KR101985612B1 - Method for manufacturing digital articles of paper-articles - Google Patents

Abstract

The present invention relates to a method for digitizing paper documents, comprising the steps of: obtaining a paper document to be digitized; scanning the obtained paper document to obtain a scanned image of the paper document; dividing the scanned image into a text area and an image area, forming, by using pixels, the text area into unit segments, which are the minimum unit segments connected to each other, character segments formed by the combination of one or more of the unit segments, character row segments formed by the combination of one or more of the character segments, and block segments formed by the combination of one or more of the character row segments, and sequentially forming group segments formed by the combination of one or more of the block segments; converting the contents of the paper document into a text file by extracting characters from the segmented scanned image, and converting the image area into an image file; and generating an extensible markup language (XML) by using the text file. Thus, it is easy to manage and use paper newspapers.

Description

[0001] METHOD FOR MANUFACTURING DIGITAL ARTICLES OF PAPER-ARTICLES [0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of digitizing paper documents issued on paper, and more particularly to a method of digitizing paper newspapers.

In recent years, as the Internet has developed, paper newspapers have been published as electronic newspapers, and conventional traditional paper newspapers are being sourced.

In addition, before it became electronic newspapers, it relied on paper newspapers, and it is becoming necessary to utilize the vast amount of past information in statistics and analysis.

In addition, even in the case of an ancient document with many Chinese characters, there is no software or tool capable of properly digitizing it, which makes it difficult to digitize an old document.

In addition, in the past, even in Korean, horizontal writing was used as well as vertical writing. Also, since there was a document written from left to right or from right to left, detailed examination and analysis are necessary to digitize such documents.

However, there is no research and development on this.

The present invention is directed to solving the above-mentioned problems and other problems. Another object is to provide a method of digitizing paper documents.

According to an aspect of the present invention, there is provided a method for digitizing a paper document, the method comprising: obtaining a paper document to be digitized; Scanning the acquired paper document to obtain a scanned image of the paper document; The scan image is divided into a text area and an image area, and the text area is divided into a unit segment, which is a minimum unit segment connected to each other by using pixels, a character segment formed by a combination of at least one unit segment, A character row segment formed by a combination of one or more of the character segments, a block segment formed by a combination of one or more character row segments, and group segments formed by combinations of one or more of the block segments are sequentially A segmentation step of forming a segment; Converting the contents of the paper document into a text file by extracting characters from the segmented scanned image, and converting the image area into an image file; And generating an extensible markup language (XML) using the text file. The method for digitizing a paper document can be provided.

According to an aspect of the present invention, the step of segmenting into the unit segments includes obtaining a pixel block including neighboring pixels located in four neighborhoods of upper, lower, left, and right with reference to a certain pixel within a predetermined rectangular area step; Determining a valid pixel mass when the pixel mass of the pixel mass has a predetermined number or more of pixels and obtaining a plurality of the effective pixel masses to obtain a set of pixel mass; And determining the set of pixel blocks as a unit segment if the length of the set of pixel blocks is equal to or greater than a predetermined length.

According to an aspect of the present invention, the step of segmenting into the character segment includes a step of selecting a first unit segment and a second unit segment adjacent to the first unit segment in the set of unit segments including the unit segments ; The area of the first unit segment is S ₁ , the area of the second unit segment is S ₂ , the overlapping area of the first unit segment and the second unit segment is S _I , When the area of the state in which the two unit segments are combined is S _u , if one or more of the following conditional expressions (1) to (4) are satisfied, the same character segment can be determined.

S _I ≥k ₁ S ₁ ------------------- (1)

_{S I ≥k 1 S 2 ------------------- (} 2)

S _1? K ₂ S _u ------------------- (3)

S _2? K ₂ S _u ------------------- (4)

In the above equation, k ₁ and k ₂ are preset constants determined according to the typeface.

According to an aspect of the present invention, segmenting into the character row segment includes: obtaining a reference width (w ₀ ) and a reference height (h ₀ ) of the character segment according to a size of the plurality of character segments; Aligning the plurality of character segments in a first direction; Constructing a character row segment by selecting character segments within a predetermined size range in consideration of the width and height of the character segments considering the reference width w ₀ and the reference height h ₀ ; And the widths or heights of the character segments are overlapped along a second direction in which the intervals between the adjacent character segments are not less than a predetermined interval along the first direction or the character segments are perpendicular to the character line segment and the first direction And if so, a new character line segment.

According to an aspect of the present invention, the step of segmenting into the block segments comprises: storing character segments in different formats according to the width and height of the character segments; Repeating combining character string segments having the same length within a predetermined range while the width and height of the character segments are within a predetermined range; The step of combining the character line segments and the character line segments spaced apart from the neighboring character line segments by more than a predetermined interval along the second direction form another block segment; And processing the character line segments having different widths and heights of the character segments into different blocks.

According to an aspect of the present invention, the step of segmenting into the group segments is performed on a lower portion of a first block segment including a first character segment having a largest width and height of character segments among adjacent block segments, Forming block segments that are at least partially overlapping the first block segment along a second direction into the same group; And a block segment which is spaced apart from the block segment formed in the same group by a predetermined interval or spaced along the first direction or the second direction, or a block segment intersecting with a second block segment or a dividing line including the first character segment, . ≪ / RTI >

According to an aspect of the present invention, the method may further include grouping the image segments located between the grouped block segments into the same group.

According to an aspect of the present invention, after the segmenting step, if the scan image is scanned at an angle, the step of correcting the scan image by rotating the scan image to a predetermined position may be further included.

According to an aspect of the present invention, the step of modifying the scanned image may include the steps of: if the width or the height of the unit segment is constant and if any one of the width and height is longer than a predetermined length, ; Determining whether the line segment is formed adjacent to a first axis of the scan image and a second axis perpendicular to the first axis; Calculating an angle between the line segment and an axis adjacent to the line segment on the scan image of the first axis and the second axis; And rotating the scan image by the calculated angle to match the line segment and the adjacent axis.

