JPH06251199A - Character recognition pre-processing device - Google Patents

Abstract

PURPOSE:To reduce burden on a central processing unit and to perform character recognition pre-processing at high speed by employing such constitution that the control of the character recognition pre-processing is performed by the central processing unit that is a first processor, and the execution of the character recognition pre-processing by an image processing device that is a second processor. CONSTITUTION:This device is constituted of the central processing unit 1 as the first processor, the image processing device 2 as the second processor, an image acquisition device 3 comprised of the reading part of an OCR, a memory device 4, and a bus line 5. The central processing unit 1 instructs the correction of inclination of an image stored in the memory device 4 and the segmentation of a character in a character area after inclination correction. The image processing device 2 corrects the inclination of the image on the memory device 4 based on such instruction, and also, segments the character stored at an instructed position on the memory device 4. Also, the image processing device 2 corrects the inclination of the image stored in the memory device 4 by affine transformation based on that of space in reading.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition preprocessing device used in an optical character reader (OCR) or the like, and more particularly to a character recognition preprocessing device for performing preprocessing for cutting out a character from an image. It is about.

[0002]

2. Description of the Related Art Conventionally, OCR or the like has been used to convert characters on paper such as handwritten characters or printed characters into corresponding code information.

However, in the process of generating a handwritten character or a printed character, the character is not always ideally generated, for example, the printed character has shading.

Therefore, in the OCR, if a multi-valued image (hereinafter referred to as an image) containing characters on a paper surface is acquired, it is designated after removing noise and the like from the image so that character recognition can be performed accurately. Pre-processing for cutting out characters one by one from the character area is performed.

This pre-processing includes: (1) Density correction: adjusting the shade of characters on the printed paper. (2) Filter: For example, the image has a portion that is not a character due to ink droplets during printing, that is, a so-called noise portion, and should be removed. (3) Binarization: Quantizing an image obtained with multiple gradations into white and black with an appropriate threshold value. (4) Clipping: Outputting one character at a time from a binarized character pattern or character area. There is.

FIG. 7 is a block diagram showing an example of a conventional apparatus for performing preprocessing.

In FIG. 7, an image acquisition device (eg,
The image input from the OCR reading unit) to the temporary storage device 21 is directly input to the processing device 22 that performs density correction processing, filtering processing, binarization processing, etc., and the image after performing these processing is stored. The image is stored in the device 23, and then the image stored in the storage device 23 is read by the central processing unit 24 through the bus line 20 to perform a cutting process for cutting out characters to be recognized one by one.

FIG. 8 is a block diagram showing another example of a conventional device for performing preprocessing, which comprises a memory device 25, a processing device 26, and a central processing device 27 connected to these by a bus line 20. There is.

FIG. 9 is a flow chart showing the preprocessing procedure in the configuration of FIG. 8. First, the input image is once stored in the storage device 25 (step 91).

The central processing unit 27 corrects the inclination and displacement of the characters in the stored image, and the stored image is stored in the storage unit 25.
Are again stored in (step 92). After that, density correction processing, filter processing, and binarization processing are performed (step 9).
3) After that, the central processing unit 27 performs the cutting process (step 94). This series of processing is performed for all characters (step 95).

By the way, when a form or the like is input to the OCR and characters are recognized, the form is input at the input stage while being inclined from the reference position shown by the solid line to the position shown by the broken line.
As a result, the read image may be skewed.

In an image tilted due to such a cause, the character feature cannot be correctly extracted in a subsequent process unless the tilt is corrected.

Therefore, as a method of correcting the inclination of the image caused by the inclination of the form, for example, in the preprocessing apparatus having the configuration of FIG. 7, the temporary storage device 21 and the processing device 22 are used.
A delay circuit is provided between the first and second scanning lines to shift the n-th scanning line portion from the top of the image and the (n + 1) th scanning line portion from the top of the image to correct the inclination. On the other hand, density correction processing and the like are performed.

