshold value or more (a threshold value or more) are integrated. That is, the line width obtained as a result of integrating in the vertical direction and the horizontal direction using the line segment extraction mask of m × n which is long in the horizontal direction is equal to or larger than the threshold value (threshold value (thd1) shown in the figure). Minutes, which is a black and white reversal candidate area. If the size of the m × n line segment extraction mask that is long in the horizontal direction is M, for example, the size relationship shown in FIG.

On the other hand, FIG. 9 shows an example in which the "blurred" portion is included in the integrated area as described above.
“,” Is a line segment obtained by integrating regions having a line width equal to or larger than the threshold value (threshold value (thd1) shown in FIG. 6). These line segments are also line segments of the black-and-white inversion candidate region. The line segment of "-" is a state in which the portion corresponding to the "blurred" portion is deleted as compared with the line segment of FIG. The reason for this is that the integration line segment distance in the vertical direction is insufficient during the integration process described above, and the integration process in the vertical direction is not performed thickly (S3-1 is (> thd)), resulting in a thin integration line segment. It is a thing.

Next, the ruled line candidate extraction process (S
4) is performed. This processing is performed by the ruled line extraction unit 3 shown in FIG.
4 is the processing performed. FIG. 10 is a flowchart specifically explaining this processing. First, an integrated line segment which is not included in the black and white inversion candidate area is extracted from the integrated line segments integrated by the above-described processing (S4-1). Next, it is determined whether or not the line width of the integrated line segment extracted by the extraction processing is narrower than a predetermined threshold value (thd1) (S4-2). Here, when the line width of the integrated line segment is narrower than a predetermined threshold value (thd1), it is determined as a ruled line candidate (S4-2).

Explaining with reference to FIG. 8 and FIG. 9 described above, when the threshold value (thd1) has the width shown in both figures, the integrated line segments ', ...' Are ruled line candidates.
Further, in the case of FIG. 9, integrated line segments, “,”, and “
Is a ruled line candidate. The difference between FIG. 8 and FIG. 9 is that the ruled line candidates are present in FIG. 9 with “blurred”. Note that the ruled line extraction processing (S5) of FIG. The extracted ruled line candidates are integrated in the mask scanning direction to create a long line segment in the horizontal direction.

Next, the isolated ruled line extraction process (S6) of FIG. 4 is performed. FIG. 11 is a flowchart specifically explaining this process. This processing is to determine whether or not there is a black-and-white reversal candidate area within the threshold value (thd2) of the distance to the left or right or up and down with respect to the ruled line extracted by the ruled line candidate extraction process described above (S6-1). . In the case of the example of FIG. 8 described above, the integrated line segments ',,' do not correspond to the above-mentioned isolated ruled line. That is, the black and white inversion candidate area or exists on the right side of the integrated line segment, but the black and white inversion candidate area does not exist on the left side of the line segment. In addition, the black and white inversion candidate region exists on the left side of the line segment in the integrated line segment “,”, but the black and white inversion candidate region does not exist on the right side of the line segment.

On the other hand, in the case of the example of FIG.
“,,” are the same as in the example of FIG. 8 described above, but are different in the case of the integrated line segment. That is, there are black and white inversion candidate regions “,” on the left and right of the integrated line segment ”(S6-1). Therefore, the integrated line segment "is extracted as an isolated ruled line.

Next, the integrated ruled line of FIG. 4 and the black-and-white reversal candidate area are integrated (S7). This process is a process performed by the black and white reversal area determination unit 35 shown in FIG. FIG. 12 is a flowchart specifically explaining this processing. This processing is to calculate a rectangular area circumscribing the black-and-white inversion candidate area on the upper and lower sides or the left and right sides surrounding the isolated ruled line extracted by the above-described processing (S6) (S7-1). That is, as described above, the isolated ruled line is, for example, an integrated line segment of "in FIG. 9, and black and white reversal regions'and" are located on the left and right thereof. Therefore, the processing for integrating the isolated ruled line as shown by "in the figure" into the black and white inversion areas', "on both sides" is performed. Specifically, a rectangular area surrounded by a square in FIG. 9 is calculated.

According to the above description, the extraction of the black and white reversal area in the horizontal direction and the extraction of the ruled line area are completed.
In addition, the above processing is performed in the black and white reversal area with respect to the horizontal
The extraction processing of the ruled line area is also performed in the vertical direction.

That is, in the process on the right side of the flowchart shown in FIG. 4, the line segment extraction process in the vertical direction (step (hereinafter referred to as S ′) 1) is performed by using the line segment extraction mask long in the vertical direction, and the extraction is performed. Line segment integration processing is performed on the selected line segment (S2 ′). This specific process is performed according to the flowcharts of FIGS. 5 and 7 as described above. next,
White / black inversion candidate area extraction processing (S3 ′) is performed, black / white inversion candidate areas are extracted, ruled line candidate extraction processing (S4 ′) is performed, and ruled line extraction is performed (S5 ′). Further, the extraction processing for the isolated ruled line is performed (S6 '), and finally the black-and-white inversion candidate area integration processing for the vertical direction (S7') is performed. It should be noted that in each of the above-described processes, as in the case of the horizontal direction, FIGS.
It is performed according to the concrete flowchart of.

