JPH07111738B2 - Area boundary extraction method in document - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] In a document reading device for automatically extracting a description area such as a character, a figure, a table and a photograph in a document, an area boundary in a document for surely detecting the boundary of the description area Regarding the extraction method, in order to accurately extract the boundaries of each area, the document is read and stored as binary data in the image memory, and the vertical and horizontal directions in the document are stored from the data stored in the image memory. A document reading device that creates a histogram showing the distribution state of black pixels and extracts each description area of each character, figure, table, photograph, etc., and the distribution amount of black pixels shown by the histogram is small. A first threshold having a plurality of comparison values from a minimum value to a maximum value for detecting a portion, and a plurality of maximum values to a minimum value for detecting the width of a portion having a small black pixel distribution amount. Has a comparison value Second
By using the parameter table in which the threshold value is set and the set first and second threshold values, the position on the document where the distribution amount of black pixels in the histogram is small is detected to determine the boundary between the description areas. Means for controlling the combination of the first threshold value and the second threshold value set in the parameter table,
The operation of extracting the portion where the distribution amount of black pixels indicated by the histogram is small is repeated, and the position of the extracted portion on the document is set as the boundary of the description area.

[Industrial application field]

The present invention relates to a document reading device for automatically extracting a description area such as a character, a figure, a table, and a photograph of a document, and in particular, area boundary extraction in a document capable of surely detecting the boundary of the description area. Regarding the scheme.

In the print document, figures, tables, photographs, etc. are mixed in addition to the character strings, but the print document reading device uses an area in which the character strings are described, an area in which the figures are described, and a table. The area in which is described, the area in which the photograph is described, and the like are separately read.

In this way, each area is read by dividing the characters into character codes in the case of character strings, converting figures and images into image data, collecting statistical values from tables, and reading documents respectively. This is because the content to be processed after reading is different.

By the way, in order to divide such an area, it is necessary to surely detect the boundary of each area.

[Conventional technology]

Various methods have heretofore been attempted in order to automatically extract a region of a document such as a character string, a chart, a photograph, or the like. A typical method is to read a document, binarize it, store it in an image memory, read the binarized data from this image memory, and display a histogram showing the distribution state of vertical and horizontal black pixels on the document. There is a method of creating and searching a portion where the distribution of black pixels is 0 or close to 0.

FIG. 6 is a block diagram for explaining the conventional technique, and FIG. 7 is a diagram for explaining the operation of FIG.

The scanner 1 scans a document (not shown) to read a character string, a figure, a table, a photograph and the like, binarizes the same, and stores image data as shown in FIG. The black pixel counting unit 3 scans the image memory 2 to obtain a total of black pixels for each pixel row in the horizontal direction, and a histogram for the character area, the photograph area, the character area, and each vertical pixel row. The total of the black pixels of is calculated, a histogram for the character area, the photograph area, the character area, and the histogram are created and stored in the histogram memory 4.

Then, a portion where the distribution of black pixels in the histogram is 0 or close to 0 is searched, and the position of the portion obtained by this search on the document is determined as the boundary between the areas.

[Problems to be Solved by the Invention]

In order to distinguish the character area, the photograph area, the character area, the character area, and the character area in the document from the histogram stored in the histogram memory 4, the distribution of black pixels in the histogram is set to 0 or 0. It is necessary to detect the near part.

When detecting the part without black pixel number corresponding to the blank part indicating the boundary with the character region and the boundary with the character region from the histogram, if the number of black pixels in the photo region is large, the blank part with few black pixels is detected. In addition, since the number of black pixels in the photographic area is added, it is difficult to detect a blank portion where the number of black pixels is originally 0 or almost 0.

That is, it becomes difficult to detect a portion having a small number of black pixels, and the boundary between the character area and the boundary between the character area cannot be accurately separated.

Also, from the histogram, the part with a small number of black pixels corresponding to the blank part indicating the boundary between the character area and the photo area can be detected relatively clearly, but if the ground of the document is dirty, it can be detected as described above. When it becomes difficult and the width of this boundary is narrow, it is difficult to distinguish it from the line spacing of the character string, and there is a problem that the boundary between the regions cannot be accurately detected.

In view of such a problem, the present invention extracts each boundary between areas regardless of the width / narrowness of the blank area between the areas and when the blank area has a background stain or the like. The purpose is to accurately extract the boundaries of regions.

[Means for Solving the Problems]

FIG. 1 is a block diagram of the principle of the present invention, and FIG. 2 is a diagram for explaining the operation of FIG.

FIG. 1 shows a judgment unit 5 and a parameter table 6 added to FIG. As described with reference to FIG. 6, the histogram memory 4 stores the histogram as shown in FIG.

The determination means 5 uses the parameter table 6 as a comparison value for detecting a blank area between the character area, the photograph area, the character area, and each area shown in FIG.
A first threshold having a value for extracting from a portion having a small number of black pixels to a portion having a relatively large number of black pixels, and a second threshold for detecting the width of the blank portion extracted by the first threshold are set. The reading is started, and the extraction with the portion having a small number of black pixels corresponding to the blank portion of the histogram stored in the histogram memory 4 is started.

