JPH064704A - Ruled line discriminating method and area discriminating method - Google Patents

Abstract

PURPOSE:To exactly discriminate and extract a horizontal or vertical ruled line by going through stages of selection of ruled line rectangle candidates, extraction of a ruled line rectangle from ruled line rectangle candidates, and condition discrimination of the ruled line rectangle. CONSTITUTION:Circumscribed rectangles of black linking components are extracted from an inputted document picture by a rectangle extracting part 102. A size W in the horizontal direction and a size H in the vertical direction of each of these extracted rectangles are compared with respective thresholds RLHTB and RLVTH by a rectangle classifying part 106 to discriminate the rectangles satisfying conditions W>RLHTH and H>RLVTH as horizontal ruled line rectangle candidates. Pictures in ranges of these discriminated rectangles are scanned in the horizontal direction to extract black runs longer than a threshold by a horizontal ruled line extracting part 108, and rectangles circumscribed to linking components of these black runs are extracted as ruled line rectangles. The rectangles satisfying a prescribed condition out of these rectangles is finally discriminated as a horizontal ruled line by a horizontal ruled line approving part 112.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to extraction of ruled lines and table areas in a document image.

[0002]

2. Description of the Related Art In the case of filing general document image data in which charts and characters are mixed in a recording medium or transmitting by G4 facsimile, horizontal ruled lines in the document image,
It is convenient to be able to automatically identify vertical ruled lines, a frame area surrounded by ruled lines, a table region defined by ruled lines, etc. However, the optimum technique for such a purpose has not been established yet.

As a related technique known so far, a black connected rectangle is extracted from a binarized image by scanning an original and comparing its size with a threshold value.
An image extraction method for discriminating a character rectangle from a line figure rectangle (Japanese Patent Laid-Open No. 55-162177), a black run is integrated into a rectangle, and a run of a predetermined length or more inside a rectangle of a predetermined size or larger is integrated. Then, there is a table area identification method (Japanese Patent Application No. 1-143456) for extracting horizontal ruled lines and determining whether or not the rectangle is a table area based on the number of horizontal ruled lines.

[0004]

The image extraction method is
The line figure is not specifically identified as a horizontal ruled line, a vertical ruled line, a table, a surrounding frame, or the like. In the table area identification method, only the horizontal ruled lines are extracted, and the table areas are determined based on the number of horizontal ruled lines. Therefore, the table area identification accuracy may be insufficient. I can't.

Therefore, an object of the present invention is to provide a technique for identifying and extracting a horizontal ruled line, a vertical ruled line, a table area, a surrounding frame area, and other areas with higher accuracy.

[0006]

According to the method of the first or second aspect of the present invention, based on the horizontal and vertical sizes of the circumscribed rectangle of the black connected component extracted from the document image or its reduced image. To select horizontal ruled line rectangle candidates or vertical ruled line rectangle candidates. By scanning the document image or its reduced image within the range of each rectangle candidate in the horizontal direction for horizontal ruled line identification and in the vertical direction for vertical ruled line identification, a length equal to or greater than a certain threshold value is scanned. Extract a horizontal ruled line rectangle or a vertical ruled line rectangle consisting of black runs. Then, based on the relationship between each direction ruled line rectangle candidate and the size of each direction ruled line rectangle extracted in the horizontal direction and the vertical direction (the invention of claim 1) or based on the positional relationship between the horizontal direction and the vertical direction. (Invention of Claim 2) It is determined whether the ruled line rectangle is a horizontal or vertical ruled line.