With the above processing, the extraction processing of the black-and-white inversion area and the ruled line area in the vertical direction is also completed, and then the processing moves to the integration processing of the horizontal / vertical black-and-white inversion candidate areas (FIG. 4).
S8).

FIG. 13 is a flowchart specifically explaining this process. Further, FIG. 14 is a diagram for explaining the processing along the flowchart of FIG. First, Fig. 1
According to the flowchart of No. 3, the horizontal black-and-white inversion candidate area is extracted (S8-1). This black-and-white inversion candidate area is', ", and in Fig. 14. Next, the vertical-direction black-and-white inversion area overlapping the horizontal black-and-white inversion area is extracted (S8-2). This is shown in Fig. 14. The black-and-white inversion candidate area in the vertical direction is the area extracted in the black-and-white inversion area extraction processing (S3 ') in Fig. 4.

Next, the area of the overlapping portion of the horizontal black-and-white inversion candidate area and the vertical black-and-white inversion candidate area is calculated (S8-3). That is, the area A in which the horizontal black and white inversion candidate area 'and the vertical direction black and white inversion candidate area in FIG. 14 overlap, the area B in which the black and white inversion candidate area "and the vertical direction black and white inversion candidate area overlap, and the horizontal black and white. The respective areas of a region C where the reversal candidate region overlaps with the black-and-white reversal candidate region in the vertical direction and a region D where the black-and-white reversal candidate region overlaps with the black-and-white reversal candidate region in the vertical direction are calculated.
The areas of B, C and D are added.

Next, the following calculation is performed by using the area of the overlapping region calculated by the above processing. That is,
(Area of overlapped area) / Area of horizontal black and white reversal area is calculated, and this value and threshold value (thd
3) are compared (S8-4). Then, when the area of (overlapped area) / horizontal black-and-white reversal region is larger than the threshold value (thd3) (≧ thd3),
A rectangular area circumscribing the horizontal black-and-white inversion candidate area and the vertical black-and-white inversion candidate area is calculated (S8-6). That is, in this case, the case where the horizontal black-and-white inversion candidate area and the vertical black-and-white inversion area are large (the area of the overlapping portion) is large, and the black-and-white inversion area may be identified. On the other hand, in the above determination (S8-4),
(Area of overlapping portion) / When the area of the horizontal black / white inversion area is smaller than the threshold value (thd3) (<thd3), (horizontal black / white inversion candidate area and vertical black / white inversion area) This is the case where the area) is so small that it cannot be identified as a black and white reversal region.

However, in this case, the next judgment is made (S8-5).
To execute. That is, contrary to the above, (the area of the overlapping portion) / the area of the black and white reversal region in the vertical direction is calculated,
This value is compared with the threshold value (thd3). When the area of (overlapping portion) / vertical black / white reversal region is larger than the threshold value (thd3) (≧ thd3), the ratio of (area of overlapping portion) to the vertical direction is large. Assuming that there is, a rectangular area that circumscribes the black-and-white inversion candidate area and the vertical black-and-white inversion candidate area is calculated (S8-6). That is, in this case, the area of the overlapping portion of the black-and-white inversion candidate area in the horizontal direction and the black-and-white inversion area in the vertical direction is sufficiently larger than the area of the black-and-white inversion candidate area in the vertical direction. This is because it can be determined that it may be identified as the inverted region.

By calculating the rectangular area in which the horizontal black-and-white inversion candidate area and the vertical black-and-white inversion candidate area are circumscribed by the above processing, the solid line area E surrounded by a square mark in FIG. 14 is obtained. That is, this area is the black and white reversal area identified from the horizontal and vertical black and white reversal candidate areas.

However, in this example, the black-and-white inversion region is not immediately determined from the result obtained by the integration process of the horizontal / vertical black-and-white inversion candidate regions, but the process of calculating the connected component in the next black-and-white inversion region ( 4 is executed, and the white / black inversion area determination processing is performed based on this result (S10 in FIG. 4).
The details will be described below.

First, the calculation of the connected component is executed according to the flow chart shown in FIG. 15, and the number of connected components is calculated by, for example, labeling the black-and-white inversion area (S9-1). That is, the calculation of the number of connected components is, for example, a process of attaching a label to an independent black print portion in the black-and-white inversion area and calculating the number of labels (the number of connected components). The number of connected components is white.
Black in the target area when reversing the black inversion area
The number of print components or white print components.
Configure the range (frame) to be performed, for example, the upper and lower, left and right boundaries
It is the number that becomes the judgment standard when connecting. That is, concatenation
The number of components is the boundary between the top and bottom and the left and right when performing the black and white reversal
It is the number that serves as a criterion for connection when connecting lines. Therefore, for example, if there are halftone dots in the black and white reversal area,
The number of black dots (the number of labels, the number of connected components) becomes an extremely large value.

Next, the process for determining the black and white reversal area is performed (S
10). This determination processing is performed by the processing shown in FIG. 16, and the above-described connected component has a predetermined threshold value (thd
4) If it is greater than or equal to, the black and white inversion candidate area is rejected, and if the connected component is less than or equal to a predetermined threshold value, the black and white inversion candidate area is determined as the black and white inversion area (S10-1).