Further, in the preprocessing apparatus having the configuration of FIG. 8, the processing for correcting the inclination is performed by the central processing unit 27.

At this time, the degree of inclination correction varies depending on which scanning line is displaced or how many scanning lines are displaced at a ratio of one scanning line.

[0016]

[Problems to be solved by the invention] However, originally,
The correction of the inclination of the image should not be performed by the parallel conversion from the parallelogram to the rectangle as in the correction method of the inclination shown in FIG. 10, but should be corrected by the rotation conversion.

Therefore, the inclination correction method shown in FIG. 10 has a problem that the character feature cannot be correctly extracted in the subsequent process.

Further, in the pre-processing apparatus having the configuration of FIG. 8, the inclination correction by the rotation conversion is performed by the CPU 27, but this causes a problem that the load of the CPU 27 increases, and as a result, the character This means that the pre-recognition process cannot be speeded up.

An object of the present invention is to provide a character recognition preprocessing device which can reduce the load on the central processing unit and can perform the character recognition preprocessing at high speed.

Another object of the present invention is to provide a character recognition preprocessing device which can correct the inclination of an image correctly.

[0021]

In order to achieve the above object, the present invention detects an inclination of a paper surface with respect to a reference position at the time of reading an image including a character stored in a storage device, and preprocesses the object. A first processing unit (central processing unit) that designates an image region and further designates a character cutting position in the region, and stores it based on the information of the inclination of the paper notified from the first processing unit. The inclination of the image stored in the apparatus is corrected, and a character area is detected from the preprocessing target area notified from the first processing apparatus for the image after the inclination correction, and the first processing apparatus is detected. And a second processing device (image processing device) for notifying and notifying the character at the cutting position instructed by the first processing device in response to the notification.

In the second processing device, the inclination of the image stored in the storage device is corrected by affine-transforming the storage position of the image based on the inclination of the paper surface at the time of reading. Is.

[0023]

According to the above-mentioned means, the first processing device is different from the second processing device in that the inclination of the image stored in the storage device is corrected and the characters in the character area of the image after the inclination correction are corrected. Instruct to cut out.

The second processing device corrects the inclination of the image on the storage device based on the instruction of the second processing device, and cuts out the character stored at the instructed position on the storage device. .

The second processing device corrects the inclination of the image stored in the storage device by affine transformation based on the inclination of the paper surface at the time of reading.

[0026]

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

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

In FIG. 1, 1 is a central processing unit (hereinafter referred to as CPU) as a first processing unit, and 2 is an image processing unit as a second processing unit.

Reference numeral 3 is an image acquisition device composed of, for example, an OCR reading unit, and 4 is a storage device for storing an image.

Reference numeral 20 is a bus line for transmitting an instruction issued by the CPU 1 to the image processing apparatus 2.

FIG. 2 is a block diagram showing the detailed arrangement of the image processing apparatus 2 in this embodiment. The image processing apparatus control unit 11, the image acquisition unit 12, the storage position calculation unit 13, the image processing unit 14 and the storage unit. It is configured by the device control unit 15.

In FIG. 1, the CPU 1 is an image processing apparatus 2
A command for image acquisition, density correction processing, filter processing, binarization processing, cutout processing, and the like is issued.

The image processing device 2 has an interface with three devices of the image acquisition device 3, the storage device 4 and the CPU 1, and in accordance with the instruction of the CPU 1, image acquisition, form inclination correction, density correction processing, filter processing and binary. The conversion processing and the cutout processing are executed.

FIG. 3 is a flow chart showing the operation procedure of this embodiment.

The operation of this embodiment will be described with reference to FIG.

First, the CPU 1 issues an image acquisition command to the image processing apparatus 2 (step 31). Then, the image processing device 2 stores the image sent from the image acquisition device 3 in the storage device 4 via the image acquisition unit 12 and the storage device control unit 15 according to this command, and at the same time, stores the acquired image. The area on the paper surface and the boundary other than the paper surface are detected, and the detection result is sent to the CPU 1 (step 32).