First threshold value read from parameter table 6 first
T ₁ is the smallest value, and the second threshold value W ₁ is the largest value. Therefore, as shown in FIG. 2 (a), when the width of the blank portion between the respective areas is wide and there is no background stain on the document or a portion such as a photograph in which a large number of black pixels develop, the black portion is black. A wide range with a small pixel distribution amount is extracted. Then, the distribution of black pixels is small, but the portion shown as narrow is not extracted.

Next, the determination means 5 keeps the first threshold value T ₁ and reads the second threshold value W ₂ from the parameter table 6, and again extracts the portion having a small number of black pixels corresponding to the blank portion of the histogram. In this case, the threshold value W ₂ <W ₁ . Subsequently, the determination means 5 keeps the first threshold value as T ₁ , reads the second threshold value W ₃ from the parameter table 6, and extracts the portion having a small number of black pixels corresponding to the blank portion of the histogram. In this case, the threshold value W ₃ <W ₂ .

In this way, when the process is executed up to the second threshold value W _m prepared in the parameter table 6, for example, FIG.
As shown in, the narrow range of the blank portion between the areas is detected. In this case the second threshold value W _m, it is necessary that not a small value enough to detect the line spacing or character spacing of the string.

Judging means 5 is subsequently a first threshold value and T _2, the second threshold is changed from W ₁ to W _m, executes an operation of extracting the portion of a small number of black pixels. In this case, the threshold value T ₂ > T ₁ . Then, further the above-described operation is repeated a first threshold value as T _3. In this case, the threshold value T ₃ > T ₂ .

In this way, when the first threshold value prepared in the parameter table 6 is executed until T _m , for example, the second threshold value
As shown in FIG. 7C, a range where there is background stain and the width of the blank portion is narrow is detected.

The determination means 5 uses the first and second methods as another method different from the above.
The threshold of is changed at the same time. That is, when the first threshold value T ₁ and the second threshold value W ₁ are read out from the parameter table 6 and the extraction operation of the portion having a small number of black pixels in the histogram is performed, the first threshold value T ₂ and the second threshold value The threshold value W ₂ is read out, the extraction operation of the portion with a small number of black pixels is performed again, then the first threshold value T ₃ and the second threshold value W ₃ are read out, and the portion with a small number of black pixels is extracted. The operation to be performed is repeated until the first threshold becomes T _m and the second threshold becomes W _m .

[Action]

With the above-described configuration, the determination unit 5 uses the first and second threshold values set in advance in the parameter table 6 to determine whether the histogram stored in the histogram memory 4 is free of background stains or the photograph. It is possible to extract from a wide blank portion which does not overlap with a portion having a large number of black pixels as described above to a blank portion which has a background stain or overlaps a portion having a large number of black pixels and which has a narrow width.

〔Example〕

FIG. 3 is a block diagram of a circuit showing an embodiment of the present invention.
FIG. 4 is a flowchart explaining the operation of FIG. 3, and FIG. 5 is a diagram showing an example of a parameter table.

6 that are the same as those in FIG. 6 have the same functions. Scanner 1
Scans a document (not shown) to scan text, graphics, tables, photos, etc., binarizes it, then performs image thinning processing to reduce the number of pixels, and prints a document of A4 size, for example, vertically. The image memory 2 has a size of 300 dots in the direction and 210 dots in the horizontal direction.

The black pixel counting unit 3 scans the image memory 2 to display, for example, the seventh
The total of black pixels in the horizontal direction of the image data as shown in the figure is obtained, the histogram for the character area, the photograph area, the character area, and the total of the black pixels in the vertical direction are obtained, and the character area, the photograph area, the character area, Histogram memory 4
To store.

The processor 7 reads m from the parameter table 6 having the contents shown in FIG. 5 as a variable for the first threshold value T _m and the second threshold value W _m . That is, 2 dots are read as the first threshold T ₁ , and 3 dots are read as the second threshold W ₁ .

Then, extraction of a portion having a small number of black pixels corresponding to a blank portion between the histogram stored in the histogram memory 4 and the area is started.

That is, as shown in FIG. 4, it is searched for a portion of the number of black pixels equal to or smaller than the first threshold value (2 dots) in the histogram,
If so, measure its width in dots. Then, it is checked whether this width is equal to or larger than the second threshold value (3 dots). And
If it is equal to or larger than the second threshold value, it is determined to be a boundary between regions.

If there is no portion of the number of black pixels equal to or less than the first threshold value (2 dots) in the histogram, or if the width is equal to or less than the second threshold value (3 dots), the processor 7 determines the boundary between the regions. Instead, it is determined to be non-boundary.

Then, the processor 7 changes the first and second threshold values.
That is, from the parameter table 6, 6 dots are read as the first threshold value T ₂ , and 3 dots are read as the second threshold value W ₂ .