According to the third or fourth aspect of the invention, the table area rectangle candidate is selected from the circumscribed rectangles of the black connected components extracted from the document image or the reduced image thereof based on the horizontal and vertical sizes. By scanning the document image or its reduced image in the horizontal direction for each table area rectangle candidate range, a horizontal ruled line rectangle consisting of black runs having a length equal to or longer than a certain threshold is extracted, and a horizontal ruled line rectangle is horizontally extracted. Horizontal ruled lines are extracted by judging the ruled line conditions, and the document image or its reduced image is scanned in the vertical direction, and vertical ruled lines consisting of black runs having a length equal to or greater than a certain threshold are extracted, and the vertical ruled line rectangle is formed. On the other hand, the vertical ruled line is extracted by performing the condition determination of the vertical ruled line. Then, according to the invention of claim 3, it is determined whether or not the table area rectangular candidate is a table area based on a determination condition that includes at least the number of horizontal ruled lines and the number of vertical ruled lines extracted from the table area rectangular candidate. To do. According to the invention of claim 4, the judgment condition includes the presence or absence of a horizontal ruled line or a vertical ruled line in a specific range of the table area rectangular candidate, and the table area rectangular candidate is a table area, a surrounding frame area, or other depending on the judgment condition. It is determined which of the areas.

[0008]

According to the method of the present invention, the horizontal and vertical ruled lines are accurately identified by the steps of selecting ruled line rectangle candidates, extracting ruled line rectangles from the ruled line rectangle candidates, and judging the ruled line rectangle conditions. , Can be extracted.

The area identification method according to the invention of claim 3 or 4 is to select a table area rectangle candidate, extract horizontal and vertical ruled lines from the table area rectangle candidate, and determine the number of horizontal and vertical ruled lines in the table area rectangle candidate. Determination of a condition including (invention of claim 3),
Alternatively, by performing a step of judging the condition including the presence or absence of horizontal or vertical ruled lines in the specific range of the table area rectangle candidate in addition to the number of horizontal and vertical ruled lines in the table area rectangle candidate (the invention of claim 4), The area can be accurately identified, and according to the invention of claim 4, the surrounding frame area other than the table area and the other area can be identified.

[0010]

EXAMPLE 1 In this example, as shown in FIG. 1, a document image memory 10 for storing document image data input by a scanner or the like.
0, a rectangle extraction unit 102 for extracting a circumscribed rectangle of a black connected component of the document image data in the memory 100, and a rectangle memory 10 for recording information of the rectangle extracted by the rectangle extraction unit 102.
4. Based on the rectangular information in the rectangular memory, the rectangular classifying unit 106 that classifies the rectangular type by comparing the width and height of the extracted rectangular with a predetermined threshold value. Information on a rectangle determined to be a horizontal separator (horizontal ruled line) (stored in the rectangle memory 104) is referred to, and an image of this rectangle (document image memory 100
The horizontal ruled line extraction unit 10 generates a rectangle from a black run having a length equal to or larger than a threshold value given in advance by performing a horizontal scan on
8. A horizontal ruled line rectangle memory 110 for storing the information of the rectangles extracted thereby, and a horizontal determination for determining whether or not the rectangle is a horizontal separator (horizontal ruled line) based on the information of the rectangles stored in the memory 110. Ruled line inspection section 11
2. It has a functional configuration including a control unit 114 that controls each processing unit 102, 106, 108, 112.

An image reducing unit may be provided in front of the rectangle extracting unit 102 to perform the following processes of rectangle extraction on an image obtained by reducing the input document image. Further, each processing unit including the image reducing unit is realized as hardware or software on a computer system.

The flow of the horizontal parameter (horizontal ruled line) determination processing in this embodiment is shown in FIG. Less than,
The details of this process will be described, but FIG. 3 is a process explanatory diagram that is appropriately referred to in the description.

First, the rectangle extraction unit 102 extracts a circumscribed rectangle of a black connected component from the input document image (step 200). This process can be performed by, for example, repeating an operation of extracting a circumscribed rectangle of a connected black run while scanning an image and integrating the extracted rectangles within a certain distance.

For the extracted rectangle, the rectangle classification unit 1
In 06, the size W in the horizontal direction and the size H in the vertical direction are compared with the respective threshold values RLHTH, RLVTH, and the judgment is made. Barrel rectangle (1)
(Steps 210 and 215).

In FIGS. 3 (a) and 3 (b), reference numeral 300 indicates an example of a rectangle (1), and the shaded portion inside thereof is a black connected component such as characters and ruled lines. The rectangle extracting unit 102 extracts a rectangle by integrating the black connected components such as characters and ruled lines, but since such a process is known, detailed description thereof will be omitted.