That is, this is a process for identifying whether or not the number of connected components calculated as described above is less than or equal to a preset threshold value. For example, when a pattern such as a halftone dot is added to the black-and-white inversion region, this process The cells that have been identified and have such a pattern added are the original black and white reversal areas (black and white reversal cells).
Is reliably identified and excluded from the black and white reversal area.

Here, in the case of the example shown in FIG. 17, for example, the black print portion is composed of one black portion of the background and independent black print portions (for example, six) included in the character string "company name", for a total of seven. is there. Therefore, in this case, the threshold value (t
If hd4) is set to “10”, the determination (S10-1) is (≦
thd4), which in this case is a black and white reversal area.

Next, the contour segment expression of the black-and-white reversal area determined by the above processing is performed (S11). This process is specifically performed according to the flowchart of FIG. That is, paying attention to the extracted line segment located outside the black-and-white inversion region, the line width of the line segment is incremented in the outward direction of the black-and-white inversion region (outward direction of the ray vector of the line segment) (S11).
-1). Then, the vertex coordinates of the rectangle of the line segment are stored (S
11-2).

Specifically, as shown in FIG. 19, the area (line segment) extracted by the line segment extraction masks in the horizontal and vertical directions may be inclined up to about 4 degrees due to the inclination of the original image. The area (line segment) to be extracted is also shifted vertically and horizontally. Therefore, as described above, the correction is performed to increase the line width of the line segment in the outward direction of the black / white reversal region.

Next, the attribute assigning process for the black and white reversal area is performed (S12). Specifically, as shown in the flowchart of FIG. 20, this process stores the position information in the black-and-white reversal region of the contour segment created by the above process as attribute data (S12-1). For example, in the example of the line segment segment sequence in FIG. 19 described above, the line segment segment sequences a1 to a4 located above the black / white inversion region are added with an attribute indicating the “upper” position of the black / white inversion region, The line segment segment columns b1 and b2 are added with an attribute indicating the “right” position of the black / white inversion region, and the line segment segment columns c1 to c4 located below the black / white inversion region are “below” the black / white inversion region. Is added to the line segment segment columns d1 and d2, and the attribute indicating the "left" position of the black and white reversal area is added.

With this configuration, the above-mentioned attributes are added to the line segment segment sequence surrounding the black-and-white inversion area. For example, these attributes are stored in the work memory 4 together with the position information of the line segment segment sequence. As a result, when the table recognition processing is performed later, it is possible to easily determine which line segment segment (column) is a segment (column) that surrounds the black-and-white inversion region and which position of the black-and-white inversion region corresponds from the attribute data. .

Next, cell division in the black / white inversion area is performed (S13 in FIG. 4). It should be noted that this process is a process performed by the in-black / white inversion area cell determination unit 36 shown in FIG. That is, the black and white reversal area may be composed of a plurality of cells, and the black and white reversal area set as described above is divided into cells.

FIG. 21 is a flowchart specifically explaining this processing. First, a ruled line in contact with the black-and-white reversal area is used to measure the distance (threshold value (thd
5)) The ruled lines separated from each other are extracted (S13-1). For example, when the black-and-white inversion area is formed in the horizontal direction, the ruled line in the vertical direction is extracted. Next, a strip area in which the ruled line is extended into the black and white inversion area is extracted (S13-2). That is, the vertical ruled lines extracted as described above are extended into the black-and-white inversion area, and the black-and-white inversion area is divided into strips, for example. Then, it is determined whether the white pixel density in the strip area is less than a predetermined threshold (thd5) (S13-).
3). That is, in this process, when the black-and-white inversion area is divided into strips, if there is a black-and-white inversion character at the position of the black-and-white inversion area where the ruled line is extended, this position cannot be separated by the ruled line. This is a process performed to confirm. Therefore, for example, a line segment extraction mask that is long in the vertical direction is used, and the black pixel number extraction processing is performed with the line segment extraction mask in the target black-and-white inversion region. When the number of black pixels is small, that is, the number of white pixels is large. In this case, it can be seen that there is a black-and-white inverted character at that position, and the position cannot be divided into cells.

If the white pixel density of the strip area is smaller than the predetermined threshold value (thd5), the contour segmentation of the black-and-white inverted area of the strip area is performed (S1).
3-4). Then, the contour segment expression and attribute assignment of each of the black and white reversal areas obtained by dividing the cells are performed (S13-).
5).

Finally, pixel inversion processing of the black and white inversion area is performed (S14). Note that this process is a process performed by the pixel inversion unit 37 in FIG. FIG. 22 is a flow chart for explaining this specific processing. First, a filling process is performed in the contour segment obtained by dividing the black and white reversal area (S14-
1). Next, black-and-white inversion processing is performed on an area other than the contour segment of the black-and-white inversion area (S14-2). That is, the outline portion of the black-and-white reversal area is filled with, for example, black and framed by the filling means, and the reversal processing is performed only on the black-and-white reversal area. For example, in the example of FIG. 19, line segment segment columns a1 to a4,
b1, b2, c1 to c4, d1, and d2 are painted in black, and the black and white reversal region is subjected to black and white reversal processing.

When the black-and-white inversion processing is performed on the black-and-white inversion area as described above, the white characters in the black-and-white inversion area are black-and-white inversion and black printing is performed. Note that FIG. 23 is a diagram showing a state after the inversion processing is performed by the above-mentioned black-and-white inversion processing.

