JP3086277B2 - Document image processing 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 document image processing apparatus , and more particularly to a document image processing apparatus for identifying underlined character strings in a document image.

[0002]

2. Description of the Related Art As a conventional method for extracting a character string area from a document image, a character string, a graphic, and the like are respectively converted into one by replacing white runs below a threshold value with black runs in the run-length data of the image. A method of extracting a region by a blurring process into a black pixel connected component (Computer
Graphics and Image Process
ssing 20, 375-390, 1982) or a method of performing area division from the peripheral distribution of an image (Transactions of the Institute of Electronics, Information and Communication Engineers, 1986/8 Vol. J69-DNo.
8 pp 1187-1195).

In the above-described extraction method, a document area is divided into a character string area, a graphic area, a photograph area, and vertical and horizontal area dividing lines (separators) according to the size of a circumscribed rectangle of the area, the aspect ratio, the density of black pixels in the area, and the like. However, a figure having the same size as a character string is erroneously identified as a character string.

[0004]

As a method for solving this problem, a method of discriminating a character string from a figure by examining the presence of long line segments in an area extracted by the blurring processing (Japanese Patent Laid-Open No. 61-175875). Publication) and a method of discriminating a character and a figure from the size of a black pixel connected component (IEICE Technical Report PRU86-115, pp33-40). However, even with these methods, there is a problem that the character string is identified as a graphic when the underline is in contact with the character string.

An object of the present invention is to provide a document image processing apparatus for identifying a character string with an underline from a document image.

[0006]

To SUMMARY OF THE INVENTION To achieve the above object, the present invention, the document image processing apparatus for extracting document elements from the document image data, in horizontal string candidate area extracted from the document image data horizontal than the predetermined threshold
Width and determines means <br/> whether there is a large black pixel connected component, if the black pixel connected component is greater than the predetermined threshold value, means for dividing the character string candidate region vertically, Extract the black pixel connected component in the divided upper region
And, if none of the black pixel connected component that issued the extract is less than the predetermined threshold value, the character string candidate region determined to underlined text area, if not, Horizontal
Means for determining the figure or the separator .

[0007]

The document image input by the image input section is stored in the storage section, and the character string candidate area extracting section extracts a character string candidate area which is an element of the document from the image data in the storage section. For the character string candidate region, the determination processing unit checks the size of the black pixel connected component in the region, and when the width of the connected component is larger than the expected character size by a threshold or more, the region dividing unit vertically moves the region up and down. To divide. The determination processing unit compares the connected component in the divided upper region with the previous threshold, and if the connected component is equal to or smaller than the threshold, identifies the character string candidate region as an underlined character string.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings. FIG. 1 is a block diagram of the present invention. 101 is an image input unit such as a scanner that takes in a document image, 102 is a storage unit that stores input image data, 103 is a character string candidate area extraction unit that extracts a character string candidate area from image data, and 104 is An area dividing unit 105 for vertically dividing the character string candidate area, and a determination processing unit 105 for identifying an underlined character string from the character string candidate area.

FIGS. 2 and 3 are flowcharts showing the processing of the present invention. Image input unit 10 such as a scanner
1 is stored in the storage unit 102 (step 201). From the image data stored in the storage unit 102, the character string candidate area extraction unit 103 uses a known method (for example, the above-described blur processing method or the method of dividing the area from the marginal distribution) into a character string candidate area that becomes an element of the document. Is extracted (step 202).

FIG. 4 is a diagram showing an extraction region of the document image 301. Reference numerals 302 to 307 denote the extracted regions.
Numbers 02 to 305 are identified as character string candidate areas based on the size of the area and the density of black pixels. Therefore, in the character string candidate area identified in this way,
There may be cases where figures and separators (vertical and horizontal area dividing lines) are included.

Therefore, in the present invention, for an area identified as a character string, the determination processing unit 105 checks the size of the black pixel connected component in the area (step 204). As a method of obtaining the connected component, for example, a method of performing labeling on a pixel or a method described above (Com
puter Graphics and Image Processing 20,37
5-390, 1982).
djacency Graph) by a known method using a graph structure.

FIG. 5 shows a character string candidate area 40 shown in FIG.
FIG. 3 is a diagram illustrating one of the character strings 1 and represents a character string with an underline 402. FIG. 6 is a diagram showing the black pixel connection region in FIG. 5, and 501 to 506 are circumscribed rectangles of the black pixel connection component. For example, the circumscribed rectangle of the area 401 is 5
01 (the circumscribed rectangle 501 is formed because the character and the underline are close to or connected to each other), and the circumscribed rectangle of the character 403 is 504. FIG. 8 is a diagram illustrating another example in which an area that should be originally identified as a graphic is identified as a character string candidate area 701 based on the density of black pixels.