In this embodiment, the threshold values RLHTH and RL are set.
Although VTH is fixed, it may be adaptively and automatically set. As an example, a histogram of the vertical size (in the case of a horizontal writing document) of the black connected component that can be regarded as a character in the rectangle extraction processing 200 is created, and the standard character size is determined based on this histogram. A threshold R is obtained by multiplying this standard character size by an appropriate coefficient.
Calculate LHTH and RLVTH. However, this is just an example.

Next, the horizontal ruled line extraction unit 108 horizontally scans the image in the range of the rectangle (1) to extract a black run having a length equal to or longer than the threshold value RUNHTH, and a connected component of only this long black run. A rectangle (rectangle (2)) circumscribing to is extracted as a ruled line rectangle (step 21).
5). In FIG. 3B, 302 is a rectangle (1) 300.
It is a rectangle (2) generated only from the long black run extracted from, and H1 and W1 are the sizes in the vertical direction and the horizontal direction. The threshold value RUNHTH is given in advance as a fixed value, but may be adaptively and automatically set similarly to the above RLHTH and RLVTH.

With respect to such a rectangle (2), the horizontal ruled line verification unit 112 makes a condition determination of W1 / W> threshold value (for example 0.8) and H1>H> threshold value (for example 0.8) (step). 220, 225), the rectangle (2) is finally determined to be a horizontal ruled line when this condition is satisfied (step 230). The rectangle (2) 302 shown in FIG. 3B is finally determined as a horizontal ruled line.

The control unit 114 determines whether to end the process for the rectangle extracted by the rectangle extraction unit 200 (step 23).
5) is performed, and if there are unprocessed rectangles, the processing from step 205 onward is restarted.

Embodiment 2 This embodiment has the functional configuration shown in FIG. 1 as in Embodiment 1. The processing content is the same as that of the first embodiment except the processing of the horizontal ruled line verification unit 112.

That is, in the first embodiment, the horizontal ruled line inspection unit 112 performs the horizontal ruled line inspection based on the ratio of the sizes of the rectangle (1) and the rectangle (2) in the horizontal and vertical directions (see FIG. 2). Steps 220-230). On the other hand, in the present embodiment, as shown in FIG. 3C, the horizontal position differences ΔW1 and ΔW2 between the start and end points of the rectangle (1) 300 and the rectangle (2) 302, and the vertical position difference. ΔH1,
When ΔH2 is ΔW1 <threshold, ΔW2 <threshold, Δh1 <threshold and ΔH2 <threshold, the rectangle (2) 302 is finally determined as a horizontal ruled line.

Embodiment 3 This embodiment has a functional structure as shown in FIG. This embodiment is different from the first embodiment in the following points due to the relationship of extracting vertical ruled lines.

In the rectangle classifying unit 106, the size of the rectangle in the horizontal direction is W and the size in the vertical direction is H, and W <RL
When the condition of HTH and H> RLVTH is satisfied,
The rectangle is classified into a rectangle (1) which is a vertical ruled line rectangle candidate. Since the vertical ruled lines are targeted, the thresholds RLHTH and RLVT are different from those of the first and second embodiments targeted for the horizontal ruled lines.
The magnitude relationship of H is reversed.

In the vertical ruled line extraction unit 408, the document image in the range of rectangle (1) is vertically scanned to extract black runs, and from the connected components of only black runs having a length equal to or greater than a predetermined threshold RUNHTH. Rectangle (2) is extracted and the information is stored in the vertical ruled line rectangle memory 410.

The vertical ruled line verification unit 412 is arranged between the horizontal size W1 and the vertical size H1 of the rectangle (2) and the horizontal size W and the vertical size H of the rectangle (1). Then, when the conditions of W1 / W> threshold value (for example 0.8) and H1 / H> threshold value (for example 0.8) are satisfied, the rectangle (2) is finally determined as a vertical ruled line.

Also in this embodiment, the vertical ruled line extraction may be performed with the image obtained by reducing the input document image as the processing target.

Fourth Embodiment This embodiment has the functional configuration shown in FIG. 4 similarly to the third embodiment. The processing content is the same as that of the third embodiment except for the processing of the vertical ruled line inspection unit 412.

