JP2007041832A - Difference image extraction apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a difference image extraction apparatus for accurately extracting an image of a difference between unwritten printed matter where no positioning mark is recorded and written printed matter. <P>SOLUTION: The image data of unwritten printed matter 201 is converted into binary image data and histograms of black pixels are created in the horizontal direction 203 and the vertical direction 204 of a paper surface. An area where black pixels alternate with white pixels in a regular fashion is determined to be a character area, which is used as a template 207. The coordinates of two points on the template 207 are calculated while one corner of the paper surface is regarded as the original point. The binary image data of written printed matter is collated with the binary image data of the template to find an area of the template that matches it, to calculate coordinates corresponding to the two points on the template. Differences between the coordinates of the written printed matter and the coordinates of the template are calculated and a correction value is determined to correct the position of the binary image data. Differences between the binary image data of both printed matters whose positions have been corrected are extracted for output. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a difference image extraction apparatus that extracts a difference image between an unwritten print and a written print.

After writing a comment on the distributed printout, you may want to extract only that comment. In such a case, when the original unwritten print and the written print are captured by an image input device such as a scanner and compared as image data, the positions of the two image data must be aligned.
When a mark for alignment is printed on the printed material, the positions of the two image data are aligned based on the mark, and then the difference between the two image data is extracted, so that only the comment written on the printed material is extracted. be able to.

However, in the printed matter on which the alignment mark is not printed, the alignment accuracy differs greatly depending on which position is used as the alignment reference. For example, if a region such as a picture or a figure is used as a reference for alignment, alignment may be difficult due to image density, color change, figure shape, or the like during scanning.
In order to solve this problem, for example, Patent Document 1 discloses a technique for performing character recognition processing of image data, extracting two characters from the image data, and performing alignment based on position information of the extracted characters. Is disclosed.
JP-A-11-175717

However, as in the prior art, man-hours are required to print the alignment mark on the printed material in advance, and the technique of Patent Document 1 requires character recognition processing, which requires man-hours for character recognition processing. Necessary.
Therefore, the present invention provides a differential image extraction apparatus that can perform high-precision alignment with a small number of man-hours without performing character recognition processing even on image data without alignment marks. Objective.

  In order to solve the above-described problems, the present invention provides a differential image extraction apparatus that extracts a difference between a last-printed printed matter and a written matter that has been written, an image obtaining unit that obtains images of two printed matters, An image extracting means for extracting the character area of the printed material to be written, an alignment means for matching the extracted character area with the image data of the written printed material to align both printed materials, and 2 in the aligned state Difference extraction means for extracting a difference image of two printed matters.

With the configuration as described above, a character area with higher accuracy than the picture or graphic area is used as a reference for alignment. Therefore, even if there is no alignment mark, high-precision alignment is performed, and the difference image is efficiently obtained. Can be extracted.
In addition, the image extraction unit includes a binarization unit that converts the image of the last written print acquired by the image acquisition unit into image data binarized into a black pixel and a white pixel for each pixel; A histogram generation unit that generates a histogram obtained by projecting image data that has been converted to black pixels in the value conversion unit into the vertical and horizontal directions of the last-printed printed matter, and a determination that determines a character area from the shape of the generated histogram And a template extraction unit that extracts M rows and N columns of image data from the character region determined by the determination unit as a template.

With such a configuration, the image data can be discriminated into black pixels corresponding to the character portion and white pixels corresponding to the background portion, and the character region can be discriminated from the picture or graphic region.
Further, the alignment means is extracted by the target binarization unit that converts the written image of the printed material into image data binarized into black pixels and white pixels for each pixel, and the template extraction unit. A matching area recognition unit for recognizing a region where the template and the image data binarized by the target binarization unit match, and a coordinate value of the region determined to be a match by the matching region recognition unit. And an alignment portion to be matched.

With such a configuration, even if the position of the image data of the last-written printed matter acquired by the image acquiring unit and the written printed matter is slightly deviated, accurate alignment is performed, so that the difference image is accurate. Can get well.
In addition, the difference extracting unit extracts a region obtained by adding a blank region that is a set of white pixels to a region determined as the character region of the last-written printed material, and a region corresponding to the written printed material. An extraction unit and a calculation unit that calculates a difference between the extracted image data of both regions are included.

With such a configuration, a differential image is extracted by excluding a picture or graphic data area that may have a large density difference in an image acquired by the image acquisition means, so an accurate differential image can be obtained. .
In addition, when the difference calculated by the calculation unit is not “0”, output means for outputting as a black pixel is further provided.

With such a configuration, writing that is the difference between an unwritten print and a written print can be accurately output as a black pixel.
In addition, when the template extraction unit can extract a plurality of M rows and N columns of image data, it extracts the largest one of them.
With such a configuration, a template used for alignment is not localized on the paper surface of the printed material, and thus accurate alignment is possible.

