JPH10233930A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate only a background accurately by correctly discriminating a character color from image data. SOLUTION: A labelling section 30 separately labels a binary image by a black pixel and a white pixel. A labelling result of a black pixel is outputted to a black label image storage section 40 and a labelling result of a white pixel is outputted to a white label image storage section 50 respectively. A character background discrimination section 60 discriminates whether each of label data of each label image is character data or background data by the minimum rectangular method and the line width discrimination method. A character color discrimination section 80 counts total pixels of a black label image and a white label image and discriminates the black label image or the white label image which has higher pixel number. A character data output section 90 provides an output of character black label image from the black level image storage section 40.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image processing apparatus for removing noise, copy-forgery-inhibited pattern, and the like included in a binary image.

[0002]

2. Description of the Related Art In an optical character reading apparatus or a facsimile, a document or the like is optically read by using a CCD sensor or the like, and an original image image of a document or the like quantized into black and white binary values is obtained. In the case of a document or the like, the original image image includes character image data (character data) and character background image data (background data). In general, background data often includes unnecessary noise that hinders image processing, and processing for removing noise from the background data is often performed for image processing.

[0003] Further, in a document, as seen in a headline of a newspaper or the like, a linear pattern (see FIG. 2) such as a vertical line, a horizontal line, or an oblique line is included in a character background portion, or a small dot pattern (See FIG. 3) may be scattered. In this case, the background data includes pattern image data (pattern data). When the image processing of the original image data including such pattern data is performed, inconveniences such as lowering the efficiency of the image processing occur. For example, when performing character recognition processing in an optical character reading device, it is very difficult to perform character segmentation without removing pattern data. In general, processing to remove background pattern data is performed. Characters are cut out from.

[0004] In a headline of a newspaper or the like, FIG.
As shown in (1), there are many images in which character lines are composed of white images (white characters). If there is a possibility that both a black character image and a white character image are input, it is necessary to determine the color of the character line before removing the background data.

Conventionally, in order to first determine the character color,
Utilizing the property that the background pattern is black when the background pattern is black and black when the background pattern is whitish, the density of black pixels around the heading is calculated. (P55 of Literature 1, "Toshiyuki Sakai," Seminar in Basic Information Science ", Corona Publishing Co., Ltd.)
reference).

Further, as a conventional method of removing the background after character color judgment, a method using a black run length and a method using contraction (degeneration) are known (see P56 of the above-mentioned document 1).
The method using the black run length is, for example, by scanning the original image data in the vertical direction and the horizontal direction, and removing black runs whose black run length is smaller than a predetermined threshold value, thereby removing a background linear pattern or a line pattern thinner than a character line. This is to remove a dot pattern. In the method using contraction, for example, contraction in the vicinity of 4 is repeated in the entire original image until there is no background line pattern, and then expansion processing is performed to restore characters.

[0007]

However, in the conventional character color judgment, when the images of the peripheral portion are actually very similar as shown in FIG. Can not.

Further, as shown in FIG. 3A, actual characters include black characters having a black background pattern and a white border,
As shown in FIG. 3B, there are many white characters with a black outline on a whitish background pattern. In such a case, the conventional method has a problem that the character color is erroneously determined and characters are not accurately extracted.

Further, in the conventional background removal method, any of the above methods cannot accurately remove the background unless the background pattern is thinner than the character line.

The character image with a line pattern shown in FIG.
As shown in FIG. 4A, when a part of the horizontal line of the character is equal in thickness to the background line, and such image data is to be removed using the black run length, the horizontal line of the character disappears as shown in FIG. Resulting in. Similarly, when contraction and expansion are used, FIG.
As shown in (b), the horizontal line of the character disappears. As shown in FIG. 2A, when characters are printed in a typeface similar to the Mincho style, the horizontal lines of the characters become extremely thin. Therefore, there is a problem that the character extraction processing and the character recognition processing at the subsequent stage are hindered.

Therefore, there is a demand for an image processing apparatus capable of correctly determining the character color even in image data in which the character color cannot be specified from the peripheral portion of the image. There is also a need for an image processing apparatus that can accurately remove only the background from image data in which some of the characters are so thin that they cannot be distinguished from the background data.