That is, in the present embodiment, the vertical ruled line inspection unit 412 has the horizontal position differences ΔW1 and ΔW2 between the start and end points of the rectangle (1) and the rectangle (2) and the vertical position differences ΔH1 and ΔH2. However, when ΔW1 <threshold, ΔW2 <threshold, Δh1 <threshold and ΔH2 <threshold, the rectangle (2) is finally determined as a vertical ruled line.

In the present embodiment as well, the processing can be performed on the reduced image of the input document image.

Embodiment 5 FIG. 5 is a functional block diagram of this embodiment. However,
Alternatively, the equivalent parts in FIG. 4 are designated by the same reference numerals. 5
Reference numeral 06 denotes a rectangular classification unit, which has different classification conditions from the rectangular classification unit 106 shown in FIG. 1 or 4. Reference numeral 514 is an added area determination unit. FIG. 6 is a process flowchart of this embodiment, and FIG. 7 is a diagram for explaining the process. The processing contents will be described below.

The rectangle extraction unit 102 extracts the circumscribed rectangle of the black connected component from the document image, and stores the rectangle information in the rectangle memory 104 (step 600). It is also possible to perform the reduction processing of the input document image and execute the processing after the rectangle extraction for the reduction processing.

In the rectangle classification unit 506, the horizontal size W and the vertical size H of the rectangle are compared with the horizontal threshold LARGEHTH and the vertical threshold LARGEVTH, and W> LARGEH and H>. A rectangle satisfying the condition of LARGEVTH is determined to be a rectangle (1) which is a table area rectangle candidate (steps 605 and 610).

Since the purpose here is to identify a table area consisting of horizontal ruled lines and vertical ruled lines and having a certain size or more, the threshold values LARGEHTH and LA are set.
RGEVTH is predetermined in consideration of the minimum size of the table area to be identified. However, these thresholds may be adaptively automatically determined based on, for example, a histogram of the height of a rectangle.

Next, the horizontal ruled line extraction unit 108 and the horizontal ruled line verification unit 112 extract horizontal ruled lines from the rectangle (1) (step 615). Similar to the first embodiment, the horizontal ruled line extraction unit 108 horizontally scans an image within the rectangle (1), which is a candidate for a table area rectangle, and is a horizontal ruled line rectangle generated from only black runs having a length equal to or greater than a predetermined value. Rectangle (2) is extracted. In the horizontal ruled line verification unit 112, regarding the horizontal size W1 of the rectangle (2), the vertical size H1, and the horizontal size W of the rectangle (1), W1 / W> threshold value (for example, 0.8) And H1 <RLH
When the condition eightTH is satisfied, the rectangle (2) is extracted as a horizontal ruled line. The threshold value RLHeight may be a fixed value or a variable value adaptively determined.

Further, the vertical ruled line extracting section 408 and the vertical ruled line checking section 412 extract vertical ruled lines from the rectangle (1) which is a candidate for the table area rectangle (step 620). Similarly to the third embodiment, the vertical ruled line extraction unit 408 vertically scans the image in the range of the rectangle (1) to obtain a rectangle (2) which is a vertical ruled line rectangle generated from only black runs having a length equal to or larger than a predetermined value. Extract and
With respect to the horizontal size W1 of the rectangle (2), the vertical size H1, and the vertical size H of the rectangle (1), the vertical ruled line verification unit 412 sets H1 / H> threshold (for example, 0.8). And W1 <RLW
When the condition of idthTH is satisfied, the rectangle (2) is extracted as a vertical ruled line. The threshold value RLwidthTH may be a fixed value or a variable value adaptively determined.

Next, in the area determination section 517, the rectangle (1) which is the table area rectangle candidate classified by the rectangle classification section 506.
For each time, the number of horizontal ruled lines and vertical ruled lines extracted therefrom and the threshold value are compared (steps 625 and 630). Here, assuming that 3 is used as the threshold value, the rectangle (1) in which both three or more horizontal ruled lines and three or more vertical ruled lines are extracted is determined as the table area (step 635). In addition,
The position coordinates of the extracted ruled lines are output from the direction ruled line verification units 112 and 412, and the area determination unit 517 compares each position coordinate with the information of the rectangle (1) stored in the rectangular memory 104 to determine each ruled line. Recognize which rectangle (1) it belongs to.