  The present invention also provides a differential image extraction method of a differential image extraction apparatus that extracts a difference between a final printed material and a written material, an image acquisition step for acquiring images of two printed materials, An image extraction step for extracting the character area of the printed material to be written, an alignment step for comparing the extracted character region and the written image of the printed material to align the two printed materials, and two states in the aligned state And a difference extraction step for extracting a difference image of the printed matter.

  By using such a method, a character area with higher accuracy than a picture or graphic area is used as a reference for alignment. Therefore, even if there is no alignment mark, the accuracy of unprinted printed matter and written printed matter is high. It is possible to extract the difference image efficiently by performing alignment.

Hereinafter, embodiments of a differential image extracting apparatus according to the present invention will be described with reference to the drawings.
(One embodiment)
FIG. 1 is an external view of a hardware configuration of an embodiment of a differential image extraction apparatus according to the present invention.
This differential image extraction apparatus includes a scanner 101 and a personal computer (hereinafter referred to as “PC”) 102.

The scanner 101 scans the placed document, converts it into digital image data, and outputs the digital image data to the PC 102.
This document includes a non-written printed matter as a reference and a printed matter on which a memo or the like is written.
The PC 102 has a built-in CPU and stores a program for extracting a difference image in the ROM. In accordance with this program, the CPU exhibits functions such as an image recognition extraction unit, a registration unit, and a difference extraction unit.

The PC 102 extracts difference image data such as a memo from the digital image data of the unwritten print and the written print input from the scanner 101, and displays the difference image on the display unit 103 including a liquid crystal display or the like. To do.
FIG. 2 is a diagram for explaining the outline of the difference image extraction procedure in the difference image extraction apparatus.

(1) First, the scanner 101 acquires digital image data of an unwritten printed matter 201 and outputs it to the PC 102.
The control unit of the PC 102 converts the input digital image data into image data 202 binarized with a predetermined threshold value. Next, a character area is extracted using the binarized image data 202. At this time, histograms 203 and 204 of black pixels in the horizontal direction and vertical direction of the paper surface of the unwritten printed matter 201 are generated to recognize the character area.

(2) Next, a predetermined part of the character area is extracted as a template 207.
FIG. 3 is a diagram for explaining an example of an image recognition extraction step for extracting the template 207 from the character area.
The image data of each pixel of the digital image data of the unwritten printed matter is converted into binary image data of black pixels and white pixels with a predetermined threshold value. This threshold value is selected so that the image data of the unprinted printed matter on the paper surface is white pixels.

A histogram 310 of black pixels in the horizontal direction and a histogram 320 of black pixels in the vertical direction are generated from the binarized image data 301 of the unwritten printed matter.
The shape of the generated histogram 310 in the horizontal direction includes regions 331, 333, and 335 in which many black pixels are present, and white pixel regions 332, 334, and 336 alternately, and includes a region 337 in which many black pixels are present. . In the area where the black pixel areas 331, 333, 335 and the white pixel areas 332, 334, 336 are regularly repeated in the shape of the histogram 310, the black pixel area 331 and the like are the character line existence areas, and the white pixel area 332 Etc. are determined to be between the lines. Since the black pixel region 337 does not include a white pixel region, it cannot be determined as a character region, and thus is determined to be a picture or graphic region 205.

  Further, the generated vertical histogram 320 has black pixel areas 341, 343, and 345 and white pixel areas 342, 344, and 346 alternately followed by a black pixel area 347. In an area where the black pixel area 341 and the like and the white pixel area 342 and the like regularly repeat, it is determined that the black pixel area 341 and the like are character string existing areas and the white pixel area 342 and the like are between the characters. The black pixel area 347 is determined to be a picture or graphic area 205.

As a result, it is determined that the character area 350, the character area 351, and the graphic area 205 sharing the area 306 exist on the paper surface 302. The other areas form blanks.
Next, the maximum M rows and N columns of image data are extracted from the character area 350. Further, the maximum M ′ row N ′ column image data is extracted from the character area 351. The value of M × N and the value of M ′ × N ′ are compared, and a larger value is adopted as image data of M rows and N columns. In the binary image data 301 of the unwritten printed matter, the M ′ row N ′ column of the character region 351 is larger than that of the character region 350, and therefore image data from the character region 351 to M row N column is used as a template. adopt.

The reason for making the template as large as possible is to make the error as small as possible when aligning with the object.
A coordinate value with the origin (0, 0) at one corner of the paper surface 302 of the position (X ₁ , Y ₁ ) of the upper left 304 and the position (X _M , Y _N ) of the lower right 305 of this template is used. Ask.
Again, a description will be given with reference to FIG.

