JP2003196592A - Program for processing image, and image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To erase a preprint and to generate a binary image where a entered character is left, so as to realize highly precise character recognition, even when the entered characters come into contact with the preprint to overlap each other. <P>SOLUTION: This image processor is provided with a binalized image generating means 105 for binalizing the minimum value of primary color components based on a color image within a field, and a means 102 for generating a binalized image in which a preprint color is dropped out from the color image within the field, a character reader 107 is made to read the binalized image from the binalized image generating means for binalizing the minimum value of the primary color components, and made to try again to read out the binalized image from the means for generating the binalized image in which the preprint color is dropped out from the color image when a result read hereinbefore by the character reader 107 is rejected. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing program and an image processing apparatus, and more particularly, to an optical character reading apparatus which optically reads a surface image of a form and recognizes a predetermined area of the read image ( The present invention relates to an image processing program and an image processing apparatus that are suitable for being read by an OCR.

[0002]

2. Description of the Related Art Conventionally, a written character written by handwriting on a form or the like has been optically read, recognized and processed by an OCR. Generally, a form is printed with characters (referred to as pre-printing) that are not necessary for reading a frame to be filled, a yen in a unit of money, or the like. Preprinting of these frames and print characters is not necessary for reading, and a color determined to be white by OCR, that is, a predetermined color called a dropout color has been used. The dropout color is fixedly and strictly determined according to the wavelength of the light source or the optical filter used in the photoelectric conversion unit. For this reason, the colors that can be used for preprinting the form are limited, and preprinting can be performed on a form printed in a color other than the specified color, such as a color form that has not been conventionally input by OCR. It was difficult to drop.

As a conventional technique capable of dropping preprinting on a form printed in a color other than a designated color, for example, a technique described in Japanese Patent Laid-Open No. 6-176194 is known. This prior art is provided with a photoelectric conversion unit capable of outputting a color image, and based on the image of the reference form, a portion to be excluded from the form image to be processed is prepared in advance by color identification. The pre-printed part of the form to be processed is dropped. According to this conventional technique, therefore, it is possible to read the entered characters and recognize them by OCR without the limitation of colors used for preprinting the forms, but a reference form in which no characters are entered is prepared in advance. There is a need.
Further, in this prior art, when preprinting such as a frame line and the written character come into contact with and overlap with each other, no consideration is given to preventing the written character from being erased at the overlapping portion.

As a conventional technique relating to image processing for reading by OCR, a technique described in Japanese Patent Laid-Open No. 6-195509 is known. This prior art is provided with a photoelectric conversion unit that gives the primary color components (red, green, blue) of each pixel from a multicolor image on the assumption that the written character color is black, and has the maximum value of those primary color components. It outputs a signal, removes the color, and extracts only the written characters. However, this prior art is intended to remove color from the multicolored background of a document such as a check, and when the written characters are colored, for example, in the case of blue written characters, the color is removed by removing the color. There is a problem that characters are also removed.

Further, as another conventional technique, for example, a technique described in Japanese Patent Laid-Open No. 10-27213 is known. In this conventional technology, the red, green, and blue primary colors can be read from the photoelectric conversion unit so that characters written in chromatic colors such as red pencils, red ballpoint pens, and blue ballpoint pens can be read in the same way as achromatic characters. Input a component, determine whether the color of the pixel belongs to the color group of achromatic color, red-based color, green-based color, or blue-based color, and if the pixel is the character color specified in advance The maximum value of the three primary color components is output as a black and white luminance signal of the pixel. If it is not the character color, the background level is output as a black and white luminance signal of the pixel. Then, the pixel is converted into a monochrome luminance signal. That is, multivalued image data is converted into binary image data. However, in this conventional technology, when preprinting such as a border line and the written characters contact and overlap, consideration is given to leaving the written characters and dropping the preprint so that the overlapping part does not disappear. Not not.

Further, as another prior art, Japanese Patent Laid-Open No. 6-
The technique described in Japanese Patent Publication No. 290302 is known. In this conventional technique, when preprinting such as a frame line and the written characters come into contact with each other and overlap each other, provisional color separation is performed on the input image data, and pixels forming each of the provisionally color-separated portions are separated. The final color separation is determined by using the geometrical information of (3) and the density information of each of the portions and the background. However, this conventional technique is provided with a binary image buffer, a gray image buffer, and a color image buffer, and when a character for a binary image is cut out and character recognition is performed, a specific portion of the input image, for example, a character When re-checking a holed portion in a stroke or a contact portion between a ruled line and a character, the character is cut out by referring to the gray image buffer and the color image buffer, and character recognition is executed.