According to an aspect of the present invention, when the width and the height of the unit segment are larger than a predetermined size, the unit segment may be included in the image segment and excluded from the segment.

According to an aspect of the present invention, the paper document may be a paper newspaper.

According to an aspect of the present invention, the unit segment, the character segment, the character row segment, the block segment, and the group segment have coordinate information, and the group segment may form a polygon according to the coordinate information.

According to an aspect of the present invention, the group segment may be formed as a closed region formed by a path that does not overlap while moving along the outermost portion of the adjacent block segment.

According to an aspect of the present invention, the text file includes a title, an article classification, a body text, and a byline, and the step of generating an extensible markup language (XML) using the text file The method may further include generating a news mark (News Markup Language) by combining the title, the article classification, the body text, and the byline with the delimiter markup language and basic environment information.

According to an aspect of the present invention, there is provided a method of converting a scanned image into a PDF file, Generating coordinates of the group segment on the PDF image; And selecting and registering a text file corresponding to the coordinates of the generated group segment among the files in the NewsML.

According to an aspect of the present invention, the conversion into the text file may be performed by an optical character recognition (OCR) program.

The effect of the digitizing method of the paper document, particularly paper newspaper according to the present invention, will be described below.

According to at least one embodiment of the present invention, it is possible to digitize a paper document issued by a paper such as a paper newspaper, thereby realizing the contents of the paper newspaper on the web.

According to at least one of the embodiments of the present invention, there is an advantage that it is easy to manage and utilize a paper newspaper by storing a text file as a file name associated with the date of issue.

According to at least one of the embodiments of the present invention, the character and image areas of the paper newspaper are stored to include coordinate information, so that it is possible to more easily confirm the details when clicking on the web.

Further scope of applicability of the present invention will become apparent from the following detailed description. It should be understood, however, that the detailed description and specific examples, such as the preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art.

1 is a flowchart of a method of digitizing a paper newspaper related to the present invention.
2A to 2E are diagrams for explaining rotation of an image related to the present invention.
3A and 3B are scanned images of paper newspapers related to the present invention.
4 is a flow chart related to tilt correction of a paper newspaper according to an embodiment of the present invention.
5A and 5B are views for explaining segmentation according to an embodiment of the present invention.
6 is an illustration for explaining a method of character line segmentation according to an embodiment of the present invention.
Figure 7 is a scanned image of a block segmented paper newspaper associated with the present invention.
8 is a scanned image of a paper newspaper containing photographs and advertisements related to the present invention.
9 is a conceptual diagram for polygon processing of a group segment of a paper newspaper related to the present invention.
10 is a diagram for explaining group segmentation related to the present invention.
11 is a diagram for explaining a crop related to the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, wherein like or similar components are denoted by the same reference numerals, and redundant explanations thereof will be omitted. The suffix " part " for the constituent elements used in the following description is to be given or mixed with consideration only for ease of specification, and does not have a meaning or role that distinguishes itself. In the following description of the embodiments of the present invention, a detailed description of related arts will be omitted when it is determined that the gist of the embodiments disclosed herein may be blurred. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are intended to provide further explanation of the invention as claimed. , ≪ / RTI > equivalents, and alternatives.

Terms including ordinals, such as first, second, etc., may be used to describe various elements, but the elements are not limited to these terms. The terms are used only for the purpose of distinguishing one component from another.

The singular expressions include plural expressions unless the context clearly dictates otherwise.

In the present application, the terms "comprises", "having", and the like are used to specify that a feature, a number, a step, an operation, an element, a component, But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

FIG. 1 is a flowchart of a method of digitizing a paper document related to the present invention. Hereinafter, a method of digitizing a paper document according to an embodiment of the present invention will be described with reference to FIG.

One embodiment of the present invention relates to a method of digitizing a paper newspaper, particularly, among paper documents, and hereinafter, a method of digitizing a paper newspaper will be described. However, the following contents are not limited to digitizing paper newspapers, but may be applied to digitizing books such as periodicals, magazines, or monographs.

First, in order to digitize a paper document, a paper document to be digitized is acquired (S110), and the acquired paper document is scanned to image the paper document (S120). The imaging is a step of acquiring a scanned image by storing the image as an image file having an extension of jpg, png, bmp, gif, tiff, or the like. The type of the image file is not particularly limited, and it is natural that an image file having an extension other than the extension of the image file listed above is also available. At this time, in the case where a document that has already been electronicized, for example, a PDF file, exists, a document composed of PDF may be digitized. In this case, a separate scan operation is unnecessary, and the step of acquiring the scan image (S120) or modifying the scan image (S130) may be omitted, and the subsequent steps are the same.

The method of digitizing a paper document according to an embodiment of the present invention may further include a step of modifying the scan image (S130). 2A to 2E are views for explaining rotation of a scan image related to the present invention. FIG. 2A is a diagram before correction of a scan image, and FIG. 2B is a diagram after correction of a scan image.

For example, in the step of acquiring the scan image, when the scan image is scanned in an inclined state, the scan image is rotated to a predetermined position so that the text is more correctly recognized in the text file conversion (S150) can do.

A method of modifying the scan image will be described with reference to FIGS. 2A and 2B. If the width or height of the unit segment within the area of the scanned image is constant and if any one of the width and the height is longer than a predetermined length, the line segment is determined as a line segment, And a second axis perpendicular to the first axis, calculating an angle between the line segment and an adjacent axis in the scan image, rotating the scan image by the calculated angle, Line segments and the adjacent axes.

2A and 2B, the reference axes are X1 and Y1, the horizontal and vertical axes of the scan image are X2 and Y2, and the intersection of the X1, X2, Y1, and Y2 axes is O. FIG. As shown in FIG. 2A, the X1 axis and the X2 axis may be discordant. In this case, there is an error between the X1 axis, the X2 axis, and the O, so that the scan image is rotated by this angle.