When the size of the black pixel connected component is obtained, the determination processing unit 105 then determines a threshold value obtained from the size (width) of the black pixel connected component and the expected character size in the area (ie, the height of the area). (Step 205). When the width of the black pixel connected component is larger than the expected character size by a threshold value or more (for example, twice), the region dividing unit 104 divides the region into upper and lower regions (step 206). In the example of FIG. 6, the black pixel connected component (the circumscribed rectangle 501) is larger than the threshold value, and thus is divided by the line 404 in FIG. 5, and in the example of FIG.
Since the black pixel connected component (the circumscribed rectangle 701) is larger than the threshold value, it is divided by the line 702 in FIG. It should be noted that the position of the dividing line is such that the ratio of the divided upper and lower heights is a predetermined value (for example,
(A value at which the up-down ratio becomes 3 to 2).

Then, the determination processing unit 105 obtains a connected component for the black pixel in the divided upper region (step 207). FIG. 7 is a diagram in which connected components of black pixels in an upper region obtained by dividing the region 401 in FIG. 5 by the line 404 are extracted, 612 is a circumscribed rectangle of the region, 613 is a division line, and 601 to 611 are newly added. This is the obtained black pixel connected component of the upper region. Similarly, FIG. 9 is a diagram in which the connected components of the black pixels in the upper region divided by the line 702 in FIG. 8 are extracted, and reference numeral 801 denotes a newly obtained black pixel connected component of the upper region, and 802 denotes a divided line. It is.

The determination processing unit 105 compares the black pixel connected component in the divided upper region with the above threshold (step 208). In the example of FIG. 7, all the black pixel connected components have a size equal to or smaller than the threshold value, that is, a character size, so that the character string candidate area 401 is identified as an underlined character string (step 209). On the other hand, in the example of FIG. 9, since the black pixel connected component 801 is larger than the threshold value, it is identified as a graphic or a separator (step 210).

By performing the above processing for all areas identified as character string candidate areas based on the size of the area and the density of black pixels, the processing for identifying an underlined character string and a figure or a separator is performed. Will be

[0017]

As described above, according to the present invention, in a region identified as a character string candidate region based on the size of the region and the density of black pixels, is there any black pixel connected component larger than the threshold value? In the case where there is a black pixel connected component larger than the threshold value, the region is further divided into upper and lower parts, and it is checked whether or not there is a black pixel connected component larger than the threshold value in the upper region. Therefore, an underlined character string and a horizontally long graphic or separator can be accurately identified.

[Brief description of the drawings]

FIG. 1 is a block diagram of the present invention.

FIG. 2 is a diagram showing a processing flowchart of the present invention.

FIG. 3 is a diagram showing a processing flowchart of the present invention as in FIG. 2;

FIG. 4 is a diagram showing an extraction area of a document image.

FIG. 5 is a diagram illustrating an example of a character string candidate area.

FIG. 6 is a diagram illustrating a black pixel connection region of FIG. 5;

7 is a diagram in which a connected component of black pixels in an upper region obtained by dividing the region in FIG. 5 is extracted.

FIG. 8 is a diagram illustrating another example in which an area that should be originally identified as a graphic has been identified as a character string candidate area.

FIG. 9 is a diagram in which a connected component of black pixels in a divided upper region in FIG. 8 is extracted.

[Explanation of symbols]

 Reference Signs List 101 Image input unit 102 Storage unit 103 Character string candidate area extracting unit 104 Area dividing unit 105 Judgment processing unit 401 Character string candidate area 402 Underline 501 Bounding rectangle of black pixel connected component

Claims (1)

(57) [Claims] A document element is extracted from document image data.
Document image processingapparatus, Extracted from the document image data
WasHorizontal writingIn the character string candidate area,Width
ButDetermine if there is a large black pixel connected componentmeans
And the black pixel connected component isSaidGreater than a given threshold
Means for dividing the character string candidate area into upper and lower parts,
Split top areaExtract the connected components of black pixels within
And the extractedBlack pixel connected componentAny ofIs the predetermined
If the value is smaller than the threshold, the character string candidate area is under
Judge as character string area with lineAnd if not, landscape
Shape or separatorMeans for determining
Document image processing characterized byapparatus.