Therefore, this conventional technique has a problem in that it is necessary to access a gray image or a color image for each processing unit such as a character cutout unit and a character recognition unit, which makes control complicated. Further, in this conventional technique, the preprint is erased, the binary image in which the written characters are left is not generated and output, and the preprint is removed and the written characters are left in order to correct the reading result. No consideration is given to displaying the value image on the screen. Further, this conventional technique uses linear geometric information such as growing a region that can be regarded as the same color, but the region growth due to the disturbance of the contour of the region in the straight line portion or the curved portion of the character or the color misregistration. The reliability of is not considered.

[0008]

The prior arts described above have various problems as explained together with the description of the respective prior arts.

An object of the present invention is to solve the above-mentioned problems of the prior art, and for a color image to be processed, for a form image in which the color of preprinting such as a frame line and the color of the entered character are different. Then, even if the preprint and the written character touched or overlapped, the preprint was erased and the written character was left. 2
An object of the present invention is to provide an image processing program and an image processing device that can generate a value image and can perform highly accurate character recognition.

[0010]

According to the present invention, the above object is based on a step of setting a preprint color and a written character color of an input form image, and based on the preprint color and the written character color. Determining the color to be binarized, dropping out the preprinted color from the form image, binarizing the image after the dropout with the determined color, This is achieved by causing the information device to perform the step of performing character recognition on the binarized image.

Further, the object is to binarize the input form image with respect to the minimum values of the three primary color components,
By causing the information device to perform the character recognition of the binarized image and the step of performing character recognition again by dropping out the pre-printed color of the form image when the recognized character is rejected. To be achieved.

[0012] Further, the above-mentioned object is means for performing character recognition for an input form image, and means for determining a character entry color and a preprint color of the form image when the character recognition is rejected. Means for determining the primary color component to be binarized from the character entry color and the pre-print color of the form image; and means for dropping out the pre-print color determined from the form image to binarize the primary color component And a means for performing character recognition of the binarized image.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of an image processing apparatus and a character reading system according to the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing the arrangement of a character reading system including an image processing apparatus according to an embodiment of the present invention. In FIG. 1, 100 is a color image scanner, 101 is an image processing unit, 102 is a color identification dropout binarization unit, 103 is a color identification unit for preprinting / written characters,
104 is a pre-printing dropout binarization unit, and 105 is 3
A primary color component minimum value binarization unit, 106 is a control unit, 107 is a character reading unit, 108 is an image storage unit, and 109 is a network.

A character reading system according to an embodiment of the present invention includes a color image scanner 100 for collecting a color image on the surface of a form, and an image processing unit 104 for binarizing color image information from the color image scanner 100 into black and white.
A character reading unit 107 such as an OCR that reads binarized image data and an image storage unit 108 are connected by a network 109.

The image processing unit 101 selects the minimum value of the three primary color components of the image data input from the color image scanner 100, that is, the data portion of the darkest portion of the three primary color components, and performs the monochrome binarization process. Three primary color component minimum value binarization unit 105 and a color identification dropout binarization unit 102 that identifies a preprinted color and performs a dropout process on the color.
And a control unit 10 that controls selection and execution of these processes
6 and. Minimum value of three primary color components Binarization unit 10
Reference numeral 5 selects the minimum value of the three primary color components of the input color image and binarizes the primary color component as a gray value. As a method of binarizing the grayscale value, for example, as disclosed in Japanese Patent Laid-Open No. 11-120333, a method of determining a binarization threshold value based on the grayscale value of each of the pixel of interest and its surrounding pixels. May be

The color identification dropout binarization unit 102
A color identification unit 103 that performs preprinting and color identification of written characters and a binarization unit 104 that drops out a preprinted color using a color identification result. Character reading unit 107
Performs character segmentation and character recognition from the binary image sent from the image processing unit 101, and the image storage unit 1
Reference numeral 08 stores the binary image sent to the character reading unit 107.

FIG. 2 is a flow chart for explaining the processing operation in the image processing unit 101 and the character reading unit 107 in the character reading system shown in FIG. 1. Next, this will be explained. Before the process shown in FIG. 2 is started,
The color image scanner 100 scans a form for reading characters, collects a color image of the form, and transfers the image to the image processing unit 101. This image is temporarily stored and held in a buffer (not shown) provided in the color image scanner 100 or the image processing unit 101.

