JP2789835B2 - Table frame detection method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting and extracting a table frame from a document image / form image input from an image input device.

[0002]

2. Description of the Related Art To read a tabular document such as a form, first, an image input device such as a scanner is used.
It is necessary to input a general form document as a form image quantized into binary, and then to extract ruled lines forming a table.

Conventionally, in such a form image, a table structure has been extracted by taking advantage of the fact that ruled lines, which are main components, are regularly arranged and that ruled lines are expressed as straight lines. .

[0004] Such a system is very sensitive to the inclination of the input image, and has a problem that it is difficult to extract the ruled line particularly when the ruled line is drawn by a thin line segment. .

[0005] An object of the present invention is to solve the above-mentioned conventional problems, because the table frame is regularly arranged as a rectangular rectangle, so that the outline inside the table has a shape close to a rectangle. And instead of detecting the linear component,
It is an object of the present invention to provide a table frame extracting method capable of stably and easily extracting a table frame with respect to an inclination by extracting a contour having a shape close to a rectangle.

Another object of the present invention is to reduce the number of pixels that need to be scanned substantially by tracing only the contours of connected pixels, thereby improving the time required for processing. It is to provide a detection method.

[0007]

According to the table frame detecting method of the present invention, when detecting a table frame from a binary quantized image of a document or form, a connected component of black pixels is first detected. Then, the contour of the connected component detected here is obtained. Next, a circumscribed rectangle of the contour is obtained. Here, the circumscribed rectangle is defined as a rectangle whose side is in the horizontal or vertical direction and which is in contact with the outermost side of the outline. FIG. 1 shows specific examples of the outline and the circumscribed rectangle. Next, the distance between each pixel constituting the contour and the circumscribed rectangle is calculated, and the values are integrated. The integration processing is performed on all the pixels constituting the outline, the peripheral length of the outline is compared with the integrated value, and it is determined whether or not the pixel is a table frame. FIG. 2 shows an algorithm for this table frame determination.

A table frame detecting apparatus according to the present invention includes a memory for storing a binary-quantized document image / form image, a connected component extracting section for extracting a connected component of black pixels, and a contour of the extracted connected component. A contour tracing unit for tracing a line, a memory for storing the tracing contour, a circumscribed rectangle detecting unit for obtaining a circumscribed rectangle of the cut connected component, a memory for storing the circumscribed rectangle, A distance detection unit that calculates the distance between the pixel and the side of the circumscribed rectangle, a distance integration unit that integrates the distance, and a comparison selection unit that compares and selects the value of the integrated distance, the perimeter of the contour line, and a predetermined threshold. , A table frame storage unit for storing selected table frame data, and a timing control unit for controlling data output from each memory and timing of each process.

[0009]

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

FIG. 3 shows an outline of a table frame detection method. Hereinafter, the processing method will be described using formulas as appropriate. The connected components of black pixels are separated from the original image, and an outline is obtained for each connected component. Next, a circumscribed rectangle of the connected component is obtained. The circumscribed rectangle is defined by horizontal and vertical line segments and tangent to the upper end, lower end, left end, and right end of the contour line, respectively. Therefore, the coordinates of the circumscribed rectangle can be detected by sequentially tracing the contour and detecting the maximum value and the minimum value of the coordinates of the contour pixel being tracked in the horizontal axis direction and the vertical axis direction, respectively. A known method can be used for the procedure for obtaining the contour.

FIG. 4 illustrates a method for obtaining the distance between one pixel on the contour and the circumscribed rectangle. First, the distance between one pixel on the contour and four sides of the circumscribed rectangle of the contour is determined. In this embodiment, the Euclidean distance is used to determine the distance between the outline and the circumscribed rectangle. As a method of defining this distance, other definitions of the distance can be used. The above procedure is expressed by the following equation. Now, it is assumed that the upper left coordinate of the circumscribed rectangle is (g _x1 , g _y1 ) and the lower right coordinate is (g _x2 , g _y2 ). The coordinates of the contour pixels which are currently examined is assumed to be (r _x, r _y). At this time, the distance to the upper side of the circumscribed rectangle is δ ₁ , and the distance to the lower side is δ
₂ , if the distance to the left side is δ ₃ and the distance to the right side is δ ₄ ,
Each distance is expressed as follows.

