KR101831204B1 - Method and apparatus for document area segmentation - Google Patents

Abstract

Proposed are a method and apparatus for segmenting a document area. The method for segmenting a document area includes the steps of: detecting text line candidates by clustering connection elements included in a document image; detecting a text line by removing a non-text element by performing filtering using two different threshold values with regard to the detected text line candidates; detecting a non-text area by performing repetitive XY cut after removing the detected text line from the document image; and detecting a document boundary by comparing the detected non-text area with an interest area. Accordingly, the present invention can accurately detect the text line with regard to a document image with a complicated layout.

Description

[0001] METHOD AND APPARATUS FOR DOCUMENT AREA SEGMENTATION [0002]

The embodiments disclosed herein relate to a method and apparatus for detecting text areas, non-text areas and document boundaries in a document image.

In recent years, the performance of cameras provided in personal devices such as smart phones has been improved. Therefore, instead of scanning a document using a scanner, a document is photographed using a camera provided in a smart phone or the like, Is often used instead of a scan image.

However, in the process of converting a document image into an electronic document, character recognition results are deteriorated when an element other than text is included in performing character recognition. Therefore, in order to improve the accuracy of the electronic document conversion, a preprocessing process is required to determine which part of the text area and the non-text area are in the document image, and the boundary of the entire document.

Korean Patent Laid-Open Publication No. 2000-0033954 discloses a method and apparatus for dividing document images based on clustering. More specifically, Japanese Patent Application Laid-Open No. 2000-0033954 discloses a method and apparatus for dividing document images based on clustering, Clustering is performed on the basis of only the character regions, and the character regions are divided and extracted based on the result.

On the other hand, the background art described above is technical information acquired by the inventor for the derivation of the present invention or obtained in the derivation process of the present invention, and can not necessarily be a known technology disclosed to the general public before the application of the present invention .

Embodiments disclosed herein are aimed at providing a method and apparatus capable of detecting a text area, a non-text area, and a document boundary in a document image with high accuracy.

According to an embodiment of the present invention, there is provided a method of dividing a document region, the method comprising: clustering connection elements included in a document image to detect text line candidates; Removing the non-text elements and detecting a text line by performing filtering using two different threshold values for the candidates, removing the detected text line from the document image, and then repeating the repetitive X-Y cut Detecting a non-text area by performing a search on the detected non-text area, and detecting the document boundary by comparing the detected non-text area with the area of interest.

According to another embodiment, a computer program for performing a document area dividing method, the method comprising: clustering connecting elements included in a document image to detect text line candidates; Removing the non-text elements and detecting a text line by performing filtering using two different threshold values for the candidates, removing the detected text line from the document image, and then repeating the repetitive X-Y cut Detecting a non-text area by performing a search on the detected non-text area, and detecting the document boundary by comparing the detected non-text area with the area of interest.

According to another aspect of the present invention, there is provided a computer-readable medium having recorded thereon a program for performing a method of dividing a document region, the method comprising: clustering connection elements included in a document image to detect text line candidates Removing the non-text elements and detecting a text line by performing filtering using the two different threshold values for the detected text line candidates, removing the detected text line from the document image, Detecting the non-text area by performing a repetitive X-Y cut, and detecting the document boundary by comparing the detected non-text area with the area of interest.

According to another embodiment, the document area dividing apparatus includes an input / output unit for receiving an input related to processing of a document image and for displaying a situation and a result of processing of the document image, a program for performing document area division And a control unit for performing an area division of the document image by executing the program, wherein the control unit detects clusters of the text line candidates by clustering connection elements included in the document image, Removing the non-text elements and detecting the text line by performing filtering using the two different threshold values for the detected text line candidates, removing the detected text line from the document image, a non-text area is detected by performing a cut operation on the detected non-text area, The document boundary can be detected by comparing with the station.

According to any one of the above-mentioned means for solving the problems, it is possible to accurately detect a text line even for a document image having a complicated layout by clustering connection elements included in the document image and then performing projection.

In addition, by applying a classifier using two different threshold values, it is possible to improve the accuracy in removing non-text elements.

In addition, by detecting the document boundary by comparing the detected non-text region with the region of interest given as the initial value, the accuracy of document boundary detection can be increased.