(1) The minimum value binarization unit 105 of the three primary color components in the image processing unit 101 is a color image scanner 100.
Image, the minimum of the three primary color components, ie,
3 Select the data part of the darkest part of the primary color component and black and white 2
The binary image resulting from the binarization processing is transmitted to the network 109 via the control unit 106 (step 200).

(2) The character reading unit 107 inputs the image binarized by the three-primary-color component minimum value binarizing unit 105, and designates an area to be read, that is, a field. The field can be specified by registering the coordinates of the field as a form format in advance and selecting the field when reading the format, or by extracting the frame line of the form. A method of determining the area to be read from the arrangement and arrangement of the frames may be used (steps 201 and 202).

(3) Next, the character reading unit 107 cuts out characters in the field and executes character recognition to read characters in the field. Then, it is determined whether the result of reading the character in the field is accept or reject. If the result of this determination is that it is an accept, the read results are integrated and simply output (steps 203, 204, 209).

(4) If it is determined in step 204 that the result of character reading is reject, the color image acquired by the color image scanner 100 is preprinted / filled in by the color identification dropout binarization unit 102. The characters are input to the color identification unit 103 (step 205).

(5) The color identification section 103 performs color identification of the preprinted portion such as the frame line of the form and the guide character string and the color identification of the entered character from the input color image (step 20).
6).

(6) Next, the pre-print dropout binarization unit 104 drops out the color using the pre-printed color from the color identification unit 103 to generate a binary image. If the preprint and the entered characters have the same color and it is determined that the dropout binarization is unsuccessful, the preprint dropout binarization unit 104 outputs a reject (step 207).

(7) The character reading unit 107 moves from the pre-print dropout binarization unit 104 to the network 109.
The characters in the field are read from the binary image transmitted to the computer, and the read results are integrated and output (steps 208 and 209).

As described above, the character reading system according to the embodiment of the present invention determines whether the result of the previous character reading in the field is Accept or Reject in the process of step 204, and if it is reject, preprint is dropped. Since character reading is performed on the binary image that has been output, it is possible to prevent the preprinted portion from interfering with character recognition, and improve the reliability of character reading.

FIG. 3 shows the color identification dropout binarization unit 10.
2 is a block diagram showing the configuration of FIG. 2, FIG. 8 is a diagram for explaining a region for identifying the colors of preprinting and written characters in the field, and FIG. 10 is a diagram showing the configuration of the primary color component selection unit 305 in FIG. In FIGS. 3 and 10, reference numeral 301 is a pre-print color identification unit, 302 is a written character color identification unit, 303 is a color component selection determination unit, 305 is a primary color component selection unit, 306 is a component value thickening / lightening unit, and 307. Is a binarization unit, 308 is a pixel identification unit, and other symbols are the same as those in FIG.

The color identification section 103 is a preprinted color identification section 30.
1, a written character color identification unit 302, and a color component selection determination unit 3
03, and performs preprinting and color identification of written characters. Then, the blue component 310, the green component 311, and the red component 31 of the color image from the color image scanner 100.
2 is a pre-print color identification unit 301 and a written character color identification unit 30
Entered in 2. The pre-print color identification unit 301 sets a processing region for identification for a partial color image in the field, detects the color distribution of pixels in the region, and identifies the pre-print color. . The pre-print color is 1
The number of colors is not limited to one, and a plurality of colors may be used. For example, two colors, red and green, may be extracted as the preprinted colors.

Next, referring to FIG. 8, an area for identifying the colors of the preprint and the written character in the field will be described. In FIG. 8A and FIG. 8B, the field 80
The frame line 800 in 2 is colored, for example, a red or blue line. Further, the written character 801 is a reading target and is a black or colored character. The preprinted frame line 800 and the entry character 801 are shown as being overlaid. If the color of the border and the color of the entered characters are different, the border can be dropped out.

In order to identify the preprinted color such as the frame line, as shown in FIG. 8A, the color identification processing areas 803, 80 are provided.
4 is set. In this processing area, pixel color identification is performed, the color pixel distribution is obtained, and the pre-printing color is determined. In addition, in order to identify the color of the written characters, FIG.
As shown in (b), a color identification processing area 813 is set. In this processing area, when the color of the pixel is identified, the distribution of the color pixel is obtained and the color of the written character is determined.

