JPH07129719A - Document picture input device - Google Patents

Abstract

PURPOSE:To improve the character recognition rate of a color document. CONSTITUTION:Image areas of the document picture read in as RGB (or CMY) components by a reader unit A are separated by a character recognition circuit 408, and document pictures (character pictures) of the color having the highest degree of character separation in areas judged to be respective character areas are compared and synthesized, and the density histogram of each character area to be synthesized is generated by a density histogram processing circuit 410. The synthesized document picture is subjected to character recognition by the circuit 408, and, if an unrecognizable character picture is found, the quantity of light of a lamp for document illumination is adaptively controlled in accordance with the state of the density histogram of the area, to which this character picture belongs, to obtain a recognizable character picture.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to control of character image input means in a character recognition device.

[0002]

2. Description of the Related Art Handwritten characters vary in character density depending on the writing tool used, the writing pressure, the material of the paper, and the like. Further, the density difference of output characters also occurs depending on the state of an output device such as a printer.

FIG. 11 is a block diagram of a conventional character recognition apparatus. When characters having different densities in the same original document are recognized by the character recognition apparatus as shown in FIG. About characters that have not reached
Occurs, and the recognition accuracy has deteriorated.

In order to solve this drawback, a method of re-recognizing a character after enlarging the line width of a character in which "blurring" has occurred, or detecting an end point of the character in which "blurring" has occurred, is detected. A method has been devised in which after the distance between the end points and the flow of a line segment near the end points are connected to each other, the characters are recognized again.

[0005]

However, according to the above method, a part which is not originally required to be crushed is partially crushed due to a partial character blurring, or a part which is not required to be crushed is connected, so that the number of character candidates is increased. There was a drawback that it could not be recognized.

In recent years, the demand for color originals has increased. Along with this, a character recognition device compatible with color originals has become necessary.

However, since the character recognition itself does not need to be in color, the conventional character recognition device does not perform character extraction corresponding to a color original. For this reason, there is a drawback in that not only the fading of characters is more likely to occur than in a black-and-white document, but also the character recognition rate is lowered if there is a character decoration with colors.

Therefore, an object of the present invention is to provide a document image input device that solves the above problems.

[0009]

In order to solve the above-mentioned problems, the present invention stores an image input means for inputting an original image for each of a plurality of color components and an original image input by the input means. 1 or a plurality of image storage means, an image area separation means for dividing a character area and a non-character area in a document image in the image storage means, a dictionary means for storing a character matching font, and the image storage means. A character recognition means for recognizing a character by extracting the characteristics of each character after cutting out one character from the character area in the image recognition means and a density histogram of the character area in the image storage means and for character recognition. When a density histogram processing means for calculating a threshold value and a character that cannot be recognized by the character recognition means are generated, the reading density of the original image of the image input means is changed. Characterized in that comprising further to means.

Further, according to the present invention, an image input means for inputting a manuscript image for each of a plurality of color components, a line memory for storing one line of each manuscript image of a plurality of color components input by the input means, A processing means for creating a density histogram for each color image stored in the line memory and detecting a high frequency component in the line direction to select a threshold for character recognition and a document image of a color for character recognition by the processing means. Image storage means for storing one screen of the original image of the selected color, image area separation means for dividing a character area and a non-character area from the image in the image storage means, and a character matching font are stored. And a character recognition means for recognizing characters by extracting the characteristics of each character after cutting out the character area from the character area in the image storage means and comparing with the dictionary means. And wherein the door.

[0011]

Embodiments of the present invention will now be described in detail with reference to the drawings.

<Embodiment 1> FIG. 1 is an external view of a copying apparatus according to an embodiment of the present invention. A reader unit A for reading an original image and a printer unit for forming the read image on a recording medium such as paper. It is composed of B and. Further, the reader unit A is provided with an operation section C shown in FIG.

The reader unit A and the printer unit B will be described with reference to FIG.