The effects obtained in the disclosed embodiments are not limited to the effects mentioned above, and other effects not mentioned are obvious to those skilled in the art to which the embodiments disclosed from the following description belong It can be understood.

1 is a block diagram showing a configuration of a document area dividing apparatus according to an embodiment.
2 is a diagram illustrating a document image according to an exemplary embodiment.
FIG. 3 is a diagram illustrating a result of applying an MSER algorithm according to an exemplary embodiment, and extracting connection elements and performing clustering.
FIG. 4 is a diagram illustrating a result of separating clusters by performing vertical projection on each of the clusters in a document image according to an exemplary embodiment.
FIG. 5 is a diagram showing a result of separating clusters by performing a horizontal projection on each of the clusters in a document image according to an embodiment.
6 is a diagram illustrating a result of detecting a text line in a document image according to an embodiment.
7 is a diagram showing a result of dividing a text area and a non-text area in a document image and detecting a document boundary according to an embodiment.
8 is a flowchart for explaining a document area dividing method according to an embodiment.

Various embodiments are described in detail below with reference to the accompanying drawings. The embodiments described below may be modified and implemented in various different forms. In order to more clearly describe the features of the embodiments, detailed descriptions of known matters to those skilled in the art are omitted. In the drawings, parts not relating to the description of the embodiments are omitted, and like parts are denoted by similar reference numerals throughout the specification.

Throughout the specification, when a configuration is referred to as being "connected" to another configuration, it includes not only a case of being directly connected, but also a case of being connected with another configuration in between. In addition, when a configuration is referred to as "including ", it means that other configurations may be included, as well as other configurations, as long as there is no specially contradicted description.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing a configuration of a document area dividing apparatus according to an embodiment.

The document area dividing device according to an embodiment may be a device for directly photographing a document and dividing an area of the photographed document image. For example, the document area dividing device may be a terminal, such as a smart phone or a tablet, equipped with an application for performing document area division.

Alternatively, the document area dividing device according to an embodiment may be a device for receiving a document image from another apparatus and dividing an area of the received document image. For example, the document area dividing device may be a desktop or a notebook installed with a program for performing document area dividing.

Alternatively, the document area dividing apparatus according to an exemplary embodiment may be a server that receives a document image from a user's terminal through a network, performs area division, and transmits the document image to the user's terminal.

1, the document area dividing apparatus 100 may include an input / output unit 110, a control unit 120, a storage unit 130, and a communication unit 140. Although not shown in FIG. 1, the document area dividing apparatus 100 may further include a photographing unit.

The input / output unit 110 receives input related to the processing of the document image from the user, and can display a screen showing the progress of the processing of the document image and the result. For example, the input / output unit 110 may include an operation panel for receiving user input and a display panel for displaying a screen.

In particular, the input unit may include devices capable of receiving various types of user input, such as a keyboard, a physical button, a touch screen, a camera or a microphone. Also, the output unit may include a display panel or a speaker. However, the present invention is not limited to this, and the input / output unit 110 may include various input / output support structures.

The control unit 120 controls the overall operation of the document area dividing apparatus 100, and may include a process such as a CPU. According to one embodiment, the control unit 120 may include at least one processor. The control unit 120 may execute the processes of dividing the document area by executing the program stored in the storage unit 130. [

Various types of data such as files, applications, programs, and the like may be installed and stored in the storage unit 130. According to an exemplary embodiment, a program for performing a document area dividing method may be installed in the storage unit 130. [

The communication unit 140 can perform wire / wireless communication with another device or a network. To this end, the communication unit 140 may include a communication module supporting at least one of various wired / wireless communication methods. For example, the communication module may be implemented in the form of a chipset. According to one embodiment, the communication unit 140 can receive a document image from another apparatus.

The wireless communication supported by the communication unit 140 may be Wi-Fi (Wireless Fidelity), Wi-Fi Direct, Bluetooth, UWB (Ultra Wide Band), NFC (Near Field Communication), or the like. The wired communication supported by the communication unit 140 may be, for example, USB or High Definition Multimedia Interface (HDMI). Also, the communication unit 140 may transmit data or messages to a destination via the Internet or a mobile communication network.