(3) The printed matter 211 on which the memo 212 is written is read by the scanner 101 to acquire digital image data and output to the PC 102.
In the PC 102, binary image data 213 binarized with a predetermined threshold value is obtained.
(4) Next, the template 207 extracted from the binarized image data 202 of the unwritten printed matter as a reference is collated with the binarized image data 213 of the written printed matter, and an area that matches the template 207 Find 214.

(5) When the region 214 is found, the coordinate values of the upper left portion and the lower right portion are obtained with the upper left portion of the paper surface 215 as the origin, and (X ₁ , Y ₁ ), (X _M , Y _N ) of the template 207 and The binarized image data of the printed material written so as to match is moved to correct the position.
FIG. 4 is a diagram for explaining the position correction step.
In the alignment step, the binarized image data 213 of the printed material in which the binary image data of the template 207 is written to indicate which region of the printed material 402 in which the template 207 of the unwritten printed material 401 matches is written. Find while changing the range.

When an area 214 having a predetermined coincidence or higher with the binary image data of the template 207 is found, the upper left 403 of the printed surface of the printed matter 402 is set to the origin (0, 0), and the upper left 404 of the area 214 is displayed. The coordinates (X ₁ ′, Y ₁ ′) and the coordinates (X _M ′, Y _N ′) of the lower right 405 are obtained.
Note that the predetermined degree of coincidence means that the degree of coincidence is compared by shifting the range in a state where the image data of each pixel of the template 207 and the image data of each pixel of the region 214 are 95% or more, for example. Also says that there are no more matching areas.

Next, in the position correction step, the coordinates (X ₁ , Y ₁ ) of the upper left portion 304 of the template 207 with the origin (0, 0) being the upper left portion 303 of the unprinted printed material 401 of the template 207 are obtained in advance. The coordinates (X ₁ ′, Y ₁ ′) of the region 214 are compared, and the coordinates (X _M , Y _N ) of the lower right 305 of the template 207 are compared with the coordinates (X _M ′, Y _N ′) of the region 214. Then, the written matter 402 that has been written is set to the same position as the printed matter 401 that has not been written yet.

For example, if X ₁ −X ₁ ′ = −1.00, X _M −X _M ′ = −1.00, Y ₁ −Y ₁ ′ = 0, Y _N −Y _N ′ = 0, writing is performed. The printed product 402 image data is translated by 1.00 to the left as indicated by an arrow 406. Of course, when the position cannot be corrected only by vertical and horizontal translation, for example, when the image data 213 is tilted, rotational movement is performed.
A description will be given with reference to FIG. 2 again.

  The area 216 of the printed material 211 that has been written corresponding to the area determined as the picture area 205 in the unwritten printed material 201 is excluded from the difference extraction targets. This is because, in pictures and the like, it is difficult to make the binarization correct as if the pixel gradation is clearly a black pixel and a white pixel as in the character area, and the unwritten printed material 201 was written. This is because it is considered that the degree of coincidence of the binarized image data with the printed material 211 is low.

Finally, in the difference extraction step, a difference between the binarized image data 202 of the unwritten printed material 201 and the binarized image data 217 of the printed material 211 that has been subjected to position correction is extracted. If the corresponding binarized data is the same, it becomes a white pixel, and if it is not the same, it becomes a black pixel.
As a result, a memo 212 is extracted as a difference image. The difference image data 208 is output, and the difference image is displayed on the display unit 103 of the PC 102.

In the above embodiment, the differential image extraction apparatus is configured by the scanner 101 and the PC 102. However, the differential image extraction apparatus may be configured by connecting a printer to the PC 102 and outputting the differential image from the printer.
In the above embodiment, the template 207 is substantially matched with the character area 351, but may be a part of the character area 351.

In the above embodiment, the character area is determined by generating a histogram. However, the page is divided, the ratio of the black pixel to the white pixel is obtained in each divided area, and the ratio is predetermined. The character area may be determined by other methods, such as determining an area within the range as a character area.
Next, the operation of this embodiment will be described with reference to the flowcharts of FIGS.

First, a reference unwritten print is converted into image data by the scanner 101 and output to the PC 102 (S502).
In the PC 102, the input image data is converted into image data binarized into black pixels and white pixels for each pixel (S504).
Next, a histogram of black pixels is generated in the vertical direction and the horizontal direction of the page of the unwritten printed matter (S506), and whether or not there is a region where the histogram shape regularly repeats substantially “0” in each direction. Is determined (S508).

If there is, the area is determined as a character area (S510), and the process proceeds to S514.
If not, the area is determined to be a graphic area (S512). Since there is no character area, the process ends.
In S514, image data of M rows and N columns is extracted from the character area (S514). It is determined whether image data of M ′ rows and N ′ columns can be extracted from other character regions (S516).