[0012]

(Equation 1)

[0013]

Note that, from the relationship between the contour and the circumscribed rectangle, g _x1
≦ r _x ≦ g _x2 , g _y1 ≦ r _y ≦ g _y2 . Of these four distances, the minimum δ _i (i = 1-4) is defined as the distance Δ between the outline pixel and the circumscribed rectangle. That is,

(Equation 2)

[0015]

## EQU1 ##

Next, since the table frames are arranged generally in the row direction and the column direction, the outline of the table frame can be expected to almost touch the containing rectangle. Therefore, if the figure is a table frame, a value obtained by integrating the distance between one pixel on each contour and the circumscribed rectangle can be expected to be sufficiently small. This integrated value is defined as a difference value. In addition, since the size of the table frame is sufficiently larger than the size of the character, it can be expected that the perimeter of the outline is sufficiently large. Therefore, it can be expected that the ratio between the difference value and the contour perimeter is sufficiently large. The table frame is extracted by comparing this ratio with a predetermined threshold value.

This procedure is represented by the following equation. The integrated value of the distance Δ between the contour pixel and the circumscribed rectangle, that is, the difference value Δ _total is

(Equation 3)

[0019]

## EQU1 ## Here, Σ is performed for all pixels constituting the contour.

A known method can be used as a method for obtaining the perimeter of the contour line. For example, the perimeter r ₁ of the contour line indicates that the distance between pixels is one unit in the vertical and horizontal directions,
The oblique direction is defined as √2 units, and can be calculated when the contour is tracked from the positional relationship with the adjacent contour pixels. This can be expressed by the following equation. The perimeter of the contour is r
₁ , the coordinates of two adjacent pixels are represented by r ₁ (r _x1 , _ry1 ), r
₂ (r _x2 , r _y2 )

(Equation 4)

[0022]

The determination of the table frame is performed by determining the ratio r ₁ / Δ _total between the contour line perimeter r ₁ and the difference value Δ _total to a predetermined threshold value α.
Is not exceeded, it is determined that the figure is a table frame.

Next, the second invention will be described. Second
The present invention is an apparatus used in the first invention. In FIG.
FIG. 2 is a functional block diagram illustrating an embodiment of a table frame detection device according to the present invention.

This table frame detecting apparatus comprises a pattern memory 1, a connected component cutout unit 2, a circumscribed rectangle memory 3, a contour tracking unit 4, a contour memory 5, a distance calculation unit 6, a distance integration unit 7, a distance memory 8, It comprises a comparison and selection section 9, a table frame storage section 10, and a timing control section 11.

The pattern memory 1 stores a binary-quantized document image / form image. Connected component extraction unit 2
Analyzes image data sent from the pattern memory 1 in an image format such as a binary image, and cuts out black pixel connected components in the image. The circumscribed rectangle memory 3 stores the coordinate values of the circumscribed rectangle of the connected graphic obtained by the connected component cutout unit 2. The contour tracing section 4 performs contour tracing on the image data of the connected component cut out by the connected component cutout section 2 and outputs contour information. As a method of expressing the outline information, for example, a data format such as a chain code can be used. Contour memory 5
Stores the contour information generated by the contour tracking unit 4. The distance calculator 6 fetches the contour data and the circumscribed rectangle coordinate data from the contour memory 5 and the circumscribed rectangle memory 3, respectively, and calculates the distance. The distance accumulating section 7 uses the distance detecting section 6 while using the distance memory 8 as a working memory.
The distance obtained by is integrated. The comparison determining unit 9 calculates the ratio of the integrated value of the distance obtained by the distance integrating unit 7, that is, the difference value, to the peripheral length of the contour stored in the contour memory 5, and compares the ratio with a predetermined threshold. I do. As a result, when it is determined that the outline is a table frame, the outline information is stored in a table frame storage unit 1 in a predetermined format, for example, a chain code format.
Transfer to 0.

