JP2001101342A - Character recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent occurrence of erroneous recognition, etc., as the result of overlapping of a stepping position and a character pattern at the time of correcting the inclination of an original in a character recognizing device. SOLUTION: This character recognizing device obtains the inclined height H4, of a reading area 301 first and further, equally divides an area width by the number of the pixels of the inclined height H4 to decide temporary stepping positions 311 to 313. Next, the vertical frame positions 321 to 323 of an image frame are judged through the use of format information and the nearest ones to the respective positions 311 to 313 are selected from these positions 321 to 323 to make stepping positions 331 to 333. Continually, reading is executed in a width direction while deviating coordinates pixel by pixel in a vertical direction at the positions 331 to 333, thereby an image 340 whose inclination is corrected is obtained. After then, a character pattern 352 is segmented from the area 351 of the image 340 to execute character recognition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device, and more particularly, to a character recognition device having a tilt correction function.

[0002]

2. Description of the Related Art FIG. 1 shows a configuration example of a conventional character recognition apparatus.
0 is shown.

FIG. 10A is a conceptual diagram showing a system configuration of a conventional character recognition device.

In this character recognition system, an operation unit 1
010, for example, input means 1011 such as a keyboard;
Display means 1012 such as a CRT (Cathode Ray Tube)
And an image acquisition unit 1013 using, for example, a general-purpose scanner or the like. In addition, CPU (Central
A ssing unit 1020 receives image information from the operation unit 1020 and performs character recognition and the like.

FIG. 10B is a block diagram schematically showing the internal configuration of CPU 1020.

In the CPU 1020, the control unit 10
21 communicates with the operation unit 1010, and
The entire U1020 is controlled. Image buffer 102
Reference numeral 2 stores image information captured from the operation unit 1010 via the control unit 1021. Format buffer 1
Reference numeral 023 stores the format information of the image. The character cutout unit 1024 detects a character position and cuts out a character pattern for one character. The character recognition unit 1025 recognizes a character pattern and outputs a character code as a recognition result.

In such a character recognition device, a reading operation is performed as follows.

The operation unit 1010 stores information for specifying a reading area and format information in advance. As shown in FIG. 11A, the reading area 1101 includes a left margin W ₁ and a top margin H of the document 1102.
₁ , the region width W ₂ and the region height H ₂ . As shown in FIG. 11B, the format information has a character height H ₃ , a character pitch P, and the number of characters N (N = 4 in FIG. 11B).

The operation unit 1010 is provided with an image acquisition unit 10
13 to obtain image information, and further obtains the information W
_1, with H _1, W _2, H _2, cut out reading area 1101 from the image information. Then, the operation unit 1010
Sends the image information of the reading area 1101 to the CPU 1020 together with the format information.

When the control unit 1021 of the CPU 1020 receives the information from the operation unit 1020, the control unit 1021 stores the image information of the reading area 1101 in the image buffer 1022.
The format information is stored in the format buffer 1023,
Store each. Next, the character cutout unit 1024
Based on the format information, a character frame 1102 (FIG. 11)
(B) is obtained, a character pattern is cut out from the character frame 1102, and sent to the character recognition unit 1025.
The character recognizing unit 1025 recognizes the character of the character pattern, and outputs a character code as a recognition result to the control unit 1021.
Send to This character code is sent from the control unit 1021 to the operation unit 1010, where display, storage, recording, and the like are performed.

In the conventional character recognition device, when image information acquired by using the image acquiring means 1013 has an inclination, the inclination is corrected as follows.

The operation unit 1010 first measures the inclination angle θ as shown in FIG. 12A, and calculates the inclination height H ₄ from the measurement result. The inclination angle θ can be measured, for example, by measuring the inclination of the upper end of the document or measuring the inclination of the detection mark. And
The document width W _3, is equally divided by the number of pixels height H ₄ gradient.
For example, as shown in FIG. 12 (A), if the inclination height H ₄ corresponds to 10 dots, the document width to 10 split. That is, in this case, the unit length of the document width after division (correction width) becomes W ₄ = W _3/10. Then, a step position is set for each correction width obtained by the equal division.