At this time, the degree of discrepancy between the X1 and X2 axes can be calculated by the contents of the scan image. For example, when a region of a character in the scan image is checked and a position of a character that is firstly started in each row is connected, a line segment L1 is generated. By making such a line segment L1 stand upright, You can modify the image. Alternatively, the recognized line segment L2 may be horizontally or vertically recognized by recognizing the line segment L2 formed in the horizontal or vertical direction in the scanned image. The line segment L2 may be a dividing line for separating adjacent regions.

That is, in the case of paper newspapers and periodicals, the position of the first letter of each line is sequentially recognized, and when these are connected, a line segment L1 is generated. Since the line segment is generally formed vertically vertically, when the line segment has a non-vertical angle, the line segment is rotated at an angle corresponding to the difference so as to stand upright. As described above, in the case where the scanned image is scanned with a slight inclination, the operator sometimes pulls the paper document in a somewhat distorted state, or the paper document is tilted while being partially unfolded in the process of scanning the paper document. At this time, the center of rotation may be the center O of the paper document or the center O in the virtual space of the program that rotates the scanned image.

In some cases, a line segment (L2) formed somewhat longer in the horizontal or vertical direction is described in the paper document. In this case, the horizontal or vertical line segment is recognized and rotated by a difference from the horizontal or vertical angle. For example, if there is a thin or thick line segment horizontally in a paper document, the line segment should have an angle of 0 °, but if it has some error, it can be rotated only by the amount corresponding to the error. This can be referred to as a fix orientation step.

In other words, in the case of paper documents, there are many ways to find the slope of the image, but in paper newspapers there is a specific method. Generally, there is a header (refer to a line segment in the red figure of FIG. 3B) indicating "date and time information" in the nature of a newspaper, and there is a line segment connecting arbitrary two points. In addition, there may be an advertisement section line (refer to a line segment in the red figure in Fig. 3B) indicating an advertisement area, and a dividing line between the article and the article (see a line segment in the red figure in Fig. 3A) necessarily exists in the lateral newspaper or the vertical newspaper. At this time, the segment, the line segment indicating the advertisement area, or the dividing line between the article and the article may correspond to the Y2 axis in FIG. 2A.

Here, the inclination can correct the inclination by finding the line segment Y2 and calculating the angle formed by the line segment Y2 with the Y1 axis. In this case, the angles can be obtained by using the trigonometric formula on the right triangle after realizing the right triangle formed by Y1 axis, Y2 axis and O.

On the other hand, when the thickness of the unit segment described later is constant and the horizontal length of a specific size is larger than a predetermined size (for example, 20 times) compared with the vertical length, or when the vertical length is larger than a predetermined size 20 times), these segments can be a line segment serving as a reference for correcting the slope.

2C is a view of a scan image in a state before the tilt is corrected, and FIG. 2D is a diagram illustrating a scan image in a state in which the tilt is corrected, FIG. 2E is a view showing a state in which margins are set (added) after margins are removed.

When the tilt is corrected, margins other than the newspaper area are removed. Margin removal is performed by increasing the area starting from the left side and obtaining one large frame segment. For example, in FIG. 2C, the area from the left end to the moment when the imaginary vertical line moves to the right and touches the image of the paper document is stored, and the imaginary vertical line is moved from the right end to the left. The area of the scanned image of the paper document in the horizontal direction can be obtained. Similarly, the areas of the upper and lower scan images can be obtained. That is, the paper newspaper is automatically cut off (AutoTrim) by excluding the frame segment from the actual scan area size of the scanned image. This corresponds to the incision in Fig. 2C. That is, the margin elimination is the same as the result of rotating the area of the actual paper document while the white background part shown in FIG.

After that, the margin is set. Since the blank newspaper is not better than the actual newspaper version, it is necessary to do this process.

The margin (the border (frame) portion of FIG. 2E) is set by adding the up, down, left, and right margins to the background to the specified size. That is, an empty frame is created by adding the margin size to the frame size with the margin removed, and then the margin is set by placing the paper newspaper in the empty frame.

Then, the image file is converted into a pdf file (S133). In this case, the file name includes information such as newspaper code, year, month, day, edition, face, section, To be intuitive.

Meanwhile, FIG. 4 is a flow chart related to the tilt correction according to an embodiment of the present invention, and is a diagram for more specifically explaining the tilt correction and PDF conversion step (S130) in FIG.

Hereinafter, a description will be made with reference to FIG.

The above-described method of correcting the tilt of the scan image can be performed after obtaining the connecting unit segment, which is the minimum unit connected to each other. The process of connecting point unit segmentation (S131) according to an embodiment of the present invention is a process of obtaining the unit segment. This process is performed by locating the unit segment in the vicinity of four pixels of up, down, left, Determining a plurality of valid pixel clusters by determining a valid pixel clusters for clusters having a predetermined number (k) or more of pixels by irradiating the black pixel clusters with the adjacent pixels, When the length of the set of pixel blocks is equal to or greater than a predetermined length, the set of pixel blocks is determined as a unit segment.

That is, valid pixels for the obtained arbitrary pixel with respect to the adjacent pixels (8 pixels excluding the arbitrary pixel positioned at the center in the 3 * 3) are identified, and when the length of the pixel block composed of valid pixel blocks is valid By including it in the calculation of the number of strokes. At this time, the outer frame of the actual character recognition area of the paper newspaper among the unit segments can be obtained.

In other words, in order to obtain a unit segment, pixels are recorded while moving from the upper left area of the paper newspaper to the right side to obtain a unit segment around pixels having an effective size, and pixels having a size enough to become noise are removed A valid unit segment is calculated.