The connected component cutout unit 2 includes a memory for storing labels, a memory for storing coordinates of pixels being scanned, a mechanism for performing pixel scanning in the raster direction, and a mechanism for determining the presence of black pixels. A mechanism for reading the content of the pixel being scanned and the eight pixels adjacent to the pixel, and scanning the same label when there is a black pixel whose label has already been allocated to the eight adjacent pixels. A mechanism for performing a process of allocating to a certain pixel, a mechanism for storing two or more labels in the memory that stores the labels as equal when eight adjacent pixels are allocated, and a mechanism for storing the labels in the memory. This can be realized by using a mechanism for rewriting the same label by referring to the relationship between the stored labels, and a mechanism such as a memory and a sequencer for controlling the operation of the mechanism. The mechanism for determining the presence of a black pixel can be realized by determining whether the corresponding bit of the pixel data is 0 or 1.

The contour tracing unit 4 includes a mechanism for performing pixel scanning in the raster direction, a memory for storing the coordinates of the pixel that has started scanning and the coordinates of the pixel being scanned, and whether the pixel being scanned is a black pixel. If the pixel is determined to be a black pixel, a mechanism for reading the pixel being scanned and the contents of eight pixels adjacent to the pixel being scanned, and a mechanism for reading the read eight adjacent pixels in a predetermined order, for example, starting from the pixel scanned immediately before A mechanism for finding a point on the contour by scanning clockwise and searching for a black pixel starting from a pixel forming an angle of 90 degrees around the pixel, and sequentially scanning contour points found in the search for the point on the contour A mechanism that replaces the coordinates of the pixel in the middle, and by detecting that the coordinates of the pixel being scanned and the coordinates of the pixel that started the scanning match, determines that the contour tracking has ended, and the contour tracking ends. Contour point up to It can be realized by using a mechanism for performing control so as to continue the tracking of the coordinates, or more memory, a mechanism such as a sequencer for controlling the operation of the mechanism.

The distance calculator 6 can be realized by using the following mechanism. First, the mechanism for calculating the distance between one point on the contour and one side of the circumscribed rectangle is an adder, a subtractor, a multiplier,
And a mechanism for finding the square root. By adding to this mechanism a register for storing the result of the mechanism for obtaining the distance, a comparator for comparing the contents of the register, and a selector for selecting data in accordance with the result of the comparator, the distance calculation unit is added. 6 can be realized.

The distance integrating section 7 can be realized by using, for example, an accumulator.

The comparison and selection section 9 can be realized by using a comparator for comparing the magnitude of input data and a selector for selecting data according to the result of the comparator.

Each of the processing units described above is configured with software by directly configuring each memory and register as a memory directly accessed by a microcomputer or the like,
It can also be realized by a microcomputer or the like.

The timing of data transfer from each memory and the timing of execution of each process are intensively controlled by the timing control unit 11 so that a plurality of connected figures extracted from the figure cutout unit 2 are subjected to table frame determination. It can be performed. Hereinafter, a control procedure of the timing control unit 11 will be described with an example.