In FIG. 7, reference numeral 700 is an example of a rectangle (1) which is a table area rectangle candidate. In this example, rectangle (1) 70
Reference numeral 0 is a table area composed of three horizontal ruled lines and three vertical ruled lines. The horizontal ruled lines are extracted as rectangles (2) 702a to 702c, and the vertical ruled lines are extracted as rectangles (2) 704a to 704c. Since they are finally determined to be horizontal ruled lines or vertical ruled lines, the rectangle (1) 700 is The condition of the number of each ruled line is satisfied and the table area is determined.

Sixth Embodiment This embodiment has the same functional configuration as the fifth embodiment. The processing content of this embodiment is the same as that of the fifth embodiment, except that the determination condition of the area determination unit 517 (corresponding to steps 625 and 630 in FIG. 6) is different from that of the fifth embodiment.

Regarding the processing of the area determination unit 517, FIG.
This will be described with reference to (b). The uppermost horizontal ruled line (70 in a range RangeUTH from the upper side of the rectangle (1) 700)
2a) is upper, and the lowermost horizontal ruled line (702) in a range RangeLTH from the lower side of the rectangle (1) 700.
Let c) be lower. Such upper and lo
If there are wers and there are three or more horizontal ruled lines and vertical ruled lines as in the fifth embodiment, the rectangle (1) is determined as the table area.

Seventh Embodiment This embodiment has the same functional configuration as the fifth embodiment. The processing content of this embodiment is the same as that of the fifth embodiment, except that the determination condition of the area determination unit 517 is different from that of the fifth embodiment.

The processing of the area determination unit 517 will be described with reference to FIG. Similar to the case of the sixth embodiment, upper and lower are calculated. Further, the vertical ruled line (704b) in the middle range excluding the range of RangeLFTHk from the left side of the rectangle (1) 700 and the range of Range RTH from the right side of the rectangle (1) is set as a middle.

There are three or more horizontal ruled lines, and
If upper and lower (see Example 6) exist and one or more middles exist, the rectangle (1) is determined as the table area.

Eighth Embodiment This embodiment has the same functional configuration as the fifth embodiment. The processing content of this embodiment is the same as that of the fifth embodiment, except that the determination condition of the area determination unit 517 is different from that of the fifth embodiment.

The processing of the table area determination unit 517 will be described with reference to FIG. Similar to the case of the sixth embodiment, upper and lo
ask for a wer. Further, the leftmost vertical ruled line (7) in the range of RangeLFTH from the left side of the rectangle (1) 700.
04a) to the left, and Range from the right side of the rectangle (1)
R the rightmost vertical ruled line (704c) in the RTH range
A vertical ruled line (704b) in an intermediate range other than the light and those ranges is obtained as a middle.

Then, in the region (1), the fifth embodiment
If any of the table area determination conditions 6 and 7 is satisfied, the area (1) is determined to be the table area. If not, upper, lower, left, ri
When one or more ghts exist, the area (1) is determined to be an enclosing area, but upper, lower, left,
If none of the light is present, the area (1) is determined as the other area (drawing, photograph, etc.) that is neither the table area nor the surrounding frame area.

[0046]

As is clear from the above detailed description, according to the invention of claim 1 or 2, the effect that the horizontal or vertical ruled line can be accurately identified can be obtained, and according to the invention of claim 3, the table area is obtained. The effect of being able to accurately identify
Further, according to the invention of claim 4, it is possible to obtain an effect that the table area can be accurately identified and the surrounding frame area and other areas can be identified.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a first embodiment and a second embodiment.

FIG. 2 is a processing flowchart of the first embodiment.

FIG. 3A is a diagram showing an example of extraction of horizontal ruled line rectangle candidates. FIG. 7B is an explanatory diagram of horizontal ruled line determination in the first embodiment. (C) It is explanatory drawing of the horizontal ruled line determination in Example 2.

FIG. 4 is a functional block diagram of third and fourth embodiments.