In order to correct the tilt, a reference line segment should be selected. In an embodiment of the present invention, a line segment serving as a slope correction reference is selected from unit segments that are the minimum unit segments. At this time, whether or not adjacent pixels are connected to form a single segment is determined by the distance between adjacent pixels having a valid size. If the distance between adjacent pixels is greater than a predetermined distance, . In this case, most of the text is formed to be adjacent. In this case, only the unit segment is formed, and the state is still before the character segment is formed. That is, it is preferable that the slope and the margin correction (S132) form a character segment, a character row segment, a block segment, and a group segment in a state in which the slope is found by searching for a segment that can be a reference among the unit segments. However, the present invention is not limited to this, and the tilt correction may be performed after the unit segment, the character segment, the character row segment, the block segment, and the group segment are simultaneously performed.

Then, the image file is converted into a PDF file (S133). The step of converting the image file into a PDF file (step S133) may be performed only before cropping, which is described later, and is not necessarily performed before the segmentation step (S140).

In one embodiment of the present invention, the scan image is divided into a text area and an image area. The text area is divided into a unit segment, which is a minimum unit segment connected to each other by using pixels, A character line segment formed by a combination of one or more of the character segments, a block segment formed by a combination of one or more character line segments, and a character string segment formed by a combination of one or more of the block segments And sequentially segmenting the group segments to be formed. As described above, the segmentation operation from the small size segment to the large segment can obtain the position information (coordinate information) about the text and image area in the paper document.

Hereinafter, segmentation according to an embodiment of the present invention will be described.

5A and 5B are diagrams for explaining segmentation according to an embodiment of the present invention, and can be applied to a character segment, a character row segment, a block segment, and a group segment, have. Since the method of unit segmentation has been described above, a method of character segmentation will be described with reference to FIGS. 5A and 5B.

A method of segmenting into a character segment according to an embodiment of the present invention includes dividing a first unit segment (blue square in FIG. 5) and a second unit segment The area of the first unit segment is S ₁ , the area of the second unit segment is S ₂ , and the area of the first unit segment and the second unit segment (1) to (4) are satisfied, when the overlapping area of the first unit segment and the second unit segment is S _I and the area of the state where the first and second unit segments are combined is S _u .

S _I ≥k ₁ S ₁ ------------------- (1)

_{S I ≥k 1 S 2 ------------------- (} 2)

S _1? K ₂ S _u ------------------- (3)

S _2? K ₂ S _u ------------------- (4)

In the above equation, k ₁ and k ₂ are preset constants determined according to the typeface.

That is, in the embodiment of the present invention, the overlapping area between adjacent segments is obtained in the process of character segmentation, which will be described in more detail with reference to FIG. 5B.

Two arbitrary segments A1 and A2 are selected in the unit segment set, and the areas S1 and S2 of them and the overlapping area S _I between the two segments and the area S _u when the two segments are combined are obtained. That is, the unit segment 'i' and the unit segment 'A1' and 'A2' in the Korean alphabet segment can be segmented into 'low' character, which is one character segment, because the areas overlap.

If any one of the conditions (1) to (4) is satisfied, the two segments are combined into one segment. At this time, the value of k1 was set to be 0.1 in the case of print chain, and the value of k2 in the case of handwriting was set to 0.95 regardless of the printed matter and the writing body. In the ideal print chain, the value of k1 may be set to 0, but in practice, it may be slightly overlapped even in the case of segments belonging to different characters due to noise, tilting, or other influences.

Since the values of k1 and k2 may vary depending on the type of the typeface, the values are merely illustrative and not restrictive. In this process, a 'low' character segment, which is one character segment, is considered in consideration of the overlapping area of each of the 'X' unit segment A1 and the 'X' unit segment A2. Furthermore, the 'before' character segment A4 is completed by combining with the 'a' unit segment A3 where the areas overlap with the 'low' character segments A1 and A2. The character segment A4 is composed of a combination of one or more unit segments and is represented by a rectangular area A4 including three unit segment areas A1, A2, and A3 in the case of FIG. 5B. That is, the character segment may be defined as a rectangular area including an area of a plurality of adjacent unit segments. As a result, a region of a character line segment, a block segment, and a group segment, which will be described later, can also be expressed as a rectangular region.

A character line segment composed of character segments is obtained by averaging the sizes of the plurality of character segments to obtain a reference width (w ₀ ) and a reference height (h ₀ ) of the character segment, and a plurality of the character segments in a first direction For example, in the horizontal direction).

FIG. 6 is an illustration for explaining a character line segmentation method according to an embodiment of the present invention, which will be described below with reference to FIG. 6. FIG.

The character segments in the predetermined size range are selected in consideration of the width and the height of the character segments considering the reference width w ₀ and the reference height h ₀ to form character line segments, If the interval between adjacent character segments is greater than or equal to the predetermined interval d along the first direction or the width or height does not overlap along the second direction perpendicular to the first direction, a new character line segment is formed . In FIG. 6, horizontal writing is an example. In vertical writing, the interval between the character segments should be based on the vertical direction rather than the horizontal direction. That is, if the first direction is the horizontal direction, d is the horizontal distance between the current character segment and the last row of the previous row segment set, and if the first direction is the vertical direction, d is the current character segment and the previous row segment set Is the vertical distance from the last character segment of < RTI ID = 0.0 > In addition, the character segment may form a word segment forming one word. At this time, the spacing between adjacent character segments may be a reference of a width or a height of a reference character.

The method of character line segmentation according to an embodiment of the present invention increases the accuracy of segmentation and shortens the segmentation time by constructing a character line based on the segment set obtained as a result of the previous step.

Generally, the width of a character line is similar to that of a character, and segments placed on one line have a large degree of overlap when viewed in the direction of a character line. When ignoring various influences such as noise and skewing, segments belonging to different lines overlap each other There is no case. Therefore, it is possible to construct a character line by looking at the degree of overlap between character segments in the character line direction.

That is, by selecting segments having a reference width (w ₀ ) and a reference height (h ₀ ) of a character segment and having a width of a character segment of 0.4 w ₀ or more and 1.2 w ₀ or less, unnecessary segments (noises) .