First, the circumscribed rectangle memory 3, contour line memory 5, and distance memory 8 are initialized (cleared). Next, image data for one screen is transferred from the pattern memory 1 to the connected component cutout unit 2. The connected component cutout unit 2 starts scanning an image and starts cutting out connected components. When a certain connected component is cut out, the connected component cutout unit 2 sends a connected component cutout signal to the timing control unit 11 and interrupts its own processing. The extraction of the next connected component is restarted with the connected component extraction continuation signal sent from the timing control unit as a trigger. The connected component cutout signal is also sent to the circumscribed rectangular memory 3, and the recording timing for the circumscribed rectangular memory 3 is reached. The cut-out graphic data is sent to the contour line tracking unit 4, and the contour line tracking process is performed. When the contour tracing process is completed, a contour tracing signal is sent to the timing controller 11 and the contour memory 5, and the contour information is recorded in the contour memory in synchronization with this signal.

Next, coordinate values of points on the contour line are sequentially sent from the contour line memory 5 to the distance calculating unit 6. The distance is calculated and integrated according to the procedure of the first invention. Contour memory 5
When the transfer of all the outline coordinates is completed, the outline coordinate signal is sent from the outline memory 5 to the comparison / selection unit 9 and the timing control unit 11, and the judgment processing of the cut-out figure is performed. As a result of the determination processing, when the cut-out figure is a table frame, the comparison and selection unit 9 transmits a contour output signal to the contour memory 5 and transfers the corresponding contour information to the table frame storage unit 10.
When the table frame determination processing ends, a determination end signal is sent to the timing control unit 11. When the timing control unit receives the determination end signal, the timing control unit initializes the circumscribed rectangle memory 3, the contour memory 5, and the distance memory 8 to start the next connected component determination process. A connected component cutout continuation signal is transmitted to instruct the restart of the next connected component cutout process. When the connected component cutout continuation signal is sent, the connected component cutout unit 2 marks the cutout connected component as cutout / determined, and restarts the search for the next connected component.

When the entire image in the pattern memory 1 has been searched, the connected component cutout unit 2 sends a connected component cutout end signal and ends all the processing.

[0037]

As described above, the connected black pixel components are regarded as one connected graphic area, and the shape and size of the outline are evaluated, thereby reducing the number of processing steps.
In addition, it is possible to stably extract the table frame even with respect to the inclination of the input image.

[Brief description of the drawings]

FIG. 1 is a diagram for explaining the definition of a contour line and a circumscribed rectangle for a graphic;

FIG. 2 is a diagram illustrating a table frame determination algorithm.

FIG. 3 is a diagram showing an outline of a procedure of the present table frame detection method.

FIG. 4 is a diagram for explaining a method of calculating a difference value between a contour of a cut-out graphic and a circumscribed rectangle of the contour.

FIG. 5 is a functional block diagram showing an embodiment of a table frame detection device according to the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 pattern memory 2 connected component cutout unit 3 circumscribed rectangle memory 4 contour tracking unit 5 contour memory 6 distance calculation unit 7 distance integration unit 8 distance memory 9 comparison selection unit 10 table frame storage unit 11 timing control unit

Claims (2)

(57) [Claims] In a method for detecting a table frame of a document image / form image quantized to binary, a connected component of pixels is regarded as one area, a contour of the connected component is obtained, and a contour of the connected component is determined. The distance between the circumscribed rectangle and the pixel on the contour is accumulated for each pixel on the contour, and the accumulated value is defined as a difference value between the contour and the circumscribed rectangle. A table frame detection method for detecting a table frame by comparing the threshold value with the threshold value. 2. A memory for storing a binary-quantized document image / form image, a connected component extracting unit for extracting a connected component of a black pixel, and a contour tracing for tracing the contour of the cut connected component. Part, a memory for storing the tracked contour, a circumscribed rectangle detector for obtaining a circumscribed rectangle of the cut-out connected component, a memory for storing the circumscribed rectangle, pixels on the contour and sides of the circumscribed rectangle. A distance detecting unit for calculating the distance of the distance, a distance integrating unit for integrating the distance, a comparing and selecting unit for comparing and selecting the value of the integrated distance, the perimeter of the contour line and a predetermined threshold, and the selected table frame data. And a timing control unit for controlling the data output from each memory and the timing of each processing.