Hereinafter, a process of dividing a document area by the controller 120 according to an embodiment will be described in detail.

The control unit 120 can sequentially perform text area detection, non-text area detection, and document boundary detection from a document image.

The controller 120 clusters the connected components included in the document image to detect the text line candidates and performs filtering using the two different threshold values for the detected text line candidates, The text area can be detected by removing the text element.

Then, the control unit 120 can detect the non-text area by removing the detected text line and the background from the document image, and then performing the repeated X-Y cut.

Finally, the control unit 120 can detect the document boundary by comparing the detected non-text region with a predetermined region of interest.

A process of detecting the text area by the control unit 120 will be described in detail.

The control unit 120 extracts connection elements from the document image. For example, the control unit 120 may convert a document image into a binary image using a Retinex binarization method, and extract connection elements from the binarized image. Alternatively, the controller 120 may extract the connection elements by extracting blobs using an MSER (Maximally Stable External Region) algorithm.

The Retinex binarization method can be expressed by the following equation (1).

In this case, B is a binary image and I is an input image. Also, G · I is an image obtained by performing Gaussian filtering on the input image, and k ₁ is a threshold value.

The MSER algorithm is an algorithm for blobly detecting an aggregate region of pixels having different intensities from neighboring neighboring pixels.

The controller 120 performs filtering on the extracted connection elements through the binarization algorithm or the MSER algorithm, thereby removing the connection elements corresponding to the non-text, and extracting only the connection elements likely to correspond to the text. For example, the control unit 120 may connect pixels having foreground values among neighboring pixels in eight directions, and may filter connected pixels in the following manner.

First, if the number of pixels of the connected connection elements is N, the following equation (2) can be obtained. In this case, the covariance matrix of C can be decomposed as shown in Equation (3) below.

는 고유값 및 고유벡터이다.At this time,

Are eigenvalues and eigenvectors.

In the case of a link element corresponding to text, the ratio of the number of pixels is higher and the ratio of the eigenvalues is not larger than the area represented by the product of the eigenvalues. Therefore, by filtering the connection elements satisfying the following equations (4) and (5) using this property, the connection elements corresponding to the non-text can be removed as much as possible.

는 문턱값에 해당된다.At this time,

Corresponds to the threshold value.

The control unit 120 may group the filtered and remaining connection elements, and may detect the text line candidates. First, the control unit 120 may perform clustering of connection elements using the fact that the texts included in the same paragraph have similar properties. For example, the control unit 120 determines whether the color distance, the stroke width ratio, the ratio of the perimeter of the bounding box, the vertical direction, , It is possible to cluster the connection elements among those having similar attributes.

FIG. 2 shows a document image according to an embodiment. FIG. 3 shows an image obtained by applying the MSER algorithm in the document image of FIG. 2, extracting connection elements, and performing clustering. Referring to FIG. 3, it can be seen that the connection elements having similar attributes among the connection elements included in the document image 300 form a single community 301. In this case, the connecting elements of the same color are included in the same cluster. For the sake of convenience of explanation, the boundaries of the cluster 301 are set in a rectangular shape even if some color connecting elements are included.

After the clustering of the connection elements is completed, the control unit 120 may detect the text line candidates by projecting the clusters. For example, the control unit 120 performs projection in at least one of vertical direction and horizontal direction for each of the clusters, divides the respective clusters into small clusters corresponding to a text paragraph and a text line, Can be detected as text line candidates.

In the example document image, the text paragraphs are divided into columns. Accordingly, the controller 120 performs projection in the vertical direction in each of the clusters in the document image 300 of FIG. 3 in which clustering has been completed, and as a result, the control unit 120 obtains a certain multiple of the average stroke widths ) If there is a large gap, it can be seen as a boundary where paragraphs are divided, and can be divided into small clusters with gaps. FIG. 4 shows a result of separating the clusters into small clusters in units of text paragraphs by projecting the clusters in the vertical direction in the document image 300 shown in FIG. Referring to the document image 400 of FIG. 4, it can be seen that the connection elements constituting one cluster 301 in FIG. 3 are divided into two smaller clusters 401 and 402. Since the length d1 of the gap between the two clusters 401 and 402 is equal to or greater than a certain multiple of the average stroke widths of the connecting elements, the controller 120 divides the two connecting elements into different clusters.