[0012]

In order to solve the above problems, an image processing apparatus according to the present invention comprises: (1) binary image data including a character is input, and black and white pixels of the binary image data are input; (2) a black label image storing means for storing a black label image linked for each black pixel by the labeling means, and (3) a white label linked for each white pixel by the labeling means. White label image storage means for storing a label image,
(4) character background determination means for determining whether each black label image stored in the black label image storage means and each white label image stored in the white label image storage means is character data or background data, and (5) characters 2 input based on each black label image and each white label image determined as character data by the background determination means.
A character color determining unit that determines a color of a character included in the value image image data.

Here, one of the black label image storage means and the white label image storage means is selected in accordance with the character color determined by the character color determination means, and the selected label image storage means is selected by the character background determination means. It is preferable to further include a character data output unit that outputs the label image determined as character data as character data.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of an image processing apparatus according to the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of this embodiment. As shown in FIG. 1, an image processing apparatus according to the present embodiment includes an image recording medium 10 on which characters, line drawings, and the like are recorded, a binary image image obtained from the image recording medium 10, and a binary image image obtained. An image storage unit 20 that stores (original image data), a labeling unit 30 that labels black and white pixels on each pixel of the original image data stored in the image storage unit 20, and a labeling unit 30. A black label image storage unit 40 for storing the obtained black pixel label image, a white label image storage unit 50 for storing the white pixel label image obtained by the labeling unit 30, a shape and size of the label image And the character / background determining unit 60 for determining whether the label image is character data or background data, and the label numbers of black pixels and white pixels from the labeling unit 30, respectively. A label table writing unit 70 that creates a label table for each white pixel and writes the character background determination result into the label table, and refers to a label table for black pixels and a label table for white pixels created by the label table writing unit 70. A character color determining unit 80 for determining the character color of the original image data, and a character data output unit 90 for receiving the label image and label table of the pixel determined by the character color determining unit 80 and outputting the image data determined to be character data It consists of.

Next, the operation of the image processing apparatus according to the embodiment comprising the above-described components will be described.

The image storage unit 20 includes a complex pattern on a background such as a newspaper headline from the image recording medium 10 storing a binary image such as a document having characters and line drawings. A character image (original image data) whose character or background color is not specified is received and, for example, white is stored as “0” and black is stored as “1”.

The labeling section 30 receives the original image data from the image storage section 20, scans the entire original image data, and performs labeling for black pixels and white pixels.

Here, the labeling process means that a different label is connected to each connected component in order to identify each connected component.
The operation of assigning these elements to pixels is also called labeling. As a method of the labeling processing, for example, four neighborhoods of the target pixel (4
Pixel), a pixel having the same color as the target pixel is selected, the same processing is performed on the selected pixel, and the same process is repeated until there is no pixel of the same color. When there are no more pixels of the same color, labeling of the continuous area ends.

FIGS. 5 and 6 show examples of labeling processing results of black characters having a linear pattern background as shown in FIG.
Shown in FIG. 5 shows the result of labeling using black pixels, and FIG. 6 shows the result of labeling using white pixels. 5 and 6
Respectively, the label of each connected component is represented by one alphanumeric character in order to make the figure easy to see, and does not correspond to the label described in the label table (FIG. 7).

The labeling unit 30 outputs the result of labeling black pixels to the black label image storage unit 40 and the result of labeling white pixels to the white label image storage unit 50. Further, the labeling unit 30 outputs the number of labels of the black pixels and the white pixels to the label table writing unit 70.

The black label image storage unit 40 and the white label image storage unit 50 output the label images to the character background judgment unit 60 and the character color judgment unit 80, respectively, at the subsequent stage. Outputs each label image as needed.

The label table writing section 70 receives the number of black image labels (NB) and the number of white image labels (NW) and creates a label table as shown in FIG. The number NB of labels of a black image indicates the number of closed regions formed of black pixels, and the number NW of labels of a white image indicates the number of closed regions formed of white pixels. In the example of the labeling result of FIG. 5 described above, 57 labeled as 1 to 9, a to z, and A to V are used.
Number of closed areas exist, and the number of labels NB of the black image is 57. In the example of the labeling result in FIG.
There are eight closed regions labeled a, and the number NW of labels of the white image is eight.

The character background determination unit 60 receives the black labeling image from the black label image storage unit 40 and the white labeling image from the white label image storage unit 50, and determines whether each label data of each label image is character data. Determine whether the data is data.

Hereinafter, the determination method by the character background determination unit 60 will be described. Here, FIG. 8 shows a detailed configuration of the character background determination unit 60, and FIG. 9 is a flowchart showing a determination process by the character background determination unit 60.