An image of a document is formed so that one or two documents stacked on the document feeding means 1 of the reader unit A can be fed or read continuously to a predetermined position on the platen glass surface 2. The surface (original surface) of the original platen glass 2
The document is placed face down and pressed against the document glass 2 by the document cover (document feeding means 1). The surface of the original is illuminated by a lamp 3, and the reflected light is focused on an image sensor 9 which is a sensor for image reading through mirrors 5 to 7 and a lens 8 and an image signal (image data) for one line in the main scanning direction is obtained. Enter. Then, the mirror 5 and the mirrors 6 and 7 are moved in the sub-scanning direction at a relative speed of 2: 1 to input the entire document surface.

The image signal input to the image sensor 9 is once stored in the image memory shown in FIG. 4 and subjected to processing using the image memory such as OCR processing.

The processed image data is sequentially read from the image memory, further subjected to the image processing selected by the operation section C, and supplied to the laser scanning optical system unit 10 of the printer unit B.

The laser scanning optical system unit 10 is composed of a semiconductor laser unit, a collimator lens, a rotary polyhedral mirror (polygon mirror), an fθ lens, a correction optical system and the like. The output light of the semiconductor laser is modulated by the image data input here, and the modulated laser light is applied to a polygon mirror rotating at high speed through a collimator lens, and the laser light reflected there is fθ lens. Photoconductor 1 according to the rotation of the polygon mirror via
1 is scanned to form an image (latent image).

The electrostatic latent image formed on the photoconductor 11 is 12
Alternatively, it is visualized with the developer (toner) supplied from the developing device 13.

The recording papers stacked and stored in the transfer paper stacking units 14 and 15 are sent to the registration roller position by the paper feed roller and the feed roller. Further, the recording paper is sent to the transfer position from the photoconductor 11 by the registration roller in timing with the image front end position of the image formed on the photoconductor 11.

After the toner image formed on the photoconductor 11 is transferred onto the recording sheet by the transfer / separation charger 16, the toner image is separated from the photoconductor 11 and fixed by the fixing device 17 via the conveying belt. .

The fixed recording paper is stacked on the paper output tray by the paper output roller 18 by the direction flapper 19 or is output.
In order to perform a multiplex / double-sided image forming process, the image is sent to an internal conveying unit to prepare for re-feeding.

FIG. 3 shows an example of the configuration of the operating section C. Operating part C
Is a copy start key 311 for instructing to start copying, a copy stop key 310 for instructing to stop the copying operation, and a reset key 30 for returning the copy mode to the standard state.
9, a set key group 313 including a ten key group from "0" to "9" and a "C" key for clearing the set number of copies, a copy number display portion 312, and a density for setting the density of the density Key 303 and its display unit 302, automatic density adjustment function selection key and its display 304, magnification selection key 306 and its display 305, paper feed tray and automatic paper selection function selection key 308, paper feed tray and paper size Display 307, OCR function selection key 315 and its display 314, and other independent key switches and display units.

The operation section C has an operation display section 301. The operation display section 301 is composed of a dot matrix liquid crystal display (LCD) and keys SW formed of transparent electrodes on the upper surface thereof and arranged in a matrix. , LCD
It is a touch panel type display / input means for directly touching (pressing) the content displayed on.

FIG. 4 is a block diagram of the control system of the copying machine according to the embodiment of the present invention. When the copy key of the operation unit 416 is pressed, the system controller 415 activates the reader unit A and the operation of reading the original image.

An original image illuminated by a lamp whose light amount is controlled by a light control circuit 401 is input to an image sensor 402 via a lens and converted into three primary colors, that is, R, G, and B electrical signals. . The original image converted into an electric signal is converted into a digital signal by an A / D converter 404 via a preamplifier 403, and a shading correction circuit 405.
The shading of the image signal is corrected by. The original image signal is stored in the R, G, and B image memories 407. These operations are timed by the output pulse of the timing pulse generation circuit 406 controlled by the system controller 415.