Next, the operation unit 1010 reads the image information of the document 1102 horizontally from the upper left end. When the read reaches the first step position S, the read position is shifted vertically by one pixel and the image information is read horizontally until the next step position. When the reading of the image information of the first line is completed by repeating such a reading operation, the second line is performed in the same manner as in the first line.
The reading of the image information after the line is executed. The read image information is sequentially written to the internal memory of the operation unit 1010.

With the above operation, corrected image information as shown in FIG. 11B can be obtained. After that, using the image information shown in FIG. 11B, the above-described cutout of the reading area 1101, the subsequent character recognition processing, and the like are executed.

[0015]

However, in the above-described conventional character recognition device, when the step position and the character pattern overlap, the character recognition unit 1025 erroneously recognizes the character pattern or recognizes the character pattern. Sometimes they could not be recognized. This is because a character or a frame has a step at the step position.

For example, as shown in FIG. 13A, when the character “7” is in contact with the lower end of the frame 1301 and the step position S is provided on the character “7”, FIG.
In some cases, the case where the character “2” is in contact with the lower end of the frame 1301 as shown in FIG.

For this reason, there has been a demand for a character recognition device that does not cause recognition errors or unrecognition even if the inclination of the document is corrected.

[0018]

A character recognition apparatus according to the present invention comprises a height determining means for obtaining the height of the image from the width and the inclination angle of the image, and equally dividing the width according to the height of the inclination. A temporary step position determining means for determining a temporary step position by means of the above; a character frame determining means for determining an assumed position of the character frame of the image; A step position determining means for determining a position, and a tilt correcting means for correcting a tilt of an image by performing reading in the width direction while shifting one pixel at a time in the height direction at the step position.

According to the present invention, after the provisional step position is determined using the width and the inclination angle of the image, the assumed position of the character frame of the image is determined, and the assumed position of the provisional step position among the assumed positions of the character frame is determined. The closest one was decided as the step position. Therefore, since the character pattern and the step position do not overlap, there is no possibility that a recognition error or a recognition failure occurs.

[0020]

Embodiments of the present invention will be described below with reference to the drawings. In the drawings, the size, shape, and arrangement of each component are only schematically shown to an extent that the present invention can be understood, and numerical conditions described below are merely examples. .

First Embodiment A character recognition device according to a first embodiment of the present invention will be described below with reference to FIGS.

FIG. 1A is a conceptual diagram showing the system configuration of the character recognition device of this embodiment.

This character recognition system includes an operation unit 110 and a CPU 120 as in the conventional apparatus.

The operation unit 110 includes input means 111 such as a keyboard and a CRT (Cat
A display means 112 such as a hode ray tube) and an image acquisition means 113 using, for example, a general-purpose scanner or the like.

As shown in FIG. 1B, the CPU 120 includes a control unit 121, an image buffer 122, a format buffer 123, a character extraction unit 124, a character recognition unit 125, a tilt correction unit 126, and a reception buffer 127.
It has.

The control unit 121 communicates with a host device such as the operation unit 110 and controls the entire CPU 120.

The image buffer 122 includes a tilt correction unit 1
26 stores the image information whose inclination has been corrected.

The format buffer 123 is provided in the control unit 1
21 stores the format information received from the operation unit 110.

The character extracting section 124 detects a character position in the image information using the format information, and extracts a character pattern for one character.

The character recognizing section 125 recognizes the character indicated by the character pattern and outputs a character code as a recognition result.

The inclination correction unit 126 performs
The inclination of the image information is corrected.

The reception buffer 127 temporarily stores the image information received from the operation unit 110 by the control unit 121.

Next, the operation of the character recognition apparatus shown in FIG. 1 will be described with reference to FIGS.