The character background determination unit 60 includes a rectangle determination unit 62 and a line width determination unit 64, and performs determination for each of the black label image and the white label image (steps 100 to 102).

The rectangle judging section 62 scans each label image, detects the smallest rectangle surrounding the label image in ascending order of the label, and finds the area (LS) of the rectangle when the smallest rectangle is detected. Set small threshold (T
HL1), if expression (1) is satisfied, that is, if the area (LS) of the rectangle is smaller than the threshold value (THL1), it is determined that the label image is background data (step 101).

LS <THL1 (1) Here, by setting a small value to the threshold value THL1 and applying Expression (1), if the label image is a minute point,
Equation (1) is satisfied, and the background of such a minute point is correctly determined as background data. It is desirable that the threshold value THL1 be set to about 1/1000 of the area of the original image data. Thereby, the white pixel having the dot pattern shown in FIG. 3A and the black pixel having the dot pattern shown in FIG. 3B are determined to be the background data.

Next, the rectangle judging section 62 calculates the area (LS) of the above-mentioned rectangle by using a predetermined large threshold (THL2).
If the expression (2) is satisfied, that is, if the area (LS) of the rectangle is larger than the threshold value (THL2), it is determined that the label image is the background data (step 10).
1).

LS> THL2 (2) Here, by setting a large value for the threshold value THL2 and applying the expression (2), if the label image has a size that cannot be considered in the character area, (2) The expression is satisfied, and such a large connected background is correctly determined as background data.
Note that the threshold value THL2 is set to に は of the area of the original image data.
It is desirable to set the degree. As a result, FIG.
FIG. 3 shows a black pixel corresponding to the background of the dot pattern shown in FIG.
It is determined that the white pixel corresponding to the dot pattern shown in (b) is the background data. Further, a white pixel corresponding to the background of the linear pattern illustrated in FIG. 2A and a black pixel corresponding to the background of the linear pattern illustrated in FIG. 2B are determined to be the background data.

The above-mentioned threshold value THL1 indicates the minimum threshold value of character data, and the threshold value THL2 indicates the maximum threshold value of character data.

If the rectangle determining unit 62 determines that the data is background data, a flag (FIG. 7) in the label table of black and white pixels created by the label table writing unit 70 is used.
(See step 1)
04, 105).

If the rectangular determination section 62 does not determine that the data is background data, the line width determination section 64 determines the line width (step 102).

The line width determining section 64 detects the line width of the label image by the following method.

As shown in FIG.
A 2 × 2 observation window is raster-scanned so that an observation window having two rows and two columns is provided, and all the pixels constituting the image data in the memory of the label image storage unit 40 or 50 are viewed through the observation window a0. Then, a pixel comparison between the 2 × 2 observation window and the memory of the label image storage unit 40 or 50 seen from the 2 × 2 observation window is performed, and the label image storage unit 40 seen from the 2 × 2 observation window is compared. Alternatively, when all the pixels on the 50 memories are the same as the label of the label image, the four points Q are sequentially calculated so that the count is increased by one. Further, when the pixel on the memory of the label image storage unit 40 or 50 viewed from the observation window a0 is the same as the label of the label image, the total score A is sequentially counted so as to increase the count by one. However, the initial values of the four points Q and the total points A are 0.

For example, when the black image area indicated by the label f of the black label image in FIG. 5 is enlarged, the result becomes as shown in FIG. The observation window (a0 to a0) shown in FIG.
When scanning a3), there is no pattern in which all of the four windows are the same as the observation window of f (shown in FIG. 11C) and all the pixels are the same. Therefore, the number of four points Q = 0. There are seven patterns with the same observation window a0. Therefore, the total score A = 7
Becomes As another example, a diagram in which the line width is increased by one pixel is shown in FIG. In this example, the observation window and all pixels are the same when the pattern is indicated by dotted hatching in the figure. Therefore, four points Q = 1, and the total points A = 8
Becomes

Next, the line width determining unit 64 calculates the average line width of the connected component by the following equation (3). When the average line width W is calculated, it is compared with a preset threshold value (THL3) as shown in Expression (4). If Expression (4) is satisfied, the label image is determined to be background data.