When the original image is stored in the image memory 407, the system controller 415 activates the character recognition circuit 408 and is stored in the image memory 407 for each of RGB by the image area division processing which is one of the functions of the character recognition circuit 408. Divide the existing document image into text and non-text areas.

The system controller 415 compares the areas divided by the image area division processing performed on the RGB image memories 407, and selects which image memory 407 of the original image is to be used for each divided character area.
(A selection method is based on a majority decision that gives priority to character area determination.) Then, for example, each selected image area is combined in the G image memory 407, and each area selected by the density histogram processing circuit 410 is combined. A density histogram is created for each.

The system controller 415 activates the character recognition circuit 408 again, cuts out the character images one by one from the combined image in the G image memory 407, compares the character images with the dictionary 409, and replaces the character images with the corresponding character font images.

When all the character images of the original image have been replaced, if there are unrecognizable characters due to blurring or the like, the density histogram for the area where the unrecognizable characters exist is checked, and the light intensity of the lamp is sent to the dimming circuit 401 according to the distribution. To reduce the number of scans, and then try scanning again. If it cannot be confirmed by the second trial, the portion is left as it is and the process proceeds.

The system controller 415 is R: 0.3, G: 0.59, B: 0.11 in a normal copy operation.
When the OCR function is selected, the matrix coefficient of the active matrix 411 (standard value) is R: 0, G:
It is switched to 1, B: 0 and the character color is printed out as black. (The G coefficient is set to 1 because the replaced character font is stored in the G image memory.) The character image that has passed through the active matrix 411 is processed by the density conversion circuit 412 in the brightness density. Image conversion circuit 413
Then, various images set by the operation unit C are applied to the character image, and the character image is sent to the image recording unit 414 (printer unit B). The image recording unit 414 modulates the laser light with the character image and records it on a recording sheet.

The above operation will be further described with reference to the operation flowchart of the system controller 415 shown in FIG.

In the operation standby state, the matrix coefficient of the active matrix 411 is set to the standard value (S50
1) Wait for copying to start (S50)
2).

When the copy start key of the operation section C is pressed, it is checked whether the OCR function is turned on (S503).
Start the reader unit A.

When the OCR function is turned on, the original image read from the reader unit A is stored in the image memory 407 as it is in the three primary colors (S504), and the character recognition circuit 408 is used to display the character area and other areas. It is divided (S505).

The areas divided by the image area division processing performed on the image memories 407 for each of RGB are compared and the original image in the image memory 407 to be used is selected for each divided character area. (The selection uses a selection method based on a majority decision giving priority to the character area determination.) (S506) The density histogram processing circuit 410 for each selected area.
Is used to create a density histogram, obtain a threshold value for character recognition, and notify the character recognition circuit 408 (S507). Then, the image areas selected in the predetermined image memory 407 (for example, G image memory) are combined (S508).

Then, the character recognition circuit 408 is activated again, character images are cut out one by one from the combined image on the image memory 407, and the characteristics of the character images are extracted and compared with the contents stored in the dictionary 409. For a character that can be determined by expanding the character font from the dictionary 409 on the image memory 407, if there are several character candidates, the corresponding character code is stored and the plurality of character codes and the dictionary 4 are stored.
09, the character is determined and the character font is expanded in the image memory 407. If it is not confirmed even by this, a flag indicating that a faint character has occurred is set and a code indicating the area is stored. (When synthesizing image areas, the code indicating the area is the system controller 415
Allocated by. ) (S509) When the replacement of all the character images of the original image is completed,
When there is an unrecognizable character due to faintness (S51
0) The density histogram for the area where the unrecognizable character exists is examined, and the dimming circuit 401 is instructed to reduce the lamp light amount according to the distribution (S512), and the document reading is performed again.

If it cannot be confirmed by the second trial (S511), that portion is processed as it is as the original character image.

Then, after changing the matrix coefficient of the active matrix 411 for OCR (S51
3), each print operation unit is activated to print out (S515).