When the extraction has been completed, it is determined whether or not M × N> M ′ × N ′ (S518). If not, the image data in M ′ rows and N ′ columns is replaced with image data in M rows and N columns (S520). ), The process returns to S516.
When M × N> M ′ × N ′ in S518, when image data of M ′ rows and N ′ columns cannot be extracted from other character areas in S516, the image data of M rows and N columns is stored as a template (S522). .

Next, the written material that has been written is converted into image data by the scanner 101 and output to the PC 102 (S602).
In the PC 102, the input image data is converted into image data binarized into black pixels and white pixels for each pixel (S604).
The template image data stored in S522 is collated with the written image data of the printed matter (S606). It is determined whether there is a matching area (S608). If not, the process is terminated, and in some cases, the position (X ₁ ′, Y ₁ ′), (X _M ′, Y _N ′) of the image data area in which the writing has been performed, and the template image data Differences ΔX ₁ , ΔX _M , ΔY ₁ , ΔY _N from the positions (X ₁ , Y ₁ ) and (X _M , Y _N ) are calculated, and correction values for moving the image data of the printed matter that has been written are calculated. Obtain (S610).

The image data of the printed material written according to the obtained correction value is moved (S612).
Next, the difference between the binarized image data between the unwritten printed material and the written printed material is extracted except for the graphic region (S614), the difference is “0” as a white pixel, and the difference is “0”. ”Are output as black pixels (S616), and the process ends.

  The differential image extraction apparatus according to the present invention can be used in the field of business processing because it can correctly extract a differential image between an unwritten printed material in which a mark for alignment is not recorded in advance and a written printed material.

It is an external view of the hardware constitutions of one Embodiment of the difference image extraction apparatus which concerns on this invention. It is a figure explaining the outline | summary which extracts a difference image from the unprinted printed material of the said embodiment, and the printed material written. It is a figure which shows an example which extracts a template from the unwritten printed matter of the said embodiment. It is a figure explaining an example of position alignment with the printed matter in which the unwritten and printed matter of the said embodiment were written. It is a flowchart (the 1) explaining operation | movement of the said embodiment. It is a flowchart (the 2) explaining operation | movement of the said embodiment.

Explanation of symbols

101 Scanner 102 PC
103 display

Claims (7)

A differential image extraction device that extracts a difference between a printed matter at the end of writing and a printed matter that has been written,
Image acquisition means for acquiring images of two printed materials;
Image extracting means for extracting the character area of the last-printed printed matter;
An alignment unit that compares the extracted character area with the written image data of the printed material to align both printed materials;
A difference image extraction apparatus comprising: difference extraction means for extracting a difference image between two printed materials in the aligned state. The image extracting means includes
A binarization unit that converts the image of the last written print acquired by the image acquisition unit into image data binarized into a black pixel and a white pixel for each pixel;
A histogram generation unit that generates a histogram by projecting image data that has been converted into black pixels in the binarization unit in the vertical direction and the horizontal direction of the paper surface of the last written print;
A determination unit for determining a character area from the shape of the generated histogram;
The difference image extraction apparatus according to claim 1, further comprising: a template extraction unit that extracts image data of M rows and N columns as a template from the character region determined by the determination unit. The alignment means includes
An object binarization unit that converts the written image of the printed material into image data binarized into black pixels and white pixels for each pixel;
A matching region recognition unit for recognizing a region where the template extracted by the template extraction unit and the image data binarized by the target binarization unit match;
The difference image extracting apparatus according to claim 2, further comprising: an alignment unit configured to match the coordinate value of the area determined by the matching area recognition unit with the coordinate value of the template. The difference extraction means includes
An area extracting unit that extracts an area determined by adding a blank area that is a set of white pixels to an area determined as a character area of the last written printed matter, and a corresponding region of the written printed matter;
The difference image extraction apparatus according to claim 3, further comprising a calculation unit that calculates a difference between the extracted image data of both regions.   5. The difference image extraction apparatus according to claim 4, further comprising output means for outputting as a black pixel when the difference calculated by the calculation unit is not "0".   3. The difference image extracting apparatus according to claim 2, wherein the template extracting unit extracts the largest one of the plurality of M rows and N columns of image data when it can be extracted. A difference image extraction method of a difference image extraction device that extracts a difference between a printed matter at the end of writing and a printed matter that has been written,
An image acquisition step of acquiring images of two printed materials;
An image extraction step for extracting the character area of the last-printed printed matter;
An alignment step in which the extracted character area and the written image of the printed material are collated to align both printed materials;
A difference image extraction method comprising: a difference extraction step of extracting a difference image between two printed materials in the aligned state.

Images