The control unit 120 then separates each of the clusters into smaller clusters in units of text lines by performing a horizontal projection on each clusters in the document image 400 shown in Fig. For example, the control unit 120 performs horizontal projection on each of the clusters in the document image 400 of FIG. 4 in which the clusters are divided in units of text paragraphs, and as a result, the average stroke width If there is a gap smaller than a certain multiple (for example, 0.5 times), it can be seen as a boundary on which the text lines are divided and can be divided into small clusters with gaps. FIG. 5 shows a result of separating the clusters into small clusters 501 and 502 in units of text lines by projecting the clusters in the document image 400 shown in FIG. 4 in the horizontal direction. Referring to the document image 500 of FIG. 5, it can be seen that some of the connecting elements that constitute one cluster 402 in FIG. 4 are divided into two smaller clusters 501 and 502. Since the length d2 of the gap between the two assemblies 501 and 502 is smaller than a certain multiple of the average stroke width of the connecting elements, the controller 120 divides the two connecting elements into different clusters.

On the other hand, in the case of the document image shown in FIG. 2, since the text paragraphs are divided into columns, the control unit 120 separates text lines and text lines by performing both vertical and horizontal projections, If the text paragraph is divided into units, it can be divided into small clusters corresponding to the text paragraph and the text line by performing only the projection in the horizontal direction. For example, the control unit 120 performs horizontal projection on each cluster in the clustered document image, and compares the width of the identified gaps with the average stroke width of the connecting element, It is possible to judge whether this paragraph is a boundary dividing line or dividing a line.

When the control unit 120 detects the text line candidates, the control unit 120 performs filtering using the classifier on the text line candidates to remove non-text elements and detect text lines. At this time, the controller 120 can increase the accuracy in removing the non-text elements by using a classifier having two threshold values.

According to one embodiment, the controller 120 may remove non-text elements from the text line candidates using a multi-layer perceptron (MLP) classifier having two thresholds. First, the controller 120 performs filtering using the lower threshold value of the two threshold values, thereby removing the non-text elements while preventing the text lines from being removed as much as possible even if some non-text elements remain. That is, the controller 120 performs first-order filtering using a low threshold value.

The control unit 120 performs clustering on the remaining text line candidates as a result of performing the first-order filtering. In one paragraph, based on the property that the height of the text lines and the spacing between the text lines are constant, the control unit 120 groups the remaining text line candidates into paragraphs.

If there are still text line candidates that are not included in the paragraph even though the clustering has been performed, the controller 120 performs the second-order filtering using the higher threshold value of the two threshold values for the remaining text line candidates .

The control unit 120 may detect the text line candidates remaining after performing the second-order filtering as a text line, and may detect an area including the text line as a text area.

Among the text line candidates extracted from the document image 500 of FIG. 5, the result of the removal of the non-text elements and the detection of the text line is shown in FIG. The document image 600 of FIG. 6 includes only text lines.

When the detection of the text area is completed, the control unit 120 detects the non-text area. A process of detecting the non-text area by the control unit 120 will be described in detail.

The control unit 120 removes the previously detected text area from the binarized image of the document image. At this time, the remaining foreground pixels, i.e., connection elements, are likely to be included in the non-text area. However, if a hardcopy document is photographed with a camera, the document image may include a background other than the document, so that some of the remaining connection elements may correspond to the background. Therefore, in order to remove the connection elements corresponding to the background, the following operations are required.

First, the control unit 120 sets at least a part of the document image as a region of interest in advance, and adjusts the size of the region of interest based on the detected text lines. At this time, the control unit 120 may set an area having a high probability of being a document to be the initial value of the ROI from among the entire document images. The control unit 120 expands the region of interest such that the region of interest includes the text line if some of the detected text lines are outside the predetermined region of interest.