Returning to the description of FIG. 3, the entered character color identification unit 3
Reference numeral 02 sets a processing area for identification for a partial color image in the field, for example, an area 813, and detects the appearance distribution of the color of pixels in the area to identify the color of the written character. . Further, the color component selection determination unit 303 uses the identified preprinted color and the entered character color to determine which of the three primary color components is the primary color component to be selected as the grayscale image to be binarized. For example, if the preprint color is identified as red,
When the written character color is identified as blue, the red component is determined to be the primary color component to be selected. If the preprinted color is identified as green and the entered character color is identified as black, the green component is determined to be the primary color component to be selected. Then, the color component selection determination unit 303 drops the pre-printed color into the dropout 2
Output to the primary color component selection unit 305 of the binarization unit 104,
In addition, the information 320 of the preprinted color and the entered character color is converted into the component value thickening / lightening unit 306 of the dropout binarizing unit 104.
Output to.

The processing in the color identification section 103 described above is
Color identification is performed on the color image of one field. Then, after the above-described color identification is performed, the dropout binarization unit 104 described below uses the color image data of the same field to create a black-and-white binary image with the preprinted color dropped out. Processing to generate is performed.

As shown in FIG. 3, the dropout binarization unit 104 includes a primary color component selection unit 305, a component value darkening / lightening unit 306, a pixel color identifying unit 308, and a grayscale binarizing unit 307.
It consists of According to the result of the color component selection determination unit 303, the primary color component selection unit 305 determines the three primary color components, that is, the blue component 310 and the green component 3 for each input pixel.
11. One primary color component is simply selected from the red component 312 and the gray value is output.

As shown in FIG. 10, the primary color component selection unit 305 selects from the input blue component 310, green component 311, and red component 312 the color type information 1104 output from the color component selection determination unit 303. Then, one primary color component is selected according to the above, and is output as a gray value 1105. Here, if the color type information 1104 is red, the red component 3
10 is output as the gray value 1105. If the color type information 1104 is green, the green component 311 is output as the gray value 1105. Further, if the color type information 1104 is blue, the blue component 311 is output as the gray value 1105.

In the above description, the color component selection determination unit 3
The color type information 1104 output from the printer 03 is, for example, 2 of red and blue among the three primary color components as preprinted colors.
When the color is indicated, the primary color component selection unit 305 selects the maximum value of the red component and the blue component of each input pixel,
The maximum value may be output as the gray value.

The component value darkening / lightening unit 306 performs processing for reducing (darkening) or increasing (lightening) the magnitude of the gray value of the input color. At this time, the pixel color identification unit 30
Reference numeral 8 identifies the color of the pixel of interest, and the corresponding color type 322
Is sent to the thickening / lightening section 306. Thickening / lightening section 306
The pre-printed color, the written character color, and the color of the pixel of interest are input, and the gray value of the pixel of interest is darkened or lightened according to the type of the input color. For example, if the preprinted color is red, the entered character color is blue, and the pixel of interest is red, the pixel of interest is dimmed. That is, the gray value of the pixel of interest is increased so that it becomes closer to white. The darkened / lightened gray value output from the darkening / lightening unit 306 is input to the light / dark binarization unit 307, and is generated as a binary image. The gradation binarization unit 307 is disclosed in, for example, Japanese Patent Laid-Open No. 11-120.
It can be realized by the method disclosed in Japanese Patent No. 333 publication.

FIG. 4 is a diagram for explaining the processing operation of the color identification of the preprinted / written characters in step 206 in the flow described with reference to FIG. 2. Next, this will be described. This processing is performed separately by setting the area for each of the color identification of the preprint and the color identification of the written character.

(1) First, the search range in the color image in the field is set. The search range is, for example, FIG.
As described above, the processing areas 803 and 804 are used to identify the preprinted color and the processing area 813 is used to identify the written character color (step 400).

(2) Next, the image in the area set in the field is scanned. The scanning procedure is performed by scanning the pixel of interest in the horizontal direction and then in the vertical direction, and repeating the processing from step 403 to step 407 described below while scanning the image (step 401).

(3) While scanning the image, first, the color of the pixel of interest is determined. Here, as the color type of the pixel of interest, color identification is performed with respect to four types of blue type, green type, red type, and black type to obtain the type of color (step 403).

(4) Next, the number of colors of pixels determined in step 403 is counted. That is, step 403
If the pixel of interest is determined to be blue in step 1, the number of blue pixels is increased by one, and if the pixel of interest is determined to be green in step 403, the number of green pixels is increased by one. Similarly, when it is determined in step 403 that the pixel of interest is red or black, the number of red pixels and the number of black pixels are increased by 1 (steps 404 to 407).