5 is a functional block diagram of Examples 5, 6, 7 and 8. FIG.

FIG. 6 is a processing flowchart of the fifth embodiment.

FIG. 7A is a diagram showing an example of a table area rectangle candidate. (B) It is explanatory drawing regarding a horizontal ruled line. (C) It is explanatory drawing regarding a vertical ruled line.

[Explanation of symbols]

 100 Document Image Memory 102 Rectangle Extraction Unit 104 Rectangle Memory 106 Rectangle Classification Unit 108 Horizontal Ruled Line Extraction Unit 110 Horizontal Ruled Line Rectangle Memory 112 Horizontal Ruled Line Verification Unit 114 Control Unit 300 Horizontal Ruled Line Rectangle Candidate (Rectangle (1)) 302 Horizontal Ruled Line Rectangle (Rectangle) (2)) 408 vertical ruled line extraction unit 410 vertical ruled line rectangle memory 412 vertical ruled line inspection unit 506 rectangular classification unit 517 region determination unit 700 table region rectangle candidate (rectangle (1)) 702a to 702c horizontal ruled line rectangle (rectangle (2)) 704a to 704c Vertical ruled line rectangle (rectangle (2))

Claims (4)

[Claims] 1. (a) A circumscribed rectangle of a black connected component is extracted from a document image or a reduced image thereof, and (b) based on the size of the rectangle extracted in (a) above in the horizontal and vertical directions. Select a ruled line rectangle candidate in a specific direction, either horizontal or vertical, and (c) scan the document image or its reduced image in the specified direction for the range of ruled line rectangle candidates selected in (a) above, and set a certain threshold. A ruled line rectangle is generated from the black run having the above length, and (d) the ruled line rectangle candidate selected in (a) above and the size of the ruled line rectangle obtained in (c) above in the horizontal and vertical directions. A ruled line identification method for determining whether or not the ruled line rectangle is a ruled line in a specific direction based on the relationship of height. 2. (a) A circumscribed rectangle of a black connected component is extracted from the document image or its reduced image, and (b) based on the size of the rectangle extracted in (a) above in the horizontal and vertical directions. Select a ruled line rectangle candidate in a specific direction, either horizontal or vertical, and (c) scan the document image or its reduced image in the specified direction for the range of ruled line rectangle candidates selected in (a) above, and set a certain threshold. A ruled line rectangle is generated from the black run having the above length, and (d) the ruled line rectangle candidate selected in (a) above and the position of the ruled line rectangle obtained in (c) above in the horizontal and vertical directions. A ruled line identification method for determining whether or not the ruled line rectangle is a ruled line in a specific direction based on the relationship. 3. (a) A circumscribed rectangle of a black connected component is extracted from a document image or a reduced image thereof, and (b) based on the size in the horizontal and vertical directions of the rectangle extracted in (a) above. , A table area rectangle candidate is selected, and (c) a document image or a reduced image thereof is horizontally scanned within the range of the table area rectangle candidate selected in (a) above to obtain a black run having a length equal to or greater than a certain threshold. The horizontal ruled line rectangle is extracted, and the horizontal ruled line is extracted by performing the condition determination of the horizontal ruled line with respect to the horizontal ruled line rectangle, and (d) the range of candidate table area rectangles selected in (a) above is documented. The image or its reduced image is scanned in the vertical direction, a vertical ruled line consisting of black runs having a length equal to or greater than a certain threshold is extracted, and the vertical ruled line is extracted by performing a condition determination of the vertical ruled line on the vertical ruled line rectangle. , (E) by the judgment condition including at least the number of horizontal ruled lines extracted in (c) above and the number of vertical ruled lines extracted in (d) above from the table area rectangle candidate selected in (b) above, An area identification method for determining whether a table area rectangle candidate is a table area. 4. The area identification method according to claim 3,
The determination condition of (c) above also includes the presence or absence of a horizontal ruled line or a vertical ruled line in a specific range of the table area rectangle candidate. Depending on the determination condition, the table area rectangle candidate is selected as a table area, a surrounding frame area, or another area. An area identification method characterized by determining whether there is an area.