In the above embodiment, when it is determined that a blur has occurred, the light amount of the illumination lamp is changed based on the result of the density histogram of the blur portion, but the gain of the preamplifier 403 may be changed. . Further, if the gain of the preamplifier 403 can be set for each of the RGB colors, the color can be selected at the time of reading. The same operation can be performed even if the active matrix 411 is provided in the front stage or the rear stage of the image processing circuit 413. However, in that case, three density conversion circuits 412 are required.

The character recognition dictionary 409 used in the above embodiment may be a removable storage medium, and one or more storage media can be connected.

<Second Embodiment> FIG. 6 is a block diagram of a copying machine control system according to a second embodiment of the present invention. Operation unit 61
When the copy key 8 is pressed, the system controller 61
Reference numeral 7 activates the reader unit A to activate the reading operation of the original image.

A document image illuminated by a lamp whose light amount is controlled by a light control circuit 601 is input to an image sensor 602 via a lens and converted into RGB electrical signals. The original image converted into an electric signal is converted into a digital signal by an A / D converter 604 via a preamplifier 603, and after shading of the image signal is corrected by a shading correction circuit 605, it is stored in each RGB line memory 606. It

A density histogram for each line is created in the density histogram processing circuit 608 for the image stored in the line memory 606, and edge detection in the line direction is performed. It is determined that the edge is sharp among the three RGB lines. The coefficient of the active matrix 609 is changed so that the stored color signal can be stored in the image memory 610 from the line memory 606, and the line memory 606 is sequentially changed.
To the image memory 6 via the active matrix 609.
Store in 10.

The matrix coefficients of the active matrix 609 are R: 0.3, G: 0.59, B: 0.11 (standard value). However, when the OCR function is selected, the density processing circuit 608 makes a determination of RGB. One of the coefficients is a through 1 and the other two are off.

When all the image signals have been stored in the image memory 610, the system controller 617 activates the character recognition circuit 611 and stores them in the image memory 610 by the image area division processing which is one of the functions of the character recognition circuit 611. The original image is divided into a character area and a non-character area.

Further, the system controller 617 activates the character recognition circuit 611 again, cuts out a character image for each character from the image of the area determined to be the character area on the image memory 610, compares it with the dictionary 612, and compares it with the corresponding character font. Replace text image with image.

When all the character images of the original image have been replaced, if there are unrecognizable characters due to blurring, the density histogram of the line where the unrecognizable characters are present is checked, and the dimming circuit 601 is ramped according to the distribution. Instruct to reduce the light intensity, and try again from the original capture. If it cannot be confirmed by the second trial, the portion is left as it is and the process proceeds.

When the OCR function is selected, the switch 613 is switched to the b side, and the contents stored in the image memory 610 can be output to the printer unit B.

A character image sequentially read from the image memory 610 is subjected to brightness / density conversion by a density conversion circuit 614, various image set by the operation unit C is applied to the character image by an image processing circuit 615, and an image recording unit 616. (Printer unit B)
Send the character image to. The image recording unit 616 modulates the laser light with the character image and records it on a recording sheet.

The above operation will be further described with reference to operation flowcharts of the system controller 617 shown in FIGS.

In the operation standby state, the matrix coefficient of the active matrix 609 is set to the standard value (S70
1) Waiting for copying to start (S70)
2).

When the copy start key of the operation section C is pressed, it is checked whether the OCR function is turned on (S703),
The reader unit A is activated and the number of lines of the input image is set (S704).

The original image for one line read from the reader unit A is stored in the line memory 606 after being subjected to shading correction (S705).

The density histogram processing circuit 608 is activated to sequentially read the image data stored in the line memory 606 to create a density histogram for each line (S706) and to detect an edge in the line direction (S707), and RGB3 An active matrix 60 is provided so that a color signal whose edge is determined to be sharp in a line can be stored in the image memory 610 from the line memory 606.
The coefficient of 9 is changed (S708), and the line memory 6
The data is stored in the image memory 610 from 06 via the active matrix 609 (S709).