(5) Next, the color type of pre-printing or written characters is determined based on the counted numbers of blue pixels, green pixels, and red pixels. Here, the color having the maximum number among the blue pixel, the green pixel, the red pixel, and the black pixel is determined as the target color type, and the process ends (step 402).

FIG. 5 is a diagram for explaining a character reading processing operation in the character reading unit 107, which will be described next.

(1) First, 2 in the set field
Input the value image and extract the character line that is the sequence of characters.
Then, the character is cut out (steps 500 to 502). (2) Character recognition of the cut out characters is performed and knowledge processing is performed. The knowledge process is a process of preparing a dictionary of addresses and surnames in advance and matching the character recognition result with the addresses and surnames in the dictionary to improve the recognition accuracy (step 5).
03, 504).

FIG. 6 is a diagram for explaining an example of a color form to be processed, which will be described next.

For example, as shown in FIG. 6, frames 601, 603, and 608 are printed on the color form 600, and in the illustrated example, characters 602 are written in the frame 601. Further, preprinted characters 604 and 605 are printed in advance in the frame 603, and characters 606 and 607 are entered. Further, a character 609 is entered in the frame 608.
As in the example shown in the figure, there are forms in which written characters are in contact with preprints, and forms that overlap. For example, in the example shown in the figure, the written character 602 and the frame 601 overlap. In addition, the input characters 606 and 607 and the preprint 604,
It overlaps with 605. If the preprint and the written characters are in contact with each other or if they overlap, the reading of characters will be hindered. Therefore, it is necessary to drop out the preprint.

Here, as an example, the frame 601 is red,
In the field in which the input character 602 is blue or the field in which the frame 603 and the preprints 604 and 605 are green, the preprinted frames 601 and 603 and the preprint character 6
04 and 605 are dropped out, that is, white,
The written characters 602, 606, 607 are black. In addition,
When the color of the frame 608 and the color of the written character 609 are the same, the preprint dropout using the color information in the embodiment of the present invention is not applicable. As an example in which the color of the entered character 609 and the color of the preprinted frame 608 are the same, the entered character 6
In some cases, 09 is green and the preprinted frame 608 is green, or the entered characters 609 are black and the preprinted frame 608 is black. As described above, in the case of the same color, it is difficult to perform the preprint dropout, and in the case of the same color, the preprint is performed by the process of step 207 of the flow described with reference to FIG.
The entered characters are output as black.

FIG. 7 is a diagram illustrating an example of a binary image as a result obtained by performing image processing according to the embodiment of the present invention. In the example shown in FIG. 7, the binary image resulting from the preprinting dropout from the color image of the color form 600 described with reference to FIG. 6 is shown as 700. The binary image 700 shown in the figure is the input characters 602, 6 in FIG.
06 and 607 are processed as black as the written characters 701, 702 and 703, and preprinted 601, 603 and 604,
605 is dropped out and is white. on the other hand,
Since the preprint 608 and the written character 609 have the same color, they are not dropped out in the binary image 700, and both are represented in black.

FIG. 9 is a flow chart for explaining the details of the processing operation in step 403 of the pixel color judgment described with reference to FIG. 4, and this will be described next.

(1) First, the three primary color components of the target pixel, that is, the red component value R, the green component value G, and the blue component value B are input. Generally, the size of the color component of a pixel is represented by 8-bit information, and each of R, G, and B takes a value of 0 to 225 (step 900).

(2) Next, the magnitudes of the red component value R and other component values are compared to determine whether or not the conditions of R> a · G and R> a · B are satisfied, If the condition is satisfied, the pixel of interest is registered as red. In the above equation, a is a predetermined constant of 1 or more, and the same applies to subsequent steps. When this constant a is set to a large value, pure color, in this case only pure red, is determined as red (steps 901 and 902).

(3) If the condition is not satisfied in the judgment of step 901, then the green component value G is compared with the other component values to determine G> a · R and G> a. It is determined whether or not the condition B is satisfied, and if the condition is satisfied, the pixel of interest is registered as green (steps 903 and 904).

(4) If the condition is not satisfied in the determination in step 903, the blue component value B is compared with the other component values in the same manner as described above, and B> a · R and B >
It is determined whether or not the condition of a · G is satisfied, and if the condition is satisfied, the pixel of interest is registered as blue (steps 905 and 906).