When the size of the ROI is adjusted, the controller 120 determines whether there is a connection element existing over the boundary of the ROI. If there is a connection element existing over the boundary of the ROI, the controller 120 compares the number of pixels existing in the ROI with the number of pixels existing outside the ROI among the pixels included in the ROI do. If the number of pixels existing inside the ROI is larger than the number of pixels existing outside the ROI, the connection element is likely to correspond to the non-text area of the document, not the background. Leave that connection element. However, if the number of pixels existing outside the region of interest is larger than the number of pixels existing in the region of interest, the corresponding connection element is likely to correspond to the background, and therefore the controller 120 removes the connection element.

After removing the connection elements that are likely to correspond to the background in the document image, the controller 120 detects the non-text area by performing a repeated X-Y cut on the remaining connection elements. For example, in the histogram obtained as a result of projecting in the horizontal direction and the vertical direction on the document image in which connection elements highly likely to correspond to the text region and background are removed, the number of consecutive zeros Is equal to or less than a predetermined threshold value, the connection elements are detected as one non-text area. Conversely, if the number of consecutive zeros between the connection elements is greater than the threshold value, the control unit 120 detects the connection elements as separate non-text areas. It is possible to detect the non-text area effectively by appropriately adjusting the size of the threshold value.

When the detection of the non-text area is completed, the control unit 120 can detect the document boundary. The process of detecting the document boundary by the control unit 120 will be described in detail as follows.

The control unit 120 compares the detected non-text area with a predetermined area of interest. If there is a non-text area existing outside the area of interest, the controller 120 expands the area of interest and causes the non-text area to enter the area of interest. If all of the non-text areas included in the document image exist within the ROI, the control unit 120 detects the ROI of the ROI as a document boundary.

An example in which a text region and a non-text region are divided from a document image and a document boundary is detected is shown in Fig. 7, in the document image 700, the text area 701 and the non-text area 702 are divided. In addition, the document boundary 703 was detected to include the text area 701 and the non-text area 702, and to exclude the background.

By using the document area dividing apparatus 100 according to an embodiment, the text lines can be accurately detected even for a document image having a complicated layout by performing clustering after connecting the connection elements included in the document image . In addition, by applying a classifier using two different threshold values, it is possible to improve the accuracy in removing non-text elements. In addition, by detecting the document boundary by comparing the detected non-text region with the region of interest given as the initial value, the accuracy of document boundary detection can be increased.

8 is a flowchart for explaining a document area dividing method according to an embodiment.

The document area dividing method according to the embodiment shown in FIG. 8 includes the steps of time-series processing in the document area dividing apparatus 100 shown in FIG. Therefore, the contents described above with respect to the document area dividing apparatus 100 shown in FIG. 1 can be applied to the document area dividing method according to the embodiment shown in FIG. 8 even if omitted from the following description.

Referring to FIG. 8, in step 801, text line candidates are detected by grouping connection elements included in a document image. Specifically, the connection elements are extracted by applying the binarization algorithm or the MSER algorithm described above to the document image, and the extracted connection elements are clustered. The text line candidates can then be detected by projecting in at least one of the horizontal and vertical directions for each cluster of connecting elements.

At this time, in extracting the connection elements from the document image, filtering is performed on the extracted connection elements by applying a binarization algorithm or MSER algorithm to the document image based on the eigenvalue and the number of pixels, thereby removing non-text elements , Only the remaining connection elements can be extracted.

In this case, when the text line candidates are detected, the cluster is divided into a plurality of detailed clusters according to the result of comparing the width of the margins between the connection elements included in the clusters and the average stroke width of the connecting elements, It can be detected as candidates.

Next, in step 802, filtering is performed using two different threshold values for the detected text line candidates, thereby removing the non-text elements and detecting the text line. Specifically, filtering is performed on the text line candidates using the first threshold value to remove the non-text elements. The remaining text line candidates are grouped into paragraphs, and text line candidates not grouped into paragraphs of the remaining text line candidates The filtering using the second threshold value larger than the first threshold value may be performed to remove the non-text elements, and the remaining text line candidates may be detected as text lines.

In step 803, the non-text area can be detected by removing the detected text line from the document image and then performing the repeated X-Y cut. Specifically, the text line detected in the document image is removed, the region of interest is adjusted based on the detected text line, the connection elements existing over the boundary of the region of interest correspond to the background, The non-text area can be detected by removing the connected elements and performing a repetitive X-Y cut on the remaining connecting elements.