When all the image signals have been stored in the image memory 610 (S710), the character recognition circuit 611 is activated and stored in the image memory 610 by the image area division processing which is one of the functions of the character recognition circuit 611. After dividing the original image into a character region and a region other than the character region, a character image is cut out one by one from the image of the region determined to be the character region (S711), and the corresponding character is selected by comparing with the dictionary 612 ( In step S712, the character image cut out with the character font image is replaced (S713).

If there are several character candidates, the corresponding character code is stored and a plurality of character codes are compared with the dictionary 612 to determine the character and the character font and the cut out character image are replaced. If it is not confirmed even by this, a flag indicating that a faint character has occurred is set and a code indicating the area is stored. (When synthesizing image areas, a code indicating the area is assigned by the system controller 617.) When the replacement of all the character images of the original image is completed,
When there is an unrecognizable character due to faintness (S71
4) The density histogram for the area where the unrecognizable character is present is examined, and the light control circuit 601 is instructed to reduce the lamp light quantity according to the distribution (S716), and the original reading is performed again.

If it cannot be confirmed by the second trial (S715), that portion is processed as it is as the original character image.

Then, each print operation unit is activated to print out (S718).

In the above embodiment, when it is determined that the blur has occurred, the light amount of the illumination lamp is changed based on the result of the density histogram of the blur portion, but the gain of the preamplifier 603 may be changed. .

The present invention can also be implemented by the operation shown in the flowcharts of FIGS. 9 and 10 as the operation of the system controller 617 in the block diagram of FIG. This will be described below with reference to the flowcharts of FIGS. 9 and 10.

In the operation standby state, the matrix coefficient of the active matrix 609 is set to the standard value (S80
1) Wait for copying to start (S80)
2).

When the copy start key of the operation section C is pressed, it is checked whether the OCR function is turned on (S803),
The reader unit A is activated and the number of lines of the input image is set (S804).

Then, the light quantity of the lamp is set to the standard light quantity (S805).

The original image for one line read from the reader unit A is shading-corrected and then stored in the line memory 606 (S806).

The density histogram processing circuit 608 is activated to sequentially read the image data stored in the line memory 606 to create a density histogram for each line (S807) and to detect an edge in the line direction (S808). Then, if the sharpness of the color signal whose edges are determined to be sharp in the RGB3 lines has not reached a predetermined level (S809), it is determined that unrecognizable characters due to blurring may occur, and the density histogram is checked. The light control circuit 601 is instructed to reduce the lamp light amount according to the distribution (S811), and the document reading process is repeated (S806).

If it cannot be confirmed even by the second trial (S810), the portion is left as it is and the original character image is processed.

Then, the active matrix 60 can be stored in the image memory 610 from the line memory 606.
The coefficient of 9 is changed (S812), and the line memory 6
The data is stored in the image memory 610 from 06 via the active matrix 609 (S813).

When all the image signals are stored in the image memory 610 (S814), the character recognition circuit 611 is activated and stored in the image memory 610 by the image area division processing which is one of the functions of the character recognition circuit 611. After dividing the original image into a character region and a region other than the character region, a character image is cut out one by one from the image of the region determined to be the character region (S815), and the corresponding character is selected by comparing with the dictionary 612 ( (S816), the character image cut out by the character font image is replaced (S817).

If there are several character candidates, the corresponding character code is stored and a plurality of character codes are compared with the dictionary 612 to determine the character and the character font and the cut out character image are replaced. If it is not confirmed even by this, a flag indicating that a faint character has occurred is set and a code indicating the area is stored. (When the image areas are combined, a code indicating the area is assigned by the system controller 617.) After the replacement of all the character images of the original image is completed, each print operation unit is activated to print out (S819). .

In the above example, when it is determined that a blur is likely to occur, the light quantity of the illumination lamp is changed from the result of the density histogram of that line.
The gain of 03 may be changed.