Among the obtained segment sets, the first character segment is used as the first line segment, and a row segment set is constructed. Then, starting from the second character segment, if it satisfies the following condition, it is put into the line segment set as a new character line, otherwise it is combined with the last line of the line segment set. At this time, if the characters are compared with the row segments and the widths do not overlap, a new row is generated and a row segment is generated. For example, if the overlap width is within 24% of the reference width, it is added to the last line.

In addition, a method of segmenting into a block segment formed by the character line segments may include storing character segments in different formats according to a reference width and a reference height of the character segment, Repeating the step of merging the character line segments having the same length and within the predetermined range, and in the step of merging the character line segments, the character line segments spaced apart from the neighboring character line segments by more than a predetermined interval in the second direction The segment is formed by forming another block segment, and then the character row segment having the reference width and the reference height of the character segment are processed into another block.

More specifically, the reference width (w ₀ ) and the reference height (h ₀ ) of the character segment of the main text are obtained and the row spacing can also be calculated. Therefore, a row having a string larger than the reference size in the row unit segment Title block segment. A set of rows having a predetermined pattern is compared to find a step interval of a certain pattern, and a step interval is set as a step interval in which a non-overlapping area within a predetermined size (1/3 of the reference width) is written vertically in horizontal writing.

FIG. 7 is a scan image of a block segmented paper newspaper related to the present invention, and FIG. 8 is a scan image of a paper newspaper containing photographs and advertisements related to the present invention, which will be described with reference to FIG. 7 and FIG. When blocking the next line segment in the current line segment, the length of the line segment must be the same within a certain range, and when you expand the line by combining the same line segments of the same length, Construct a new block.

At this time, if a row that is larger than the reference width (w ₀ ) and the reference height (h ₀ ) of the text segment is continuous, the block is composed of one block when the corresponding row does not overlap with other rows, . If the rows repeated with a continuous row length based on the reference width (w ₀ ) and the reference height (h ₀ ) are combined, the text becomes a body. In this way, most of the blocks are completed when the merging of row segment units is completed.

The area of the blocked segments can be classified based on the character size. In the case of the title and the subtitle, the character size is determined based on a predetermined reference, and the text is determined based on the reference character size. On the other hand, most newspaper titles are already defined by newspapers. That is, it can be classified into a major title segment, a subheading segment, and a body segment depending on the character size.

On the other hand, if the width and height of the unit segments are larger than the predetermined size, they are included in the image segment and excluded from the character row segment. For example, a segment having a width of 5 times (500%) or more of the standard character size is extracted from the unit segment, and when overlapping portions of the unit segments are overlapped by 500% or more of the standard character size, they are included in the image segment Deletes from the character row segment. In this way, the image area can be segmented, and the area where pixels of other colors other than black pixels are formed can be segmented into image areas. This is because texts are written in black, while ads or photographs contain colors other than black. If the unit segment is formed of a rectangular box and is larger than the reference size by a predetermined size, the unit segment may be determined as an advertisement or a photograph segment regardless of the presence or absence of text.

Generally, the advertisement of a paper newspaper is managed by the advertisement department of the newspaper company, and the position of the advertisement is always fixed. That is, from the bottom to the 1st to 15th stage (full advertisement), the position of the deformation stage is different according to the hole surface and the even surface. In addition, the advertisement is generally formed from left to right to distinguish between newspaper and advertisement, or there is a dividing line formed from the upper side to the lower side, or the advertisement is surrounded by a box. This is also true of deformed ends. In a paper newspaper, a segment of the photo area is judged as an advertisement by using such a pattern.

Thereafter, it is necessary to construct those of the block segments that form the same content article into a single group. For example, a group segment is formed by grouping a title, an article content (body text), a byline, and the like into one group. This will be referred to as group segmentation.

A method of segmenting into group segments according to an embodiment of the present invention is characterized in that the segment segment is located at a lower portion of a first block segment including a first character segment having a largest width and height of character segments among adjacent block segments, Grouping block segments that overlap at least partially along a second direction with one block segment in the same group and spacing or spacing a block segment formed in the same group along the first direction or the second direction, And a second block segment including the first character segment or a block segment intersecting the dividing line among the block segments formed in the same group are grouped into different groups.

This is because the title of the newspaper with the largest letter size is located at the top of the newspaper, and the subtitle or the text is located beneath it.

To group them into the same group, first compares the first block segment with the adjacent block segment rectangle, and if there is an intersection, adds a new segment. Blocks existing at the bottom of the title based on the horizontal newspaper are grouped into a single article until the next title and the advertisement separator line. Then, the flow of the text box is grouped while continuing from the left to the right, and the text box without the title box in the area which is not overlapped with the title title area is combined into the same group. At this time, if the group formed is not a rectangle, the region is created as a polygon.

Most of vertical newspapers written in vertical writing are polygons. Vertical newspapers are also based on titles. They are divided into single lines and dividing lines, and the flow of writing flows from top to bottom from right to left . Vertical newspaper consists of 15 steps from top to bottom. For example, if you meet another title block segment, segment segment, or photo block segment on the left side of the text box, the text box is located on the right side of the bottom.

Meanwhile, in an embodiment of the present invention, the unit segment, the character segment, the character row segment, the block segment, and the group segment have coordinate information, and the group segment forms a polygon by the coordinate information. The group segment is determined as a closed region formed by a path that does not overlap while moving along the outer periphery of the adjacent block segment. That is, if the contents of the same article are formed, they should be placed in one group segment. The group segments are generally formed in a rectangular shape, but the shape is not limited thereto, and may be formed in various shapes of polygons.

FIG. 9 is a conceptual diagram of a method of processing a group segment according to an embodiment of the present invention, and FIG. 10 is a flowchart of a method of polygon processing of a group segment according to an embodiment of the present invention.

Hereinafter, a method of polygoning a group segment according to an embodiment of the present invention will be described with reference to FIGS. 9 and 10. FIG.