Next, the CPU 1 calculates the position and the inclination (see (x, y) and θ in FIG. 4) of the paper surface on the storage device 4 from the boundary position of the paper surface that has been sent, and this is imaged. The processor 2 is notified (step 33).

When the image processing apparatus 2 finishes storing one sheet of the image in the storage device 4, the image processing apparatus 2 finishes the image acquisition processing by the CPU.
1 is notified (step 34).

Next, when the CPU 1 receives the notification of the end of the image acquisition processing, the position on the paper of the area where the density correction processing, the filtering processing and the binarization processing are performed ((xl0, yl in FIG. 4).
0)) is decided, this area is notified to the image processing apparatus 2 (step 35), and the execution of processing for this area is subsequently instructed (step 36).

The storage position calculation unit 13 executes C in step 33.
Affine transformation is performed based on the position and the inclination of the paper surface on the storage device 4 notified from PU1, and the storage position on the storage device 4 of the pixel from the position on the paper surface notified in step 35 (see FIG. 4). (Xp0, yp0)) is calculated (step 37).

The image processing unit 14 follows the instructions of the density correction process, the filter process, and the binarization process in step 36.
For the pixel obtained in step 37, density correction processing,
When the filter process and the binarization process are executed and a series of processes is completed, the CPU 1 is notified of the end of the density correction process, the filter process, and the binarization process. At the same time, the distribution of the black pixels forming the character is sent to the CPU 1 (step 38).

The CPU 1 estimates the character area on the storage device 4 from the distribution of the sent black pixels, determines the cutout position, and notifies the image processing device 2 (step 39).

After receiving the notification of the end of the density correction processing, the filtering processing, and the binarization processing, the CPU 1 issues a cutout processing command designating the cutout position on the storage device 4 (step 40).

The image processing device 2 cuts out the binarized character stored in the character area on the storage device 4 designated by the cutout processing command and sends it to the CPU 1. When the cutout process is completed, the CPU 1 is notified of the end of the cutout process (step 41).

The CPU 1 repeats the issuance of the cutout processing command until there are no more images to be preprocessed.

When the image processing apparatus 2 completes the processing for all the character areas (step 42),
The CPU 1 issues a next paper image acquisition command to the image processing apparatus 2 and returns to step 31.

FIG. 4 is an explanatory diagram of a method for correcting the inclination of an image by the affine transformation used in the present invention.

In FIG. 4, how much the form in which the handwritten character or the printed character is generated is input to the image acquisition device 3 by inclining (rotating) is determined from the boundary of the paper surface (continuous black dots (outside the form). It is obtained by detecting continuous white dots (points that change into a form).

Therefore, the coordinates on the storage device 4 are (x,
If the CPU 1 calculates that the document is input by parallel translation by y) and rotation by the angle θ, the CPU 1 uses this rotation angle θ before causing the image processing apparatus 2 to perform the tilt correction processing. , Cos θ, sin θ are calculated,
The image processing apparatus 2 is notified of the calculation result and the parallel movement amount (x, y).

After that, the CPU 1 causes the image processing apparatus 2 to set an area for density correction processing, filter processing, and binarization processing to the coordinates (x1, x1) of the starting point (usually the upper left point) on the form.
y1) and the size of the form (xsize, ysize)
Specify with.

The image processing apparatus 2 sets the coordinates (x1, y1) of the starting point of the given form to

[0052]

[Equation 1]

As described above, the affine transformation (rotation and parallel movement) is executed by the storage position calculation unit 13 to obtain the coordinates (x1, y1) of the starting point of the form on the storage device 4.

The adjacent point is (x1 + 1, y1), and this is replaced with the portion (x1, y1) of the equation (1) to perform the above-mentioned processing, whereby the coordinate of the adjacent point on the storage device 4 is converted. Is obtained.

The coordinates of the point on the storage device 4 thus obtained are (xp + cos θ, yp-sin θ).