As described above, the black / white reversal area identification device of this embodiment uses the line segment extraction masks that are long in the horizontal direction and the line segment extraction masks that are long in the vertical direction. This is a device for identifying the inversion region, and is a device that can accurately identify the black and white inversion region.

In the black-and-white inversion area identifying apparatus of this example, the number of connected components is calculated to determine the black-and-white inversion area, and when the number of connected components is equal to or less than the predetermined threshold value (thd4), it is determined to be the black-and-white inversion area. The present invention does not necessarily require the calculation of the number of connected components and the determination processing based on the number of connected components.

Further, in the explanation of this example, the process (S9) of FIG. 4 is executed to set the black and white reversal area, and the area of the (overlapped portion) / vertical black and white reversal area is calculated. And the threshold value (thd3) are compared with each other to set the black-and-white inversion area, but even if the black-and-white inversion candidate area in the horizontal direction and the vertical direction is surrounded by a contour line and the area is set as the black-and-white inversion area. Good.

Further, a line segment segment row in the horizontal direction is extracted, and the ruled line information in the vertical direction is used for this line segment segment row to determine the area partitioned by the line segment segment row in the horizontal direction and the ruled line in the vertical direction. It may be set in the black and white inversion area. Also, it is not always necessary to use the line segment sequence,
The horizontal black-and-white inversion candidate area may be used as it is, and the horizontal black-and-white inversion candidate area may be partitioned by vertical ruled lines to be set as the black-and-white inversion area.

FIG. 24 is an example of setting the black and white reversal area by the above-mentioned processing. That is, FIG. 7A shows a horizontal black-and-white reversal candidate area and a vertical ruled line ',
The area separated by 'is set to the black and white inversion area. In another example shown in FIG. 9B, when the black and white inversion candidate areas in the horizontal direction have the same length, both black and white inversion areas are recognized as black and white inversion areas. With such a configuration, it is possible to accurately extract the black and white reversal area.

On the other hand, contrary to the above, a vertical line segment segment string is extracted, and the horizontal ruled line information is used for this line segment segment string, and the vertical line segment segment string and the horizontal ruled line are used. The area partitioned by may be set as the black and white inversion area. Further, it is not always necessary to use the line segment segment row, and the black-and-white inversion candidate area in the vertical direction may be used as it is, and the black-and-white inversion area may be set by partitioning with horizontal ruled lines. <Second Embodiment> Next, a second embodiment of the present invention will be described.

In this example, the black-and-white inversion area identification device extracts a black-and-white inversion area, performs black-and-white inversion processing on the area, and performs recognition processing on the black-and-white inversion area and a character string in a normal cell. Is a table recognition device for performing the recognition process for the entire table.

The processing of this example is also executed by the CPU 1 according to the table recognition processing program stored in the program memory 2. First, FIG. 25 is a diagram showing an example of a form document to be recognized by the table recognition device of this example. The form of this example is, for example, a list of employees of a certain company, in which items such as company name, employee gender, telephone number, name, and person in charge are described. Has been.

This form original is read by the scanner 5, and the corresponding image data is stored in the image memory 3 via the bus 11. Then, first, line segment extraction processing (S1) is performed using a line segment extraction mask that is long in the horizontal direction, and line segment integration processing is performed on the extracted line segments (S2). In the vertical direction as well, after the simultaneous or horizontal processing is completed, a line segment extraction process (S1 ′) is performed using a line segment extraction mask that is long in the vertical direction, and the extracted line segment is processed. Line segment integration processing is performed (S2 '). Further, this line segment extraction and line segment integration processing is performed according to the flow charts of FIGS. 5 and 7 described above, and ruled lines including black and white inversion area cells or normal cells are extracted.

FIG. 26 shows a table recognition result in a state after the above-described line segment extraction and line segment integration processing, for example. As shown in the figure, a part of the black-and-white reversal region has a region where the line segment integration processing cannot be performed sufficiently. That is, there is a part where the line segment cannot be surely integrated in a part of the "store name", a part of the "company name", etc. in the item description. Incidentally, this portion is indicated by T, U, V and W in the same figure. This portion is a portion where "blurring" or the like exists in the original document state shown in FIG.

Next, an isolated ruled line extraction process is performed as a process for the above-mentioned "blurred" portion (S6, S).
6 ') and the above-mentioned "blurred" portion are extracted as isolated ruled lines. FIG. 27 shows an output result when the state of the isolated ruled line extracted by the above-described processing is table-recognized and output. As shown in the figure, isolated ruled lines are formed at the locations corresponding to T, U, V, and W described above. In addition, these ruled lines are shown by T ', U', V ', and W'in the same figure.

The table recognition apparatus of this example corrects the isolated ruled line as described above into the black and white inversion candidate region by integrating the isolated ruled line and the black and white inversion candidate region (S).
7). That is, the isolated ruled lines T ′, U ′, V ′ of FIG.
In either case, there are black-and-white inversion candidate areas on the left and right sides, and the isolated ruled lines T ', U', V ', and W'are integrated into the black-and-white inversion candidate areas on both sides by the above-described integration processing (S7). It FIG. 28 shows a result of outputting the state in which the isolated ruled line is integrated with the black-and-white inversion candidate area by the above-described processing.