First, an aggregate is formed of each block segment, and if the first block segment is compared with the next block segment, or is included in the first block segment by 100%, the small segment is removed. Then, it is expanded based on the horizontal and vertical snap sizes and replaced with a set of new block segments. Here, the size of the snap means the distance between the traverse and the vertical line of the article. In other words, in the case of a horizontally-typed newspaper, the horizontal spacing distance between adjacent block segments is referred to as a short interval, and the horizontal spacing between block segments forming the same group segment located below the title segment . And extends the area of the block segment based on the step interval. By doing so, the area overlapping with the adjacent block segment is generated, and the same group segment as the neighboring block segment is formed.

This will be described in more detail with reference to FIG.

The intersection of the first block segment B1 and the subsequent segments B2 and B3 is added. Information on each corner of the block segment rectangle is generated to construct an array. (Denoted by x) included in the group segment in the array. Compares all points from the first position (B11) and finds the next coordinate (next) information. Do not compare added points from now on. If there are any remaining points, treat them as new single polygons.

That is, the block segments B1, B2, and B3, which are adjacent to each other and overlap at least a part of the area, are disposed, the position information of the overlapping area (x marked points) is ignored, ⊙ point) is newly added. Then, from the first block segment (B1), positional information that has the largest area along the outline is found. This is obtained by moving along the arrow direction in FIG. 9, and the area of the group segment is obtained by the line segments connecting the outermost points.

10A to 10D, FIG. 10A shows a block segmented state by a title, a subhead, and a body. FIG. 10B shows a state where block segments larger than adjacent block segments are formed in FIG. 10A FIG. 10C shows a state in which text segments other than an image segment are grouped, and FIG. 10D shows a group segmentation of an image segment and a text segment.

More specifically, FIG. 10A shows a state in which a plurality of block segments are divided into a major title segment, a subheading segment, a main segment, and an image segment, respectively. In this case, grouping between the headline segment and the subheading segment is not performed. Then, the major title segment and the subheading segment are grouped by extending the area of the major title segment by a multiple of the height of the major title segment, which is the snap size. At this time, the body segments are grouped into body segments of a larger area by expanding the block segment by a multiple of a short interval (which is not necessarily an integral multiple). By this process, adjacent block segments are grouped as shown in FIG. 10B. Further, as shown in FIG. 10C, text segments forming the same article are grouped underneath the heading segment. This is accomplished by overlapping the extended heading segment with the extended region of the body segment. In this way, text segments are grouped. Next, as shown in FIG. 10D, the group segment is completed by overlapping with the text segment by extending the area of the image segment.

The image segment may include a photo segment and an advertisement segment, the photo segment means a photo region, and the advertisement segment means an advertisement region. However, the image segment may be converted into an image file in the text conversion step (S150) without segmenting.

In the segmentation step, the segments are not simply recognized as images but recognized as characters. Thus, it can be understood that a text segment (including a character, a character row, a block, and a group segment) means a segment other than an image segment.

In the embodiment of the present invention, the content of the scanned image, more specifically, the various text segments, is read out to extract the characters recorded in the paper document (newspaper) (S150), and the text file has a txt extension as an example.

At this time, the conversion into the text file may be performed by a character recognition program, for example, an optical character recognition (OCR) program, and the type is not limited.

In an embodiment of the present invention, text is extracted from a scanned image by a character recognition program, and it is not easy to extract text in the case of an old document. This is especially true for the case of kanji. For this, in an embodiment of the present invention, the character region may be divided into a predetermined size to generate a segment, and characters recognized in the generated segment may be recognized and stored as text.

At this time, it may further include modifying the segment by specifying information to be added to the generated segment. In the step of modifying the segment, a sample of a similar character may be extracted, a representative character may be extracted, and then the character may be transformed into a representative character so that repeated characters can be recognized without errors. In order to correct the segment, a corrected character for the already learned error character may be applied.

Thereafter, the text extracted from the paper document may be subjected to a step of correcting text such as spelling, spacing, and punctuation of extracted characters (S160). From this point on, the contents of the paper document can be recognized as characters.

In this case, the correcting step may include a step of recommending a replaceable word when an erroneous word different from a word included in a predetermined dictionary is found, and a step of replacing the erroneous word with a recommended word by selecting the recommended word It goes through. That is, a word that is different from a word included in a predetermined dictionary (module) is found to be an erroneous word, and the erroneous word is replaced with an appropriate word.

It is not necessary to replace the error word with a recommended word. In this case, if there is no candidate word to replace the error word, a new word is added to the error word, So as to continuously correct the new word appearing in the new word. This is intended to add a new word in the absence of a predefined dictionary, but replace it with the same word if it appears later. As described above, in the embodiment of the present invention, the same correction process is applied to characters (words) identified as the same error.

According to an embodiment of the present invention, an inspection module for checking whether a word is erroneous is used for detecting the erroneous word. The inspection module includes a dictionary module in which a plurality of words are stored, A restoration module for restoring the original form of the word by analyzing the morpheme of the word, a Hanja conversion module for converting the Chinese character into Hangul, a morpheme analysis module for analyzing the word morpheme, Managed speech information management module, and a rule encoder module that processes words or phrases that are frequently mistaken for spelling, foreign language, input errors, spacing, or pasting as separate rules.

When the text file conversion (S150) is completed, the text file is converted into an extensible markup language (XML) (S170), and the XML file is stored.

The extensible markup language (XML) is a language for electronic commerce, a number of web portals, content services, and information processing applications implemented on the Internet. Such XML is a general-purpose language for structuring documents for transmission over the World Wide Web (WWW). In addition, the extensible markup language (XML) is designed to describe content that is not intended for screen output, and allows the user to define a new tag to describe the structure. XML has the advantage of providing various characteristics according to data structure such as data structure reuse, flexibility of output format, search function of data structure, and so on, since data structure, contents, and output format are defined as whole data. The basic unit in XML is an Element, which allows the user to freely define the element, and can have a text or element or a mixture of text and elements within the element. Such XML has become a standard for information representation and electronic data exchange on the Internet due to the flexibility of the information expression of the language itself. Since the XML is a known language, a detailed description will be omitted below.