In step 804, the document boundary can be detected by comparing the detected non-text area with the area of interest. Specifically, if the non-text area exists outside the area of interest, the area of interest may be expanded so that the non-text area is included in the area of interest, and the boundary of the expanded area of interest may be determined as the document boundary.

The term " part " used in the above embodiments means a hardware component such as a software or a field programmable gate array (FPGA) or an ASIC, and the 'part' performs certain roles. However, 'part' is not meant to be limited to software or hardware. &Quot; to " may be configured to reside on an addressable storage medium and may be configured to play one or more processors. Thus, by way of example, 'parts' may refer to components such as software components, object-oriented software components, class components and task components, and processes, functions, , Subroutines, segments of program patent code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables.

The functions provided within the components and components may be combined with a smaller number of components and components or separated from additional components and components.

In addition, the components and components may be implemented to play back one or more CPUs in a device or a secure multimedia card.

The document area segmentation method according to the embodiment described with reference to FIG. 8 can also be implemented in the form of a computer-readable medium for storing instructions and data executable by a computer. At this time, the command and data may be stored in the form of program code, and when executed by the processor, a predetermined program module may be generated to perform a predetermined operation. In addition, the computer-readable medium can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. The computer-readable medium can also be a computer storage medium, which can be volatile and non-volatile, implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, Volatile, removable and non-removable media. For example, the computer recording medium may be a magnetic storage medium such as an HDD and an SSD, an optical recording medium such as a CD, a DVD and a Blu-ray Disc, or a memory included in a server accessible via a network.

The document area dividing method according to the embodiment described with reference to FIG. 8 may also be implemented as a computer program (or a computer program product) including instructions executable by a computer. A computer program includes programmable machine instructions that are processed by a processor and can be implemented in a high-level programming language, an object-oriented programming language, an assembly language, or a machine language . The computer program may also be recorded on a computer readable recording medium of a type (e.g., memory, hard disk, magnetic / optical medium or solid-state drive).

Therefore, the document area dividing method according to the embodiment described with reference to FIG. 8 can be implemented by the computer program as described above being executed by the computing device. The computing device may include a processor, a memory, a storage device, a high speed interface connected to the memory and the high speed expansion port, and a low speed interface connecting the low speed bus and the storage device. Each of these components is connected to each other using a variety of buses and can be mounted on a common motherboard or mounted in any other suitable manner.

Where the processor may process instructions within the computing device, such as to display graphical information to provide a graphical user interface (GUI) on an external input, output device, such as a display connected to a high speed interface And commands stored in memory or storage devices. As another example, multiple processors and / or multiple busses may be used with multiple memory and memory types as appropriate. The processor may also be implemented as a chipset comprised of chips comprising multiple independent analog and / or digital processors.

The memory also stores information within the computing device. In one example, the memory may comprise volatile memory units or a collection thereof. In another example, the memory may be comprised of non-volatile memory units or a collection thereof. The memory may also be another type of computer readable medium such as, for example, a magnetic or optical disk.

And the storage device can provide a large amount of storage space to the computing device. The storage device may be a computer readable medium or a configuration including such a medium and may include, for example, devices in a SAN (Storage Area Network) or other configurations, and may be a floppy disk device, a hard disk device, Or a tape device, flash memory, or other similar semiconductor memory device or device array.

It will be apparent to those skilled in the art that the above-described embodiments are for illustrative purposes only and that those skilled in the art will readily understand that other embodiments can be readily modified without departing from the spirit or essential characteristics of the embodiments described above You will understand. It is therefore to be understood that the above-described embodiments are to be considered in all respects only as illustrative and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

It is to be understood that the scope of the present invention is defined by the appended claims rather than the foregoing description and should be construed as including all changes and modifications that come within the meaning and range of equivalency of the claims, .