First, the operation unit 110 reads a document using a general-purpose scanner or the like. Thereby, the image information 201 as shown in FIG. 2 can be obtained.

Next, the operation unit 110 measures the tilt angle θ. The inclination angle θ is measured, for example, by measuring the inclination of the upper end portion of the document or the inclination of the detection mark, as in the conventional apparatus.

Subsequently, the operation unit 110 reads the read area specifying information W ₁ , W ₁ , stored in the internal memory (not shown).
Using H ₁ , W ₂ , and H ₂ (see FIG. 12), a reading area is cut out from the image information 201 of the document in the same manner as in the conventional apparatus. As a result, image information of the reading area 301 as shown in FIG. 3A can be obtained.

The operation unit 110 converts the image information 301 into
Send to CPU120. Further, the operation unit 110 sends the format information to the CPU 120. This format information is the same as the format information of the conventional device, that is, the character height H ₃ , the character pitch P, and the number of characters N (FIG. 3).
(A) includes, in addition to N = 4), the document tilt angle θ and the width W ₃ of the reading area 301 obtained by the above-described measurement.

The control unit 121 in the CPU 120 receives the image information and the format information of the reading area 301 from the operation unit 110. Then, the image information is stored in the receiving buffer 127 and the format information is stored in the format buffer 123.

Subsequently, the inclination correction unit 126 reads the inclination angle θ and the width W ₃ of the reading area 301 from the format buffer 123, and calculates the inclination height H ₄ from these data θ and W ₃ . The inclination correction unit 126 according to the height H ₄ slope, equally dividing the original width W _3. Although the method of equal division is not limited, in this embodiment, the width of the reading area 301 is divided into the same number as the number of pixels of the inclination height. For example, as shown in FIG.
If the inclination height H ₄ corresponds to 4 dots, divided into four document width. In this case, the unit length of the document width after division (correction width) becomes W _3/4. Then, the inclination correction unit 1
Reference numeral 26 sets temporary step positions 311, 312, and 313 for each correction width obtained by the equal division.

Next, the inclination correction unit 126 reads the character pitch P and the number N of characters from the format buffer 123. Then, the inclination correction unit 126 assumes the position of the format frame 320 using the data P, N, and the like. Among the format frames 320, the vertical frames 321 and 322 located between the character patterns are problematic in the subsequent processing.
323. Therefore, in this process, the vertical frame 32
Only 1,322,323 may be assumed.

Further, the inclination correction unit 126 determines the provisional step positions 311, 312, 313 and the vertical frames 321, 322, 3
Compare with position 23. Then, as shown in FIG. 3C, the positions 331, 332, and 333 of the vertical frame closest to the provisional step positions 311, 312, and 313 are respectively
Determine the final step position. That is, in this embodiment, the step position corresponding to the temporary step position 311 is the position of the vertical frame 321, the step position corresponding to the temporary step position 312 is the position of the vertical frame 322, and The step position corresponding to 313 is the position of the vertical frame 323.

Thereafter, the inclination correcting section 126 reads the image information of the reading area 301 horizontally from the upper left end.
When the read reaches the first step position 331, the read position is shifted by one pixel in the vertical direction, and the image information is read horizontally to the next step position 332. When the reading of the image information of the first line is completed by repeating such a reading operation, the reading of the image information of the second and subsequent lines is executed similarly to the first line. The read image information is stored in the image buffer 12.
2 are sequentially written. Thus, a read area image 340 after correction as shown in FIG. 3D is obtained.

Subsequently, the character cutout unit 124 outputs the character height H ₃ and the character pitch P from the format buffer 123.
And the number of characters N are read, and a character frame 350 of the corrected read area image 340 is obtained using the format information. Further, the character cutout unit 124
The character pattern 352 is cut out from the character area 351 in the character frame 350 and sent to the character recognition unit 125. The character recognizing unit 125 recognizes the character of the character pattern and sends a character code as a recognition result to the control unit 121. This character code is sent from the control unit 121 to the operation unit 110, and is displayed, saved, recorded, and the like.