Average line width W = A / (A−Q) (3) W <THL3 (4) Here, by setting the threshold value THL3 to a value smaller than the character line width, the label image becomes a line. It can be determined whether or not the pattern is a pattern. In the case where the threshold value THL3 is often composed of one line like a newspaper headline, the smaller of the width and height of the original image data is determined to be the size of one character, and 50 minutes It is desirable to set about 1. As a result, the black pixel having the linear pattern shown in FIG. 2A and the white pixel having the linear pattern shown in FIG. 2B are determined to be the background data. In the example of FIG. 11A, the average line width W = 7 /
(7-0) = 1, average line width W = 8 in the example of FIG.
/ (8-1) = 8/7.

In the image shown in FIG. 2, a part of the character is equivalent to the line width of the background. However, in this method, since the average line width is detected for each label image, a part of the character is Don't worry about judging it as data.

The label image which is not determined to be background data in the line width determination unit 64 is determined to be character data, and the label table writing unit 70 stores the label image in the label table of black pixels and white pixels. Flag (“F MOJI ”is set to be character data (steps 103 and 10).
5).

When the above-described determination process for each label image by the character background determination unit 60 is completed, the character color determination unit 8
Shift to the process of 0.

The character color determination unit 80 receives the black label image from the black label image storage unit 40 and the white label image from the white label image storage unit 50, and determines the character color. First, the character color determination unit 80 scans the black label image, and sets the flag in the label table of the black pixel to F MOJI
All the pixels of the black label image with the label of are counted. Similarly, the character color determination unit 80 scans the white label image and sets the flag in the label table of the white pixel to F MOJI
All the pixels of the white label image to which the label is attached are counted. Then, the character color determination unit 80 determines the one with the larger number of counted pixels as the character color.

The reason for this is that the flag in FIG. MOJI
, The number of pixels of the white label, and the flag in FIG.
The number of pixels of the black label that is OJI, and the flag in FIG.
F The number of pixels of the white label that is MOJI and the number of pixels in FIG.
Is F The number of pixels of the MOJI black label is 0
Individual. This is because the result of determination by the character background determination unit 60 is positive.
Because it is.

Here, the reason for determining the color having the larger number of pixels as the character color is to ensure that the character color is correctly determined even if the result of the determination by the character background determination unit 60 is slightly wrong.

The character data output unit 90 includes a character color judgment unit 8
When the character color determination result is received from 0, and the determination result is black, a black label image is received from the black label image storage unit 40 and a label table of the black image is received from the label table writing unit 70. F The MOJI label portion is output as black pixels, and the others are output as white pixels.
Similarly, when the character color determination result from the character color determination unit 80 is white, the character data output unit 90 outputs a white label image from the white label image storage unit 50 and a label table of the white image from the label table writing unit 70. And outputs the portion of the label whose flag in the label table is FMOJI as a black pixel and the others as white pixels. FIG. 12 shows the output result. Input images are shown in FIGS. 2 (a), 2 (b), 3 (a),
In either case of (b), the output result is as shown in FIG.

As described above, according to the above embodiment, it is possible to accurately determine the character color of an image that cannot be determined only from the peripheral portion of the image. As a result, it is possible to obtain an accurate ground tint (background) removal image.

Further, according to the above embodiment, since the line width is detected for each label image, a relatively stable line width can be obtained even if a part of the character is thin. The line pattern of the background and the character line can be easily distinguished, and only the line pattern can be removed without erasing the thin character line portion. Therefore, even when character recognition or the like is performed at a later stage of the image processing apparatus, character cutout processing and character recognition processing can be performed accurately.

In the description of the above-described embodiment, the rectangle determination and the line width determination are used as a method for distinguishing the character data from the background data. However, the present invention is not limited to this. Any method that can distinguish data can be applied to the present invention.

Further, in the description of the above embodiment, the thresholds for determining the rectangle (THL1, THL2) and the threshold for determining the line width (THL)
Although the example of 3) is shown, this is merely an example and the present invention is not limited to this.

Further, the order of the rectangle judgment and the line width judgment is shown in FIG.
The rectangle determination may be performed after the line width determination instead of the order shown in FIG.

In the description of the above embodiment, the four-neighbor connection method is used for the labeling method. However, the present invention is not limited to this, and another method may be used as long as a connected component can be detected.

The input image data to the image processing apparatus of the present invention is not limited to the input from the image recording medium as described in the above embodiment, but may be the input image data from another image processing apparatus such as a scanner or a facsimile. It may be.

Further, the configuration of the above embodiment up to obtaining the character color may be applied to an image processing apparatus for a different purpose from the above embodiment. For example, the configuration of the above embodiment up to obtaining a character color can be applied to an image processing apparatus that converts an input image into a black character image when the input image is a white character.