Further, the lower point is (xl, yl + 1), which is replaced with the portion (x1, y1) of the equation (1) and the above processing is performed, so that the next point on the storage device 4 is changed. The coordinates of are obtained.

The coordinates of the point on the storage device 4 thus obtained are (xp + sin θ, yp + cos θ).

As a result, the affine transformation is performed only on the starting point, and from the next point, the coordinates of all points on the storage device 4 can be obtained by adding (cos θ, −sin θ) or (sin θ, cos θ).

Therefore, the area formed by the points thus obtained is an image area without inclination. Therefore, an image without inclination can be obtained by reading the value of each pixel in this region. That is, it is possible to obtain an image in which the tilt is corrected.

As described above, the CPU 1 needs to recognize how the image of the form is stored in the storage device 4 only by first giving the rotation amount and the parallel movement amount to the image processing device 2. There is no.

FIG. 5 is a block diagram showing the configuration of another embodiment of the present invention. In this embodiment, the storage device 4 is provided with two areas A5 and B6, and when image acquisition processing is being performed in one area, density correction processing, filter processing, binarization processing, and clipping are performed in the other area. The feature is that the processes are executed simultaneously. The other structure is similar to that of the previous embodiment.

FIG. 6 is a flow chart showing the operation procedure of the embodiment in the configuration of FIG.

First, the CPU 1 issues an image acquisition command to one area of the storage device 4, for example, the area A5, to the image processing apparatus 2 (step 61).

While the image processing apparatus 2 stores the image sent from the image acquisition apparatus 3 in the area A5 of the storage device 4 via the image acquisition unit 12 and the storage device control unit 15, among the acquired images, it is on the paper surface. The boundary between the area (1) and the area other than the paper surface is detected, and the detection result is sent to the CPU 1 (step 62).

Next, the CPU 1 calculates the position and inclination of the paper surface on the storage device 4 from the boundary position of the paper surface that has been sent, and notifies the image processing device 2 of this (step 6).
3).

When the image processing apparatus 2 finishes storing one sheet of the image in the area A5 of the storage device 4, it notifies the CPU 1 of the end of the image acquisition processing (step 64).

Further, in parallel with the issuance of the above-mentioned image acquisition command, the CPU 1 performs density correction processing and filter if there is an acquired image in the other area of the storage device 4 of the image processing apparatus 2, for example, the area B6. The area to be processed and binarized is determined at a position on the paper surface, the image processing apparatus 2 is notified of this (step 65), and the execution of the processing is instructed (step 66).

In accordance with the instruction of step 65, the image processing apparatus 2 performs affine transformation in the storage position calculation unit 13 using the position and the inclination of the paper surface on the storage device 4, and from the position on the paper surface of the preprocessing target area, The storage position on the storage device 4 is obtained (step 67).

Next, the image processing apparatus 2 follows the instructions of the density correction processing, the filter processing, and the binarization processing in step 66.
The image processing unit 14 performs density correction processing, filter processing, and binarization processing on the image of the preprocessing target area.
When the process is completed, the CPU 1 is notified of the end of the density correction process, the filter process, and the binarization process. At the same time, the distribution of black pixels forming the character is sent to the CPU 1 (step 68).

The CPU 1 estimates the character area from the distribution of the black pixels sent, determines the cut-out position, and notifies the image processing apparatus 2 (step 69).

Next, in step 68, the CPU 1
Density correction processing, filter processing, 2 from the image processing apparatus 2
When the notification of the end of the value conversion process is received, the cutout processing command designating the cutout position is issued (step 70).

Then, the image processing apparatus 2 cuts out the binarized character in the designated area and sends it to the CPU 1 in accordance with the cutout processing command. When the cutout process is completed, the CPU 1 is notified of the end of the cutout process (step 71).

The CPU 1 repeatedly commands the image processing apparatus 2 to perform the density correction processing, the filtering processing, the binarization processing, and the clipping processing until the area to be processed is exhausted. (Step 72).