Next, a horizontal / vertical black / white inversion candidate area integration process (S8), a connected component calculation process in the black / white inversion region (S9), and a black / white inversion region determination process (S10) based on the result are performed. After that, a contour segment is created for the determined black and white reversal area (S11). FIG. 29 is a diagram showing a state in which a contour segment is formed in the determined black-and-white inversion area.

Further, FIG. 30 shows a result of performing table division processing (S13) in order to extract a cell area included in the black-and-white inversion area and performing table recognition output. As shown in the figure, in the cell division processing of this example, for example, a black-and-white inversion region formed in the horizontal direction is applied to a vertical ruled line 4
Cell division is performed by extending 0 to 45 into the black / white inversion area, and the cell area in the black / white inversion area can be easily divided. Although there is no part corresponding to FIG. 30,
When dividing the cell area in the black and white inversion area, if there is a character string on the extension of the vertical ruled line, the number of black pixels is detected using the vertical line segment extraction mask, so it is not a simple division Can be recognized.

After the black-and-white inversion area is extracted as described above, the black-and-white inversion area is subjected to the inversion processing by executing the black-and-white inversion processing (S14). FIG. 31 shows an output result obtained by inverting the item area of “company name” in the black and white inversion area.

On the other hand, a line width is likely to be abnormal in the character string reversed by the black-and-white inversion process. That is,
Due to black and white reversal processing, white characters are not printed correctly in black,
For example, a part having a very narrow line width or a part having a wide line width is generated.
Therefore, the table recognition device of this example performs shaping processing of the character string in order to deal with this problem.

FIG. 32 is a flow chart for explaining the character string shaping processing after the pixel inversion processing. The details will be described below. First, the average line width in the horizontal direction of the pattern in the area is calculated (step (hereinafter referred to as ST) 1). In this calculation, the counting of the number of dots is started from the position where the white dot changes to the black dot in the horizontal direction, and the counting is ended at the position where the black dot changes to the white dot next. Then, the number of dots counted during that time is set as the line width in the horizontal direction. Next, the same is done in the vertical direction, and the average line width in the vertical direction of the pattern in the region is calculated (ST2). Also in this calculation, the counting of the number of dots is started from the position where the white dot is changed to the black dot in the vertical direction, and the counting is ended at the position where the black dot is changed to the white dot next. Then, the number of dots counted during that time is taken as the line width in the vertical direction. Then, the average line width of the pattern is calculated for the result of performing the above-mentioned processing a plurality of times ((horizontal + vertical)).
/ 2) (ST3).

After that, the average line width of the character (column) after the black-and-white reversal processing obtained as described above is compared with a predetermined threshold value (thw) (ST4). When it is determined that the average line width is less than or equal to the predetermined threshold value (thw) (ST4 is <thw), expansion processing is performed (ST
5). On the other hand, when it is determined that the average line width of a character (column) is equal to or larger than a predetermined threshold value (thw) (ST4 is>
thw) and contraction processing (ST6). That is, the average line width of the reversed character (column) is the threshold value (th
If it is narrower than w), the expansion process is performed, if the line width is thick, the contraction process is performed, and the character (column) after the inversion process is shaped.

As described above, the cell area extraction 23 is performed after the shaping processing is performed on the character (string) after the inversion processing.
At this time, the table recognition apparatus of this example has already added the attributes to the line segment segments corresponding to the ruled line extraction and the black-and-white inversion area during the black-and-white inversion area extraction processing. It is possible to distinguish between a ruled line surrounding a cell and a ruled line surrounding a normal cell.

For example, FIG. 33 is a diagram for specifically explaining the above contents. As described above, since the vertical and horizontal attributes are added to the line segment segment row corresponding to the black / white inversion area, it is possible to easily determine which position in the black / white inversion area each line segment (row) corresponds to. For example, as shown in the figure, the area surrounded by the upper and lower attributes of the line segment segment row is a black-and-white inverted area, and the area surrounded by the line segment segment rows having the opposite upper and lower attributes is a normal cell area. .

After the above processing, the character string in each cell is recognized, and the table form including the recognition result for the character string is output together with the form form. Note that FIG. 30 described above is also the final output result of the table recognition device of this example.

[0150]

As described above, according to the present invention, the following effects can be obtained. That is, according to the black / white inverted area identifying device of the present invention, even if a black / white inverted area exists in a part of one cell, the area can be accurately identified. Further, even when a large-sized inverted character exists in the black-and-white inversion area, the line segment extraction mask used in the present invention uses an extremely thin and relatively short mask, so that the black-and-white inversion area can be accurately identified. Further, even when the front image is read with a tilt, since the black-and-white inversion area is identified by the integration process while recognizing the number of black pixels in a narrow range,
It can be recognized as a black-and-white reversal region with a slope.

Further, according to one aspect of the present invention, even if there is, for example, "blurring" in the black and white reversal area, the black and white reversal area can be accurately identified because it is repaired by the isolated ruled line processing.

Further, according to one aspect of the present invention, the position of the created line segment segment sequence with respect to the black-and-white inversion area is clarified, which can be easily performed when distinguishing the black-and-white inversion area from the normal cell area, for example. You can

Further, according to one aspect of the present invention, the black-and-white inversion area is divided into, for example, strips, and the area partitioned by the ruled lines in the horizontal direction or the vertical direction is used as the cell area. Can be divided into cells.