[0054]

As described above, according to the present invention, (1) labeling means for inputting binary image data including a character and labeling each of black pixels and white pixels of the binary image data; (2) a black label image storage unit for storing a black label image connected for each black pixel by a labeling unit, and (3) a white label image for storing a white label image connected to each white pixel by a labeling unit. Storage means, and (4) a character background determination means for determining whether each black label image stored in the black label image storage means and each white label image stored in the white label image storage means is character data or background data, (5) Character color judgment for judging the color of characters included in the binary image image data input based on each black label image and each white label image judged as character data by the character background judgment means Because it has a stage, it is possible to determine the character color correctly in the image data which can not be identified the text color from the image periphery.

Further, one of the black label image storing means and the white label image storing means is selected according to the character color determined by the character color determining means, and the character image is determined by the character background determining means from the selected label image storing means. When a character data output unit that outputs a label image determined as data as character data is further provided, only the background can be accurately removed even in image data including many portions other than characters such as a copy-forgery-inhibited pattern.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a basic configuration of an image processing apparatus according to an embodiment.

FIG. 2 is an explanatory diagram showing an example of a linear pattern.

FIG. 3 is an explanatory diagram showing an example of a dot pattern.

FIG. 4 is an explanatory diagram showing an example of the related art.

FIG. 5 is an explanatory diagram illustrating a black label image of a black character according to the embodiment.

FIG. 6 is an explanatory diagram illustrating a white label image of white characters according to the embodiment.

FIG. 7 is an explanatory diagram illustrating a label table according to the embodiment;

FIG. 8 is a block diagram illustrating a detailed configuration of a character background determination unit 60 according to the embodiment.

FIG. 9 is a flowchart illustrating a character background determination process according to the embodiment.

FIG. 10 is an explanatory diagram illustrating an observation window for detecting a line width according to the embodiment;

FIG. 11 is an explanatory diagram illustrating a line width determination operation according to the embodiment.

FIG. 12 is an explanatory diagram illustrating a copy-forgery-inhibited pattern removal result by the image processing apparatus according to the embodiment;

[Explanation of symbols]

20 image storage unit, 30 labeling unit, 40 black label image storage unit, 50 white label image storage unit, 60 character background determination unit, 62 rectangular determination unit, 64 line width determination unit, 7
0: label table writing unit, 80: character color judgment unit,
90 ... Character data output unit.

Claims (7)

[Claims] 1. Binary image data including a character is inputted, and labeling means for labeling each of black pixels and white pixels of the binary image image data, and the labeling means are connected for each black pixel. A black label image storage unit for storing a black label image, a white label image storage unit for storing a white label image connected for each white pixel by the labeling unit, and each black label stored in the black label image storage unit A character / background determining means for determining whether the image and each white label image stored in the white label image storing means are character data or background data; and each of the black label images and each of the black label images determined to be character data by the character / background determining means A character color determining means for determining a color of a character included in the binary image data input based on the white label image; The image processing apparatus according to claim and. 2. The character color determining means counts the number of pixels of each black label image and each white label image determined as character data by the character background determining means for each black label image and white label image. The image processing apparatus according to claim 1, wherein one having a larger number of pixels is determined to be a character color. 3. The character background determination means detects a minimum rectangle surrounding each label image, and when the area of the detected minimum rectangle is smaller than a predetermined minimum threshold, and The image processing apparatus according to claim 1, wherein the label image is determined to be background data if the label image is larger than the background image. 4. The character background determination means detects a line width of each label image, and determines that the label image is background data when the detected line width is smaller than a predetermined line width threshold. The image processing apparatus according to claim 1, wherein: 5. A method according to claim 1, wherein one of said black label image storage means and said white label image storage means is selected in accordance with the character color determined by said character color determination means, and said character background is selected from said selected label image storage means. 5. The image processing apparatus according to claim 1, further comprising a character data output unit that outputs a label image determined as character data by the determination unit as character data. 6. The character background determining means has a label table for storing a determination result of character data or background data for each label image, and the character data output means is recorded as characters in the label table. 6. The image processing apparatus according to claim 5, wherein each of the label images is read out from the black label image storage unit or the white label image storage unit based on the determination result of the character color determination unit. 7. The image according to claim 6, wherein said character data output means outputs each label image recorded as characters in said label table as black pixels, and outputs the others as white pixels. Processing equipment.