In the embodiment of the present invention, the paper document may be a periodical publication. Hereinafter, the paper document will be mainly described as a periodical, especially a paper. However, the scope of rights of the present invention is not limited thereto, and the case where a paper document is a monograph is also included.

On the other hand, in the embodiment of the present invention, the text file and the image file (segment) store the file name including the date, edition, and face on which the paper newspaper is issued. As described above, since the file name itself includes the contents of the file by storing the file name in association with the issue date, utilization and management are easier. More specifically, it is possible to easily determine whether a scan image is missing or duplicated by extracting a file name, and can easily confirm the number of faces issued on a specific date.

In addition, in an embodiment of the present invention, a file name can be stored using a path from an upper folder (or directory) to a lower folder (or directory). That is, the file names of the files in the specific folder can be named including the folder names in the path of the folder. By naming the file along the path, it is easy to find the location of a specific file, and it is easy to check whether the file is stored in the correct location.

As described above, by storing the text file of the paper newspaper as the file name associated with the issue date, edition, and number of pages, it is possible to easily check whether or not the text file corresponding to the paper document is missing. In other words, it is possible to collectively check whether there is a missing page or an entire advertisement page by collating the information of the paper file with the text file, and the information of the already held paper document is, for example, Image. In addition to the scan image, a pdf file may be used when the scan image is converted into a pdf file.

Hereinafter, the step of inputting an article will be described in more detail.

First, the contents of newspaper articles are created and stored in units of pages. In other words, if there is a plate, the name of the file is saved as newspaper (2), publication date (8) __ plate (1) __ (2) .txt, Save it in the same format as the date of issue (8) __ (2) .txt. The number in parentheses means the number of characters. For example, the text file name may be stored as GJ_19890102_01.txt, GJ_19890102_2_01.txt. The numbers in parentheses are only examples.

The text file makes a file on a surface basis. The text file includes a title, an article classification, a body text, and a byline, and the title, the article classification, the body text, and the byline are separated by respective delimiters. For example, all articles can be typed as @@@ title, ### article classification, $$$ body, %%% @ @@, ###, $$$, %%% Title, article classification, text, journal name (or byline). Even if you are not a journalist, you should mark it as %%% even at the end, such as a hospital or a professor. At this time, the byline refers to a person who writes articles such as journalists and contributors.

The contents can be easily classified by the text file thus created. That is, after predetermined delimiters, predetermined contents are inputted, so that it is easy to perform statistical analysis. For example, the article classification may be determined by words appearing repeatedly in the text, or the words frequently appearing in paper newspapers during a specific period may be analyzed to know what issues are in the period.

In addition, in the embodiment of the present invention, classification information may be constructed in advance, for example, the article classification may be determined according to the appearance frequency after character filtering. In addition, similar classification information may be constructed in advance according to the user, so that the error of the article classification can be reduced according to the user. At this time, the user can be specified by the name of journalist or e-mail. That is, by extracting the byline information from the input article and analyzing the byline information, the article name, e-mail, or article according to the individual contributor can be classified so that information can be provided at the time of subsequent article classification.

At this time, the bi-line information is extracted from the big data article including a plurality of text files, the bi-line information is analyzed and the journal name, e-mail, and the individual contributor are stored. Then, Line. This is because the name of a specific reporter often appears in periodicals such as newspapers, so that if a specific reporter's name is wrong, it can be corrected.

In an embodiment of the present invention, in the case of a paper newspaper, the text file, the title, the article classification, the body text, and the bi-line can be converted into NewsML (S170). For example, text, work XML, etc. can be combined with basic environment information and converted into NewsML. It is converted into NewsML by extracting tags from items such as title, classification, and byline and then creating a table. More specifically, a title and / or a class are extracted to perform an inspection, and a bi-line check list is generated. Check the generated list for errors, and if there are no errors, go to the table creation step. In the table creation step, a table can be created to correct the error journal name or e-mail extracted from the contents check, and the errors occurring afterwards are automatically corrected.

The NewsML is a standard format based on an Extensible Markup Language (XML) for defining news in News Markup Language, and manages all types of multimedia news from production to storage, It is an XML-based standard for

After the NewsML and the PDF file of the scanned image are finally completed to extract various data lists, various tests and lists are generated. More specifically, NewsML article entry list, PDF file list, NewsML + PDF cross list, NewsML full list, byline extraction, without line line extraction, NewsML basic edition, PDF basic edition, monthly (PDF + text + . The PDF refers to a file converted from paper newspaper to PDF. In this way, statistical analysis is possible by various data lists, and big data can be utilized.

Meanwhile, in one embodiment of the present invention, the ground mapping information can be generated to provide services on the web. For this purpose, in an embodiment of the present invention, the region of characters and images is recognized, . That is, coordinate information of a text area (segment) is generated on the ground, and an article image or a PDF cropped according to the coordinates is generated to obtain the same effect as viewing a paper newspaper on the web. At this time, in one embodiment of the present invention, crop refers to extracting a paper document as an image.

11 is a view for explaining cropping of a paper newspaper related to the present invention. Referring to FIG. 11, first, the scan image is converted into a PDF file, and coordinates of the group segment are generated on the PDF image Next, a text file corresponding to the coordinates of the generated group segment among the files in the NewsML is selected and registered.

More specifically, coordinate information is generated from an image such as a PDF, JPG, or the like, which is to be linked with a NewsML article on a digital basis and cropped. Then, the corresponding article is selected from the article list located on the right side of FIG. 11, and the corresponding block segment is selected with the mouse on the left screen. When the mapping is selected in this manner, the mapping article number is registered in the list, and when all the mapping is completed, the surface is completed.

At this time, since the digital surface to be subjected to block processing has already been segmented into blocks, selection is sufficient. That is, since the block segment and the group segmentation are made earlier, the corresponding area of the paper newspaper is displayed as it is when the specific article is selected in the digital aspect.