According to the table recognition apparatus of the present invention, even if a black-and-white inversion area exists in a part of one cell, the inversion area can be accurately identified, and a large-sized inversion character can be identified in the black-and-white inversion area. Even if the image exists, the area can be accurately identified, and the black-and-white inversion area can be accurately identified even when the front image is read with an inclination.

Further, according to the present invention, the processing program is stored in a storage medium such as a hard disk, an IC memory card, a magnetic tape, a floppy disk, or an optical disk, and is read out as necessary to identify the black / white inverted area and to perform the table recognition processing. It can be performed.

[Brief description of drawings]

FIG. 1 is a system configuration diagram of a table recognition device including a black-and-white inverted region recognition device according to the present embodiment.

FIG. 2 is a diagram illustrating a recognition process performed by a CPU.

FIG. 3 is a system configuration diagram specifically describing a process performed by a CPU.

FIG. 4 is a flowchart illustrating processing of the first embodiment example.

FIG. 5 is a flowchart specifically explaining a line segment extraction process (S1).

FIG. 6 illustrates a case where a horizontal direction component is extracted by a line segment extraction mask designating a pixel (dot) area in which m × n (vertical × horizontal) is 1 × 6.

FIG. 7 is a flowchart specifically illustrating a line segment integration process.

FIG. 8 is a diagram showing a result of line segment integration processing.

FIG. 9 is a diagram showing a result of line segment integration processing.

FIG. 10 is a flowchart of ruled line extraction processing.

FIG. 11 is a flowchart illustrating isolated ruled line extraction processing.

FIG. 12 is a flowchart specifically illustrating a process of integrating an isolated ruled line and a black and white inversion candidate area.

FIG. 13 is a flowchart specifically illustrating a process of integrating horizontal / vertical white / black inversion candidate regions.

FIG. 14 is a diagram illustrating a process of integrating horizontal / vertical black / white inversion candidate regions.

FIG. 15 is a flowchart specifically illustrating a process of calculating a connected component.

FIG. 16 is a flowchart for determining a black / white reversal area.

FIG. 17 shows a total of 7 black print parts including one black part in the background and independent black print parts (for example, 6) included in the character string “company name”.
It is a figure explaining the number of individual connected components.

FIG. 18 is a flowchart specifically illustrating a process of expressing a contour segment of a black and white reversal region.

FIG. 19 is a diagram illustrating an example in which an area (line segment) to be extracted is displaced and extracted.

FIG. 20 is a flowchart specifically showing the attribute assigning process for the black and white reversal region.

FIG. 21 is a flowchart of cell division in the black-and-white inversion area.

FIG. 22 is a flowchart illustrating pixel inversion processing in a black / white inversion area.

FIG. 23 is a diagram showing a state after performing inversion processing by black-and-white inversion processing.

FIG. 24 is an example of setting a black / white inversion area.

FIG. 25 is a diagram illustrating an example of a form document that is a target of table recognition for explaining the second embodiment.

FIG. 26 is a flowchart of ruled line extraction processing which is an output result from the table recognition apparatus of this example, assuming that the table recognition output is performed after the line segment extraction and line segment integration processing is performed.

FIG. 27 is an output result when the state of the extracted isolated ruled line is output by table recognition.

FIG. 28 is a result of outputting a state in which an isolated ruled line is integrated into a white / black inversion candidate area.

FIG. 29 is a diagram showing a state in which a contour segment is formed in the determined black-and-white inversion area.

FIG. 30 shows a result of table recognition output after performing cell division processing in the black and white inversion area to extract a cell area included in the black and white inversion area.

FIG. 31 is a diagram showing an output result obtained by inverting the item area of “company name” in the black and white inversion area.

FIG. 32 is a flowchart illustrating a shaping process of a character string after the pixel inversion process.

FIG. 33 is a diagram illustrating that a ruled line surrounding a black / white reversal region cell or a ruled line surrounding a normal cell can be easily distinguished.

FIG. 34 is a diagram illustrating a conventional table recognition system.

FIG. 35 (a) is a diagram illustrating a run configuration;
(B) is a figure explaining a black-and-white inversion area, (c) is a figure which shows that all the areas in a black-and-white inversion area are not a black-and-white inversion area.

[Explanation of symbols]

1 Central processing unit (CPU) 2 Program memory 3 image memory 4 work memory 5 scanner 6 memory 7 dictionary file 8 display 9 Printer 10 Interface circuit 11 bus 12 drivers 13 hard disk 14 IC memory card 15 magnetic tape 16 floppy disks 17 Optical disc 19 Communication interface 20 image input 21 Line segment extraction 22 Black and white reversal area extraction 23 Cell area extraction 24 Cell item extraction 25 character recognition 30 Image input section 31 Image area storage 32 line segment extractor 33 Line segment integration section 34 Ruled Line Extraction Unit 35 Black / white reversal area determination unit 36 Cell determination unit in black and white reversal area 37 Pixel inversion section 40-45 ruled lines

─────────────────────────────────────────────────── ─── Continuation of the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06K 9/00-9/82 G06F 17/21 G06F 19/00

Claims (30)