[0071]

As described above, according to the present invention, it is possible to adaptively select a document color component for character recognition of a color document and to perform character recognition at a high rate.

[Brief description of drawings]

FIG. 1 is an external view of a copying apparatus embodying the present invention.

FIG. 2 is an internal sectional view of a copying apparatus embodying the present invention.

FIG. 3 is a diagram showing an operation unit of a copying apparatus embodying the present invention.

FIG. 4 is a circuit block diagram of a first embodiment of the present invention.

FIG. 5 is a flowchart of the same operation.

FIG. 6 is a circuit block diagram of a second embodiment of the present invention.

FIG. 7 is a diagram showing a part of the same operation flowchart.

FIG. 8 is a remaining operation flowchart.

FIG. 9 is a diagram showing a part of another operation flowchart.

FIG. 10 is a remaining operation flowchart.

FIG. 11 is a conventional circuit block diagram.

[Explanation of symbols]

 401 Light control circuit 402 Image sensor 404 A / D converter 407 Image memory 408 Character recognition circuit 409 Dictionary 410 Density histogram processing circuit 411 Active matrix 415 System controller

Front page continuation (72) Inventor Masahiro Serizawa 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Inventor Tokuharu Kaneko 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Hirohiko Kishimoto 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Inventor Masanori Hira 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Invention Akihiko Sato 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Inventor Noriaki Matsui 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Yoshio Mizuno Inventor 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Inventor Yasushi Fukada 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Yoshihiro Hosomi Ota-ku, Tokyo Shimomaruko 3-30-2 Canon Inc.

Claims (12)

[Claims] 1. An image input unit for inputting a document image for each of a plurality of color components, one or a plurality of image storage units for storing a document image input by the input unit, and a document in the image storage unit. Image area separating means for dividing a character area and a non-character area in an image, a dictionary means for storing a font for character matching, a character area is extracted from the character area in the image storing means, and the characteristics thereof are extracted. Character recognition means for recognizing characters in comparison with the dictionary means, density histogram processing means for creating a density histogram of a character area in the image storage means and calculating a threshold value for character recognition, and the character recognition means A document image input device comprising means for changing the reading density of the document image of the image inputting means when an unrecognizable character occurs. . 2. The document image input device according to claim 1, wherein the reading density changing unit is a light adjusting unit of a document illumination light source. 3. The original image input device according to claim 1, wherein the reading density changing means is an amplifying means of an image sensor output. 4. The document image input device according to claim 1, wherein the dictionary means is removable. 5. The manuscript image input device according to claim 1, wherein after the character recognition means recognizes a character, the character image of the manuscript is replaced with a font in the dictionary means. 6. An image input unit for inputting an original image for each of a plurality of color components, a line memory for storing one line of each original image of a plurality of color components input by the input unit, and the line memory. A processing means for creating a density histogram for each color image stored in the memory and detecting a high-frequency component in the line direction to select a threshold for character recognition and a document image of a color for character recognition, selected by the processing means. Image storage means for storing one color original image in one screen, image area separation means for dividing a character area and a non-character area from the image in the image storage means, and a dictionary storing character matching fonts. Means and character recognition means for recognizing characters by extracting the characteristics of each character after cutting out one character from the character area in the image storage means and comparing with the dictionary means. Document image input device. 7. The manuscript image input device according to claim 6, further comprising means for changing a reading density of an input manuscript image when a character that cannot be recognized by the character recognition means is generated. 8. The document image input device according to claim 7, wherein the reading density changing unit is a light adjusting unit of a document illumination light source. 9. The document image input device according to claim 7, wherein the reading density changing unit is an image sensor output amplifying unit. 10. The document image input device according to claim 6, wherein the dictionary means is removable. 11. The document image input device according to claim 6, wherein the character recognition unit replaces the character image of the document with the font of the dictionary unit after recognizing the character. 12. An original image input device comprising: means for switching between a normal input mode and a character recognition mode of an original image and a display means for the means.