On the other hand, in some cases, a newspaper company has a pdf file. In such a case, a new copy may be processed in a text copy process. In this case, if you are copying text from a pdf file into the article area of a paper newspaper, extract the newline before pasting from the clipboard, and then replace the contents of the clipboard with the replacement document. For example, a new line can be arranged based on "~~". In this manner, the readability of the reader can be improved by correcting the sentence automatically processed by the new line.

The foregoing detailed description should not be construed in all aspects as limiting and should be considered illustrative. The scope of the present invention should be determined by rational interpretation of the appended claims, and all changes within the scope of equivalents of the present invention are included in the scope of the present invention.

Claims (16)

Obtaining a paper document to be digitized;
Scanning the acquired paper document to obtain a scanned image of the paper document;
The scan image is divided into a text area and an image area, and the text area is divided into a unit segment, which is a minimum unit segment connected to each other by using pixels, a character segment formed by a combination of at least one unit segment, A character row segment formed by a combination of one or more of the character segments, a block segment formed by a combination of one or more character row segments, and group segments formed by combinations of one or more of the block segments are sequentially A segmentation step of forming a segment;
Converting the contents of the paper document into a text file by extracting characters from the segmented scanned image, and converting the image area into an image file; And
And generating an Extensible Markup Language (XML) using the text file,
Wherein segmenting into the character line segments comprises:
Obtaining a reference width (w ₀ ) and a reference height (h ₀ ) of the character segment according to a size of the plurality of character segments;
Aligning the plurality of character segments in a first direction;
Constructing a character row segment by selecting character segments within a predetermined size range in consideration of the width and height of the character segments considering the reference width w ₀ and the reference height h ₀ ; And
Wherein the spacing between adjacent character segments in the sequence is greater than or equal to a predetermined interval along the first direction or the width or height does not overlap along a second direction in which the character segment is perpendicular to the character line segment and the first direction And a new character line segment if the character string is a new character line segment. The method according to claim 1,
Wherein segmenting into unit segments comprises:
Obtaining a pixel block including neighboring pixels located in four neighborhoods of upper, lower, left, and right with respect to a certain pixel within a predetermined rectangular area;
Determining a valid pixel mass when the pixel mass of the pixel mass has a predetermined number or more of pixels and obtaining a plurality of the effective pixel masses to obtain a set of pixel mass; And
And determining the set of pixel clusters as a unit segment if the length of the set of pixel clusters is greater than or equal to a predetermined length. 3. The method of claim 2,
Wherein segmenting into the character segments comprises:
Selecting a first unit segment and a second unit segment adjacent to the first unit segment in the set of unit segments including the unit segments;
The area of the first unit segment is S ₁ , the area of the second unit segment is S ₂ , the overlapping area of the first unit segment and the second unit segment is S _I , (1) to (4) when an area of a state in which two unit segments are combined is S _u , the step of determining the same character segment.
S _I ≥k ₁ S ₁ ------------------- (1)
_{S I ≥k 1 S 2 ------------------- (} 2)
S _1? K ₂ S _u ------------------- (3)
S _2? K ₂ S _u ------------------- (4)
In the above equation, k ₁ and k ₂ are preset constants determined according to the typeface. delete The method according to claim 1,
Wherein segmenting into the block segments comprises:
Storing character segments in different formats according to the width and height of the character segments;
Repeating combining character string segments having the same length within a predetermined range while the width and height of the character segments are within a predetermined range;
The step of combining the character line segments and the character line segments spaced apart from the neighboring character line segments by more than a predetermined interval along the second direction form another block segment; And
And processing the character line segments having different widths and heights of the character segments into different blocks. 6. The method of claim 5,
Wherein segmenting into the group segments comprises:
A block segment located at a lower portion of a first block segment including a first character segment having a largest width and a height of a character segment among adjacent block segments and at least partially overlapping the first block segment along a second direction, Forming into the same group; And
A second block segment or dividing line spaced apart from the block segment formed in the same group by a predetermined interval or along the first direction or the second direction or including the first character segment among the block segments formed in the same group, And grouping the intersecting block segments into different groups. The method according to claim 6,
Grouping the image segments located between the grouped block segments into the same group. ≪ Desc / Clms Page number 20 > 8. The method of claim 7,
After the segmenting step,
Further comprising the step of correcting the scanned image by rotating the scanned image to the correct position when the scanned image is scanned in a skewed manner. 9. The method of claim 8,
Wherein the modifying the scan image comprises:
Determining one of the unit segments as a segment if the width or the height is constant and either one of the width and height is longer than a predetermined length;
Determining whether the line segment is formed adjacent to a first axis of the scan image and a second axis perpendicular to the first axis;
Calculating an angle between the line segment and an axis adjacent to the line segment on the scan image of the first axis and the second axis; And
And rotating the scan image by the calculated angle to match the line segment and the adjacent axis. 8. The method of claim 7,
If the width and the height of the unit segment are greater than a predetermined size, excluding the character string segment from the image segment. 11. The method of claim 10,
Wherein the paper document is a paper newspaper. 12. The method of claim 11,
Wherein the unit segment, the character segment, the character row segment, the block segment, and the group segment have coordinate information, and the group segment forms a polygon according to the coordinate information. 13. The method of claim 12,
Wherein the group segment is formed as a closed area formed by a path that does not overlap while moving along the outermost part of the adjacent block segment. 14. The method of claim 13,
Wherein the text file includes a title, an article classification, a body text, and a byline, and the step of generating an extensible markup language (XML) using the text file includes: The byline is separated by each delimiter,
Generating News News Markup Language (NewsML) by combining the extended markup language and basic environment information. ≪ Desc / Clms Page number 19 > 15. The method of claim 14,
Converting the scan image into a PDF file;
Generating coordinates of the group segment on the PDF image; And
And selecting and registering a text file corresponding to the coordinates of the generated group segment among the files in the NewsML. The method according to claim 1,
Wherein the conversion into the text file is performed by an optical character recognition (OCR) program.

Images