100: Document area dividing device 110: Input /
120: control unit 130:
140:

Claims (16)

A document area dividing method,
Clustering connection elements included in a document image to detect text line candidates;
Removing the non-text elements and detecting a text line by performing filtering using two different threshold values for the detected text line candidates;
Detecting a non-text area by removing the detected text line from the document image and then performing a repeated X-Y cut; And
Detecting the document boundary by comparing the detected non-text region with a region of interest. The method according to claim 1,
Wherein the step of detecting the text line candidates comprises:
Extracting connection elements by applying a binarization algorithm or an MSER (Maximally Stable External Region) algorithm to the document image;
Clustering the extracted connection elements; And
Detecting the text line candidates by projecting in at least one of a horizontal direction and a vertical direction for each cluster of the connecting element. 3. The method of claim 2,
Wherein the extracting of the coupling elements comprises:
Filtering the extracted connection elements based on the eigenvalues and the number of pixels by applying the binarization algorithm or the MSER algorithm to the document image to remove the non-text elements, and then extracting only the remaining connection elements. How to. 3. The method of claim 2,
Wherein the step of detecting the text line candidates comprises:
Dividing the cluster into a plurality of detailed clusters according to a result of comparing a width of a margin between connection elements included in the cluster and an average stroke of the connection element, and detecting the detailed clusters as text line candidates. How to. The method according to claim 1,
Wherein the step of detecting the text line comprises:
Performing filtering using the first threshold value for the text line candidates to remove non-text elements, and grouping the remaining text line candidates into paragraphs; And
Performing filtering using a second threshold value larger than the first threshold value for text line candidates not grouped into paragraphs of the remaining text line candidates to remove non-text elements, and detecting remaining text line candidates as text lines ≪ / RTI > The method according to claim 1,
Wherein the detecting the non-text area comprises:
Removing a text line detected in the document image;
Adjusting the region of interest based on the detected text line;
Determining whether connection elements existing over a boundary of the ROI correspond to a background;
Removing a connection element determined as a background; And
And performing an iterative XY cut on the remaining connection elements. The method according to claim 1,
Wherein the step of detecting the document boundary comprises:
Expanding the region of interest such that the non-text region is included within the region of interest if the non-text region is also outside the region of interest; And
And determining the boundaries of the extended region of interest as document boundaries. A computer-readable recording medium on which a program for carrying out the method according to claim 1 is recorded. A computer program stored in a medium for performing the method recited in claim 1, which is performed by a document area dividing device. A document area dividing device comprising:
An input / output unit for receiving an input related to processing of a document image and for displaying a status and a result of processing of the document image;
A storage unit for storing a program for performing document area division; And
And a control unit for performing area division of the document image by executing the program,
The control unit detects text line candidates by clustering connection elements included in the document image and performs filtering using the two threshold values different from each other for the detected text line candidates, Detects a text line, removes the detected text line from the document image, and then performs a repetitive XY cut to detect the non-text area, and compares the detected non-text area with the area of interest Detecting a document boundary. 11. The method of claim 10,
Wherein,
A connection element is extracted by applying a binarization algorithm or an MSER (Maximally Stable External Region) algorithm to the document image, and the extracted connection elements are clustered, and in each of the clusters of the connection element, And detecting the text line candidates by projecting the text line candidates in at least one direction. 12. The method of claim 11,
Wherein,
Filtering the extracted connection elements based on the eigenvalues and the number of pixels by applying the binarization algorithm or the MSER algorithm to the document image to remove the non-text elements, and then extracting only the remaining connection elements. . 12. The method of claim 11,
Wherein,
Dividing the cluster into a plurality of detailed clusters according to a result of comparing a width of a margin between connection elements included in the cluster and an average stroke of the connection element, and detecting the detailed clusters as text line candidates. . 11. The method of claim 10,
Wherein,
Filtering the text line candidates by using a first threshold value to remove non-text elements, grouping the remaining text line candidates into paragraphs, and selecting text line candidates that are not clustered as paragraphs of the remaining text line candidates And performing filtering using a second threshold value that is greater than the first threshold value to remove non-text elements, and detects remaining text line candidates as text lines. 11. The method of claim 10,
Wherein,
The text line detected in the document image is removed, the attention area is adjusted based on the detected text line, the connection elements existing over the boundary of the ROI are judged to correspond to the background, Wherein the non-text area is detected by removing a connected element and performing a repetitive X-Y cut on the remaining connecting elements. 11. The method of claim 10,
Wherein,
Expanding the ROI such that the non-text area is included in the ROI if the non-text area exists outside the ROI, and determining the boundary of the ROI as the document boundary. .

Images