(5) Further, in the judgment of step 905,
If the condition is not satisfied, then the red component value R,
Both the green component value G and the blue component value B are smaller than a predetermined value c (R, G, B <c), and the maximum values of the red component value R, the green component value G, and the blue component value B are Less than the value obtained by multiplying the minimum value by a constant d (Max (R, G, B)
<D · Min (R, G, B)), that is, whether the values of the three primary color components are both small, and those values are substantially the same, and the values of the three primary color components are both small. , And if those values are substantially the same, the pixel of interest is registered as black and the process ends (step 90).
7, 908).

FIG. 11 is a block diagram showing the structure of the component value thickening / lightening unit 306 in FIG. 3, and FIG. 12 is a reference table for thickening / lightening process determination used to determine the type of thickening / lightening process. It is a figure explaining an example, and these are explained next. In FIG. 11, reference numeral 1200 is a darkening / lightening processing type determining unit, 1201 is a darkening multiplying unit, 1202 is a lightening multiplying unit, and 1203 is a selecting unit.

As shown in FIG. 11, the component value thickening / lightening unit 306 includes a thickening / lightening processing type determining unit 1200.
It includes a thickening multiplication unit 1201, a lightening multiplication unit 1202, and a selection unit 1203. Then, the gray value 321 output from the primary color component selection unit 305, the written character color and preprinted color information 320 from the color component selection determination unit 303, and the color information 322 of the pixel of interest from the pixel color identification unit 308. Is a component value thickening / lightening unit 306 shown in FIG.
Entered in.

The darkening / lightening processing type determining unit 1200
Using the information 320 of the preprinted color and the entered color and the color information 322 of the pixel of interest, the darkening or lightening output of the input gray value 321 or the darkening of the selection of the same value as the input is output. .Reference table 13 for determining thinning processing
00 to determine.

The reference table for determining the darkening / lightening processing used in the darkening / lightening processing type determination unit 1200 is
It is configured as shown in FIG. 12 and has a horizontal item 1340.
Indicates the type of pre-print color, and the pre-print color is red 130
The case of 1, green 1302, blue 1303, and black 1304 is shown. The vertical item 1342 indicates the type of the written character color, and shows the case where the written character color is red 1305, green 1306, blue 1307, and black 1308. Further, the vertical subdivision item 1341 indicates the type of the target pixel color, and the color of the target pixel is red 131 for each of the entered character colors.
0, green 1311, blue 1312, and black 1313 in some cases.

Then, in the illustrated lookup table 1300,
In the case of each set of the preprinted color, the entered character color, and the pixel color of interest, the contents indicating three types of darkening, lightening, and the same value are set in advance.
Stored in memory. For example, if the preprinted color 1304 is green 1302 and the entered character color 1342 is red 1305, and if the pixel color of interest is red 1310, the thickening processing indicated by 1320 is selected as the type of darkening / lightening processing.
If the pixel color of interest is green 1311, 132
The lightening process shown in 1 is selected as the type of the darkening / lightening process. If the target pixel color is blue 1312 or black 1
In the case of 313, it is determined that the same value as the type of the thickening / lightening processing, that is, the thickening / lightening is not performed and the value is output as it is.

The darkening multiplication unit 1201 multiplies the inputted gray value 321 by a predetermined value to reduce the gray value, that is, outputs a dark density. Further, the thinning multiplication unit 1202
Multiplies the input gray value 321 by a predetermined value to increase the gray value, that is, output a light density. Selector 1
Reference numeral 203 designates either the thickened output or the lightened output of the inputted gray value 321 or the same value as the input according to the result output of the darkening / lightening processing type determination unit 1200, and outputs it as 1213. .

The embodiment of the present invention described above is
Although it has been described that the preprinted color and the written character color of the form to be read are automatically identified and processed, the present invention is
It is also possible to include the preprinted color and the entered character color in advance in the form identification number indicating the type of the read form, and read the form identification number to set the type of the read form. In this case, the color identification dropout binarization unit 102 shown in FIG.
Can be configured only by the pre-print dropout binarization unit 104.

FIG. 13 is a flowchart for explaining the processing operation in the image processing unit 101 and the character reading unit 107 in the character reading system according to another embodiment of the present invention when the form identification number is read and the type of the read form is set. Then, this will be described below. In addition,
This process is performed after it is determined to be rejected in the process of step 204 in the flow described with reference to FIG.