(57) [Claims] 1. The tabular binary image data is horizontally scanned using a line segment extraction mask long in the horizontal direction, and vertically scanned using a line segment extraction mask long in the vertical direction to obtain black. Area extracting means for extracting areas each having a pixel density higher than a predetermined threshold value, and first integrating means for integrating the areas having a high black pixel density extracted by the area extracting means in a direction orthogonal to each mask scanning direction. Second integration means for integrating the black and white inversion candidate areas into respective mask scanning directions if the line width of the areas integrated by the first integration means is equal to or larger than a predetermined threshold value. And a line segment segment string creating means for creating horizontal and vertical line segment segment strings for the outline of the black and white reversal candidate area integrated by the second integrating means, and the line segment segment string creating means. Tone reversal region identification apparatus characterized by comprising: a tone reversal region setting means for setting a region surrounded by horizontal line segments rows and vertical line segment row in tone reversal region, the. 2. A tabular binary image data is scanned horizontally using a line segment extraction mask long in the horizontal direction, and vertically scanned using a line segment extraction mask long in the vertical direction to obtain black. Area extracting means for extracting areas each having a pixel density higher than a predetermined threshold value, and first integrating means for integrating the areas having a high black pixel density extracted by the area extracting means in a direction orthogonal to each mask scanning direction. Second integration means for integrating the black and white inversion candidate areas into respective mask scanning directions if the line width of the areas integrated by the first integration means is equal to or larger than a predetermined threshold value. And a black and white reversal area setting means for setting the black and white reversal area according to the overlapping state of the horizontal direction black and white reversal candidate area and the vertical direction black and white reversal candidate area integrated by the second integrating means. That black-and-white inversion region identification device. 3. A ruled line determining means for determining that the area integrated by the first integrating means is a ruled line when the line width of the area integrated by the first integrating means is less than or equal to a predetermined threshold value. The black-and-white reversal area identification device according to claim 1 or 2, further comprising: 4. The black-and-white reversal area identifying device according to claim 3, further comprising ruled line creating means for creating ruled lines by integrating the ruled lines judged by the ruled line judging means in respective mask scanning directions. 5. When the ruled line creating unit creates a vertical ruled line, a region surrounded by the horizontal line segment segments created by the line segment segment string creating unit and the vertical ruled line is created. The black-and-white inversion area identifying device according to claim 4, wherein the black-and-white inversion area is defined. 6. When the ruled line creating unit creates a horizontal ruled line, a vertical line segment segment string created by the line segment segment string creating unit and an area surrounded by the horizontal ruled lines are created. The black-and-white inversion area identifying device according to claim 4, wherein the black-and-white inversion area is defined. 7. The black-and-white inverted area identifying device according to claim 1, wherein a margin for absorbing irregularities of the contour of the black-and-white inverted area is added to the line segment segment sequence. . 8. The black-and-white inversion area identifying device according to claim 1, wherein the black-and-white inversion area setting means sets the black-and-white inversion area when the number of connected components is equal to or less than a predetermined threshold value. 9. A calculation means for calculating the number of connected components,
9. The black-and-white reversal area identifying device according to claim 8, further comprising: identifying means for identifying that the number of connected components calculated by the calculating means is less than or equal to a predetermined threshold value. 10. An isolated ruled line integrating means for extracting an isolated ruled line candidate, and integrating the isolated ruled line candidate also with the black and white inversion candidate region when the black and white inversion candidate region exists around the isolated ruled line candidate. The black-and-white reversal area identification device according to claim 1 or 2, further comprising: 11. The extraction of the isolated ruled line candidate is provided with a means for distinguishing a character region and a ruled line region, and a black and white inverted region is extracted after the ruled line extraction processing.
The black and white reversal area identification device described. 12. The white / black inverted area identifying device according to claim 1, wherein an attribute indicating that the line segment segment string is a black / white inverted area is added to the line segment segment string. 13. The monochrome according to claim 12, wherein the position of the line segment segment with respect to the black-and-white reversal region is added to the line segment segment sequence as position information of upper and lower sides or left and right sides. Inversion area identification device. 14. When analyzing the cell structure of the table, if the position information of the cell is the same as the position information of the upper and lower sides or the left and right sides added to the line segment sequence, it is determined to be a black and white inverted cell region. 14. The black-and-white reversal area identification device according to claim 13, wherein if it is different, it is determined as a normal cell area. 15. The cell area in the black / white inversion area is extracted by extending a ruled line connected to the black / white inversion area when the black / white inversion area includes a plurality of cell areas. Alternatively, the black-and-white reversal area identification device described in item 2. 16. The method according to claim 15, further comprising means for obtaining a black pixel density by subdividing an area in which a ruled line connected to the black-and-white inversion area is extended when the cells in the black-and-white inversion area are divided. Black and white inversion area identification device. 17. A binary image data in a tabular format,