As described above, according to the character recognition device of this embodiment, the vertical frames 321 and 32 of the image
Since the step position can be arranged at the position of 2,323, the character pattern and the step position do not overlap, and therefore, there is no possibility that the character pattern is distorted due to the inclination correction. Therefore, the character recognition device according to the present embodiment can reduce the occurrence rate of recognition errors and unrecognition as compared with the conventional device.

Further, since the character recognition device according to this embodiment assumes the format frame 320 using the format information stored in the operation unit 110 in advance, it is necessary to read the format frame using the image acquisition unit 113. Therefore, regardless of whether the character frame of the original is a dropout color frame (frame that cannot be read by a scanner or the like) or a non-dropout color frame (frame that can be read by a scanner or the like), the inclination is corrected. It is possible to do.

Second Embodiment Hereinafter, a character recognition apparatus according to a second embodiment of the present invention will be described with reference to FIGS.

The system configuration of the character recognition device according to this embodiment is the same as that of the device according to the first embodiment (FIG. 1A).
), And a description thereof will be omitted.

FIG. 4 shows a CPU 12 according to this embodiment.
FIG. 3 is a block diagram showing an internal configuration of the ０0 ’. In the figure,
The components denoted by the same reference numerals as those in FIG.
The same thing as FIG. 1 (B) is shown.

The CPU 120 according to this embodiment has a frame detection unit 401.

The frame detecting section 401 detects the position of the image frame using the image information, as described later.

Next, the operation of the character recognition apparatus according to this embodiment will be described with reference to FIGS.

In this embodiment, a non-dropout color frame is used as a document.

First, as in the first embodiment, the operation unit 1
10 reads the document using the image acquisition unit 113 and acquires image information.

Then, the operation unit 110 measures the tilt angle θ in the same manner as in the first embodiment.

Subsequently, the operation unit 110 uses the read area specifying information W ₁ , H ₁ , W ₂ , and H ₂ stored in the internal memory to convert the image information of the original into the image data in the same manner as in the conventional apparatus.
A reading area 501 as shown in FIG.

The operation unit 110 sends the image information and the format information of the reading area 501 to the CPU 120. This format information includes the character height H ₃ , character pitch P, number of characters N, document tilt angle θ, and reading area 50.
Including one of the width W _3.

When the control unit 121 receives the image information and the format information of the reading area 501 from the operation unit 110, the control unit 121 stores the image information in the reception buffer 127 and stores the format information in the format buffer 123.
To be stored.

Next, the frame detection unit 401 sets the reception buffer 1
With respect to the image information stored in 27, a black point peripheral distribution as shown in FIG. What is the sunspot distribution?
This is the result of counting the number of black pixels for each line in the reading area 501 in the vertical direction. The peak positions of this black point peripheral distribution are represented by vertical frames 511, 512, 51 of the image frame 510.
3, 514, 515, 516.

Subsequently, the inclination correction unit 126 reads the inclination angle θ and the width W ₃ of the reading area 501 from the format buffer 123, and calculates the inclination height H ₄ from these data θ and W ₃ (FIG. 5). (C)). The inclination correction unit 126 according to the height H ₄ slope, equally dividing the original width W _3. For example, as shown in FIG. 5 (D), if the inclination height H ₄ corresponds to 2 dots, two split the document width. That is, in this case, the unit length of the document width after division (correction width) becomes W _3/2. Then, based on the correction width obtained by the equal division, the inclination correction unit 126 determines the provisional step position 5 as shown in FIG.
Set 20.

Further, the inclination correction section 126 searches the above-mentioned distribution around the black point to detect the peak position closest to the temporary step position 520. Then, as shown in FIG. 6A, the peak position obtained by this detection is determined as the final step position 610. That is, in the case of this embodiment, the step position is the position of the vertical frame 513.