(1) First, the type of target form is set. Here, the form identification number written on the form may be read to set the form type, or
The layout of frame lines may be extracted from the form image, and the type of form may be set based on the layout of the frame lines prepared in advance and the form identification to be compared (step 1000).

(2) Next, the format information prepared in advance for each form type set in step 1000, that is, the format information is input. As the format information, the position coordinates of the field, the preprint color, the entry color, and the like are prepared in advance for each form (step 1001).

(3) Next, the processes from step 1003 to step 1008 described below are repeated until the fields set as the format information are exhausted, and the process ends (step 1002).

(4) The position coordinates of the field are set from the format information, and the pre-printing color of the field is set from the format information. Further, the written character color of the field is set from the format information (steps 1003 to 1005).

(5) Next, the color image in the field is input, and binarization processing is performed to drop out the pre-print color using the pre-print color and the entered character color set previously.
Then, the characters in the field are read from the binarized image (steps 1006 to 1008).

The processing according to the embodiment of the present invention described above is
It can be configured as a processing program, and this processing program is HD, DAT, FD, MO, DVD-RO.
It can be provided by being stored in a recording medium such as an M or a CD-ROM.

As described above, according to the embodiment of the present invention, the means for identifying the color of the preprinted / written character from the color image in the field and the binarized image in which the identified preprinted color is dropped out are provided. Since the character reading is retried for the dropped-out binarized image, the reliability of the character reading in the field where the preprint and the written character overlap can be improved. . Further, it is possible to reduce the processing time by providing only a unit for generating a binarized image in which the preprinted color is dropped out.

Further, according to the embodiment of the present invention, since the means for setting the preprinted color and the entered character color of the field in the form are provided, the preprinted color is different for each form and each field. Pre-printing can be dropped out even on the existing forms, and the reliability of character reading can be improved. Further, based on the preprinted color, the written character color, and the target pixel color, means for generating a grayscale image by selecting the maximum value of a single color component or a plurality of components from the three primary color components of the color image based on the preprinted color. Since a means for darkening or lightening the grayscale image is provided, a grayscale value can be used for a single color color dropout or a plurality of colors, for example, simultaneous dropout of red and blue, and a color dropout image The image quality of can be improved.

Further, according to the embodiment of the present invention, means for pre-registering the pre-printing color or the entered character color as the form format for each field of the form image, and the pre-printing of the registered field for each field Since the means for reading out the color or the written character color is provided, it is possible to improve the image quality of the color dropout image for a form whose format is known.

Further, according to the embodiment of the present invention, means for setting a plurality of search areas in each field of the form image, means for identifying the preprinted color in the predetermined search area, or Since the means for identifying the written character color in another predetermined search area is provided, it is possible to obtain the effect that color dropout can be performed even for a form whose format relating to the preprinted color is unknown.

The other embodiments of the present invention described above are as follows.

1. In an image processing apparatus for generating a black-and-white binary image for reading characters from a field of a form image, a binary image generating means for binarizing a minimum value of three primary color components from a color image in the field, and the field And a unit for generating a binarized image in which a preprinted color is dropped out from the color image in the image processing apparatus.

2. In an image processing apparatus for generating a black-and-white binary image for reading characters from a field of a form image, a binary image generating means for binarizing a minimum value of three primary color components from a color image in the field, and the field Means for generating a binarized image in which the preprinted color is dropped out from the color image in the field, and the preprinted color is dropped out from the color image in the field.
A means for generating a binarized image is a means for setting a preprinted color or a written character color in a form field, and selecting the maximum value of a single color component or a plurality of components from three primary color components of a color image based on the preprinted color. Means for generating a grayscale image, a means for darkening or graying the grayscale image based on the preprinted color, the written character color and the pixel color of interest, and a binarization of the grayscale image which has been darkened or darkened with a threshold value. And an image processing device.

3. The means for setting the preprinted color or the entered character color of the form is a means for preregistering the preprinted color or the entered character color as a form format for each field of the form image, and the registered field for each field. 3. The image processing apparatus according to claim 2, further comprising means for reading out the preprinted color or the written character color.

4. The means for setting the preprinted color or the entered character color of the form, for each field of the form image, means for setting a plurality of search areas in the field, means for identifying the preprinted color in the predetermined search area, 3. The image processing apparatus according to claim 2, further comprising means for identifying a written character color in another predetermined search area.