Region extraction that scans in the horizontal direction using a line segment extraction mask that is long in the horizontal direction and in the vertical direction using a line segment extraction mask that is long in the vertical direction, and that extracts regions in which the black pixel density is higher than a predetermined threshold. Means, first integrating means for integrating the areas with high black pixel density extracted by the area extracting means in a direction orthogonal to each mask scanning direction, and line width of the areas integrated by the first integrating means Is greater than or equal to a predetermined threshold value, it is determined to be a black-and-white inversion candidate area, and the black-and-white inversion candidate area is integrated in each mask scanning direction, and the black-and-white inversion candidate area integrated by the second integration means. Line segment segment creating means for creating line segment segments in the horizontal and vertical directions with respect to the contour of the line segment, and a line segment segment in the horizontal direction and vertical line segment segment created by the line segment segment creating means. The black-and-white inversion area identifying section having a black-and-white inversion area setting means for setting the area surrounded by the column and the black-and-white inversion area, and the black-and-white inversion area for the black-and-white inversion area identified by the black-and-white inversion area recognition section. A table recognition device comprising: a black-and-white inversion processing means for performing the recognition processing; and a recognition means for performing a recognition processing for a character string located in the area subjected to the inversion processing by the black-and-white inversion processing means. 18. A table format binary image data,
Region extraction that scans in the horizontal direction using a line segment extraction mask that is long in the horizontal direction and in the vertical direction using a line segment extraction mask that is long in the vertical direction, and that extracts regions in which the black pixel density is higher than a predetermined threshold. Means, first integrating means for integrating the areas with high black pixel density extracted by the area extracting means in a direction orthogonal to each mask scanning direction, and line width of the areas integrated by the first integrating means Is greater than or equal to a predetermined threshold value, it is determined to be a black-and-white inversion candidate area, and the black-and-white inversion candidate area is integrated in each mask scanning direction, and the horizontal black and white integrated by the second integration means. A black-and-white inversion area identifying section having a black-and-white inversion area setting means for setting the black-and-white inversion area according to the overlapping state of the inversion candidate area and the black-and-white inversion candidate area in the vertical direction, and the black-and-white inversion area identification section identified by the black-and-white inversion area identifying section. A white area inversion processing means for performing a black and white inversion processing on the transposed area, and a recognition means for performing a recognition processing for a character string located in the area inverted by the black and white inversion processing means. Table recognizer. 19. The black-and-white reversal processing means comprises: a filling means for filling a contour segment of the black-and-white reversal area, and a processing means for black-and-white reversal of a black-and-white reversal area other than the contour segment string. 17
The described table recognition device. 20. The table recognition device according to claim 19, wherein the black and white reversal processing means has a cell area filling means for filling an area divided by cells. 21. A line width of a character string located in a black and white reversal region subjected to the reversal process by the black and white reversal processing means is calculated, and the character string is expanded or contracted according to the calculated line width to obtain the character string. 19. The table recognition device according to claim 17, which is shaped. 22. The table recognition device according to claim 17, wherein the black / white inversion area setting means sets the black / white inversion area when the number of connected components is equal to or less than a predetermined threshold value. 23. The area extracting means extracts an isolated ruled line candidate, and when the black / white inversion candidate area exists around the isolated ruled line candidate, the isolated ruled line candidate is also integrated into the black / white inversion candidate area. 19. The table recognition device according to claim 17, further comprising ruled line integrating means. 24. The table recognizing apparatus according to claim 17, wherein an attribute indicating that the line segment segment string is a black-and-white inverted area is added to the line segment segment string. 25. When the black-and-white inversion area is composed of a plurality of cell areas, the ruled line connecting to the black-and-white inversion area is extended to extract the cell area in the black-and-white inversion area. Alternatively, the table recognition device according to item 18. 26. When dividing cells in the black-and-white inversion area, there is provided means for obtaining a black pixel density by subdividing an area in which a ruled line connected to the black-and-white inversion area is extended. Table recognition device. 27. For binary image data in tabular form,
With a function to scan in the horizontal direction using a long line segment extraction mask in the horizontal direction, and to scan in the vertical direction using a long line segment extraction mask in the vertical direction, and to extract each area in which the black pixel density is higher than a predetermined threshold value. A function of integrating the areas with high black pixel density extracted by the function in a direction orthogonal to each mask scanning direction, and a black-and-white reversal candidate area if the line width of the area integrated by the function is a predetermined threshold or more And a function of integrating the black-and-white inversion candidate area in each mask scanning direction, and a function of creating horizontal and vertical line segment segments for the outline of the black-and-white inversion candidate area integrated by the function. , A function of setting an area surrounded by the horizontal line segment segment row and the vertical line segment segment row created by the function as a black / white inversion area, and an instruction for causing a computer to execute The computer-readable storage medium storing the program. 28. For binary image data in tabular form,
With a function to scan in the horizontal direction using a long line segment extraction mask in the horizontal direction, and to scan in the vertical direction using a long line segment extraction mask in the vertical direction, and to extract each area in which the black pixel density is higher than a predetermined threshold value. A function of integrating the areas with high black pixel density extracted by the function in a direction orthogonal to each mask scanning direction, and a black-and-white reversal candidate area if the line width of the area integrated by the function is a predetermined threshold or more Then, the black and white inversion candidate area is set in each mask scanning direction, and the black and white inversion candidate area is set by the overlapping state of the horizontal direction black and white inversion candidate area and the vertical direction black and white inversion candidate area integrated by the function. And a computer-readable storage medium that stores a program including instructions for causing a computer to execute the function. 29. The black-and-white inverted area identifying device according to claim 1, wherein the black-and-white inverted area includes two arbitrary colors created by three color elements. 30. The table recognition device according to claim 17, wherein the black and white shown in the black-and-white inversion area includes any two colors created by three elements of color.