Then, it is checked whether the set value in the parameter table 6 is completed. In this case, since it is a new set value, the histogram is searched again for a portion having the number of black pixels equal to or smaller than the first threshold value (6 dots), and if there is, the routine returns to the routine for measuring the width, and this width is It is checked whether it is equal to or larger than the second threshold value (3 dots).

If this width is greater than or equal to the second threshold value, it is determined that it is a boundary between regions, and there is no portion of the number of black pixels equal to or less than the first threshold value in the histogram. If there is no portion having a width, after determining that it is a non-boundary, the processor 7 further changes the first and second threshold values. Therefore, 11 dots are read from the parameter table 6 as the first threshold value T ₃ , 3 dots are read as the threshold value W ₃ of 2.

Then, it is checked whether the setting value of the parameter table 6 is completed,
Since it is a new setting value, the histogram again searches for a portion with the number of black pixels equal to or less than the first threshold value (11 dots), and if there is, returns to the routine for measuring the width, and this width is set to the second value. It is checked whether it is equal to or more than the threshold value (3 dots).

Similar to the above, if there is a part corresponding to the blank part, it is determined to be a boundary, and if not, it is determined to be a non-boundary, and then the processor 7 further reads 18 dots from the parameter table 6 as the first threshold T ₄ , and Read out 3 dots as the threshold value W ₄ of 2,
It is checked whether the setting value of the parameter table 6 is finished, and the setting value is new. Therefore, the histogram is searched again for a portion having the number of black pixels equal to or smaller than the first threshold value (18 dots), and the width thereof is measured. Then, it is checked whether this width is equal to or larger than the second threshold value (3 dots).

Similar to the above, if there is a part corresponding to the blank part, it is determined to be a boundary, and if not, it is determined to be a non-boundary, and then the processor 7 further reads 18 dots from the parameter table 6 as the first threshold T ₅ , 2 dots are read out as the threshold value W ₅ of 2,
It is checked whether or not the set value in the parameter table 6 is finished, and since it is a new setting, the histogram is searched again for a portion having the number of black pixels equal to or smaller than the first threshold value (18 dots), and if there is, the width is measured. Returning to the routine, it is checked whether this width is the second threshold value (2 dots) or more.

Similar to the above, if there is a portion corresponding to the blank portion, it is determined to be a boundary, and if not, it is determined to be a non-boundary, and the processor 7 further changes the first and second threshold values. The threshold value T ₆ and the second threshold value W ₆ are read, but since there is no new set value in the parameter table 6, it is recognized that the set value has ended, and the distribution of black pixels corresponding to the blank part is detected from the histogram. The operation of extracting the portion with the small amount is ended.

〔The invention's effect〕

As described above, according to the present invention, it is possible to accurately divide an area even when there are various widths of blank portions indicating the boundaries between the writing areas in a document, and the blank portions between the writing areas are Even if there is a background stain or a blank portion that overlaps with a region where black pixels are distributed as in a photograph when a histogram is created, the region can be accurately divided.

[Brief description of drawings]

 1 is a block diagram of the principle of the present invention, FIG. 2 is a diagram for explaining the operation of FIG. 1, FIG. 3 is a block diagram of a circuit showing an embodiment of the present invention, and FIG. 4 is of FIG. FIG. 5 is a flow chart for explaining the operation, FIG. 5 is a diagram showing an example of the parameter table, FIG. 6 is a block diagram for explaining the conventional technique, and FIG. 7 is a diagram for explaining the operation of FIG. In the figure, 1 is a scanner, 2 is an image memory, 3 is a black pixel counting unit, 4 is a histogram memory, 5 is determination means, 6 is a parameter table, and 7 is a processor.

Claims (1)

[Claims] 1. A character, a figure, a table, a photograph, etc. are read from a document and stored as binarized data in an image memory (2). From the data stored in the image memory (2), the document is read. A document reading device that creates a histogram showing the distribution of black and white pixels in the vertical and horizontal directions in FIG. 2 and then divides and extracts each description area of each character, figure, table, photograph, etc. based on the histogram. A first threshold value having a plurality of comparison values from a minimum value to a maximum value for detecting a portion where the distribution amount of black pixels indicated by the histogram is small, and a width of a portion where the distribution amount of black pixels is small. A parameter table (6) in which a second threshold value having a plurality of comparison values from a maximum value to a minimum value for detection is preset, and first and second threshold values set in the parameter table (6) Using the hiss A unit (5) for detecting a position on the document of a portion where the distribution amount of black pixels indicated by the gram is small and determining the detected position as a boundary between the description regions is set in the parameter table (6). The first threshold and the second threshold thus combined are combined, and the operation of extracting a portion where the distribution amount of black pixels indicated by the histogram is small is repeated,
An area boundary extraction method in a document, wherein a position of the extracted portion on the document is used as a boundary of the description area.