5. In a character reading system including a character reading device for reading a black-and-white binary image generated from a field of a form image and an image processing device for generating a black-and-white binary image from a field of the form image, the image processing device comprises: , A binarized image generating means for binarizing the minimum value of the three primary color components from the color image in the field, and means for generating a binarized image in which the preprinted color is dropped out from the color image in the field, The character reading device is caused to read the binarized image from the binarized image generating means for binarizing the minimum value of the three primary color components, and when the result is reject, the color image is read. And a means for retrying the reading of the binarized image from the means for generating the binarized image in which the preprinted color is dropped out. Character reading system that.

[0079]

As described above, according to the present invention, in the color form to be processed, the form image in which the color of the preprint such as a frame line and the color of the written character are different,
Even if the preprint and the written characters touch or overlap,
Pre-printing can be erased and a binary image in which the written characters are left can be generated, and highly accurate character recognition can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a character reading system including an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a processing operation in an image processing unit and a character reading unit in the character reading system shown in FIG.

FIG. 3 is a block diagram showing a configuration of a color identification dropout binarization unit.

FIG. 4 is step 2 in the flow described with reference to FIG.
It is a figure explaining the processing operation of the color identification of the preprinting / writing character in 06.

FIG. 5 is a diagram illustrating a character reading processing operation in a character reading unit.

FIG. 6 is a diagram illustrating an example of a color form to be processed.

FIG. 7 is a diagram illustrating an example of a binary image as a result obtained by performing image processing according to the embodiment of the present invention.

FIG. 8 is a diagram illustrating an area for identifying a color of a preprint and a written character in a field.

FIG. 9: Step 4 of pixel color determination described with reference to FIG.
3 is a flowchart for explaining the details of the processing operation in 03.

10 is a diagram showing a configuration of a primary color component selection unit in FIG.

11 is a block diagram showing a configuration of a component value thickening / lightening unit in FIG. 3;

[FIG. 12] Concentration used to determine the type of thickening / lightening
It is a figure explaining the example of the reference table for lightening process determination.

FIG. 13 is a flowchart illustrating a processing operation in a character reading system according to another embodiment of the present invention.

[Explanation of symbols]

100 color image scanner 101 image processing unit 102 Color identification dropout binarization unit 103 Pre-printing / coloring character identification area 104 Pre-print dropout binarization unit 105 3 primary color component minimum value binarization unit 106 control unit 107 character reading unit 108 image storage unit 109 network 301 Pre-print color identification section 302 Input color color identification section 303 Color component selection judgment unit 305 Primary color component selection section 306 Component value darkening / lightening part 307 binarization unit 308 pixel identification unit 1200 Densation / lightening processing type determination unit 1201 Concentration multiplication unit 1202 Lightening multiplication unit 1203 Selector

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) H04N 1/60 H04N 1/46 ZF term (reference) 5B029 CC26 5B050 BA06 BA16 BA20 CA07 DA02 DA04 DA06 EA02 EA04 EA09 5B057 AA11 BA02 BA11 CA01 CA08 CA12 CA16 CB01 CB06 CB08 CB12 CB16 CE12 CE16 DB02 DB06 DB09 DC25 5C077 LL20 MP08 PP21 PP32 PQ22 RR02 5C079 HB01 LA02 LA34 LA39 MA01 NA29

Claims (5)

[Claims] 1. A step of setting a preprint color and a written character color of an input form image; a step of discriminating a color to be binarized based on the preprinted color and the written character color; Form image skeleton Dropping out the preprinted color, binarizing the image after the dropout with the determined color, and performing character recognition on the binarized image A program that causes an information device to execute and. 2. The preprinted color and the written character color are
The program according to claim 1, which is set for each field of the form image. 3. The preprinted color or the entered character color of the form is set in advance by registering the preprinted color or the entered character color as a form format for each field of the form image. 3. The program according to claim 2, which is executed by reading out a preprinted color or a written character color of the field. 4. A step of binarizing an input form image with respect to a minimum value of three primary color components, a step of recognizing a character of the binarized image, and a case of recognizing the recognized character. A program for causing an information device to execute the step of dropping out a preprinted color of the form image and performing character recognition again. 5. A means for performing character recognition on an input form image, a means for determining a character entry color and a pre-print color of the form image when the character recognition is rejected, and the form image. Means for determining a primary color component to be binarized from the character entry color and the preprinted color, means for dropping out the preprinted color determined from the form image, and binarizing the primary color component; An image processing apparatus, comprising: means for recognizing a character of a binarized image.