When the CPU 1 receives the notification of the completion of all the cutout processing, it switches the area of the storage device 4 to the image processing apparatus 2, that is, an image acquisition command for the area B6 and a density for the area A5. Notifications and commands for correction processing, filter processing, binarization processing, and clipping processing are issued (step 73).

Therefore, in this embodiment, it is possible to simultaneously perform the pretreatment for two sheets of paper.

In each of the above-described embodiments, an example in which the image acquisition device is composed of the reading unit of the OCR has been described, but it can be applied to the case of acquiring an image from a reading device such as an image scanner. .

[0077]

As described above, according to the present invention, with respect to an image containing characters stored in a storage device, the inclination of the paper surface with respect to the reference position at the time of reading is detected and the image area to be preprocessed is designated. A first processing device for instructing a character cutting position in the area;
Of the image stored in the storage device on the basis of the information on the inclination of the paper notified from the processing device, and the pre-processing notified from the first processing device to the image after the inclination correction. The first character area is detected by detecting a character area from the target area.
And a second processing device that cuts out the character at the cutout position instructed by the first processing device in response to the notification, and the control of the character recognition preprocessing is performed by the first processing device. The central processing unit which is the second processing unit executes the character recognition preprocessing, and the image processing apparatus which is the second processing unit executes the character recognition preprocessing. Therefore, the load of the central processing unit as the first processing unit is reduced.

Therefore, there is an effect that the character recognition preprocessing can be speeded up by the amount of reduction of the load, and it can be applied to control of other processing to improve the processing speed and processing accuracy of the preprocessing. .

The image processing apparatus, which is the second processing apparatus, rotates and parallel coordinates the inclination of the image stored in the storage device, based on the inclination of the paper surface at the time of reading, the original storage position. Since the correction is performed by (affine transformation), and the storage position without inclination is calculated and read out, there is an effect that the inclined image can be corrected correctly and the character feature can be accurately extracted in the subsequent process.

[Brief description of drawings]

FIG. 1 is a block configuration diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a block configuration diagram showing details of an image processing apparatus in the embodiment of FIG.

FIG. 3 is a flowchart showing an operation procedure of the embodiment shown in FIG.

FIG. 4 is an explanatory diagram of affine transformation applied to the embodiment of FIG.

FIG. 5 is a block diagram showing the configuration of another embodiment of the present invention.

6 is a flowchart showing an operation procedure of the embodiment of FIG.

FIG. 7 is a block diagram showing an example of a conventional pretreatment device.

FIG. 8 is a block diagram showing another example of a conventional pretreatment device.

9 is a flowchart showing an operation of the pretreatment device in the configuration of FIG.

FIG. 10 is an explanatory diagram of a conventional method for correcting the inclination of a character.

[Explanation of Codes] 1 ... Central processing unit (first processing unit), 2 ... Image processing unit (second processing unit), 3 ... Image acquisition unit, 4 ... Storage unit, 5 ... Region A of storage unit 4 , 6 ... Area B of the storage device 4, 11 ... Image processing device control unit, 12 ... Image image acquisition unit,
13 ... Storage position calculation unit, 14 ... Image processing unit, 15 ... Storage device control unit, 20 ... Bus line.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yasuo Sanbe 3-3-3 Toyosu, Koto-ku, Tokyo NTT Data Communications Corporation

Claims (2)

[Claims] 1. With respect to an image containing characters stored in a storage device, an inclination of a paper surface with respect to a reference position at the time of reading is detected, a preprocessing target image area is designated, and a character cutting position in the area is detected. The first processing device that gives an instruction and the tilt of the image stored in the storage device are corrected based on the information of the tilt of the paper notified from the first processing device. A character region is detected from the pre-processing target region notified from the first processing device, the first processing device is notified of the character region, and the cut-out position specified by the first processing device in response to the notification is detected. A character recognition preprocessing device comprising a second processing device for cutting out a character. 2. The character recognition preprocessing device according to claim 1, wherein the inclination of the image stored in the storage device is affine-transformed and corrected based on the inclination of the paper surface at the time of reading the storage position of the image. .