After that, the tilt correcting section 126 reads the image information of the reading area 501 horizontally from the upper left end.
Then, when the reading reaches the step position 610, the image information is read horizontally by shifting the reading position by one pixel in the vertical direction. Reading of the image information of the second and subsequent lines is executed in the same manner. The read image information is sequentially written to the image buffer 122. Thereby, a read area image 620 after correction as shown in FIG. 6B is obtained.

Subsequently, the character cutout unit 124 outputs the character height H ₃ and the character pitch P from the format buffer 123.
And the number of characters N are read, and the position of the character frame 630 of the read area image 620 after correction is determined using the format information. Furthermore, the character cutout unit 1
24 cuts out the character pattern 632 from the character area 631 in the character frame 630 and sends it to the character recognition unit 125. The character recognizing unit 125 recognizes the character of the character pattern, and outputs a character code as a recognition result to the control unit 121.
Send to This character code is transmitted from the control unit 121 to the operation unit 1.
The data is sent to the server 10 for display, storage, recording, and the like.

As described above, also in the character recognition device according to this embodiment, since the character pattern and the step position do not overlap, there is no possibility that the character pattern will be distorted due to the inclination correction. Can be reduced.

The character recognition device according to the present embodiment assumes an image frame using a black point peripheral distribution.
Even if the error of the format information or the error of the reading area specifying information is large, accurate character recognition can be performed.

Third Embodiment A character recognition device according to a third embodiment of the present invention will be described below with reference to FIGS.

The system configuration of the character recognition device according to this embodiment is the same as that of the first embodiment (see FIG. 1).
Therefore, the description is omitted.

The character recognition device according to this embodiment has a C
The internal configuration of the PU 120 is different from that of the first embodiment.

FIG. 7 shows a CPU 12 according to this embodiment.
FIG. 3 is a block diagram showing an internal configuration of the ０0 ’. In the figure,
The components denoted by the same reference numerals as those in FIG.
The same thing as FIG. 1 (B) is shown.

The CPU 120 according to this embodiment includes a character interval detecting unit 701.

As will be described later, the character interval detection unit 701
The inter-character area is detected using the image information, and the position of the image frame is assumed based on the detection result.

Next, the operation of the character recognition apparatus according to this embodiment will be described with reference to FIGS.

In this embodiment, a dropout color frame is used as a document.

First, as in the first embodiment, the operation unit 1
10 reads the document using the image acquisition unit 113 and acquires image information.

Then, the operation unit 110 measures the tilt angle θ in the same manner as in the first embodiment.

Subsequently, the operation section 110 uses the read area specifying information W ₁ , H ₁ , W ₂ , and H ₂ stored in the internal memory to convert the image information of the original into the image data in the same manner as in the conventional apparatus.
A reading area 801 as shown in FIG.

The operation unit 110 sends image information and format information of the reading area 801 to the CPU 120. This format information includes the character height H ₃ , character pitch P, number of characters N, document tilt angle θ, and reading area 80.
Including one of the width W _3.

When the control unit 121 receives the image information and the format information of the reading area 801 from the operation unit 110, the control unit 121 stores the image information in the reception buffer 127 and stores the format information in the format buffer 123.
To be stored.

Next, the inter-character detection unit 701 obtains a black point peripheral distribution as shown in FIG. 8B for the image information stored in the reception buffer 127. In this embodiment, since the dropout color frame is used, the position of the vertical frame does not appear in the distribution around the black point. That is, in the black point peripheral distribution, a portion having a large number of black points represents a region where a character pattern exists, and a portion having a small number of black points represents a region where a character pattern does not exist. Therefore, the vertical frame of the image frame exists in an area having a small number of black points.

Subsequently, the inclination correction unit 126 reads the inclination angle θ and the width W ₃ of the reading area 801 from the format buffer 123, and calculates the inclination height H ₄ from these data θ and W ₃ . The inclination correction unit 126 according to the height H ₄ slope, equally dividing the original width W _3. For example, if the inclination height H ₄ corresponds to 2 dots, 8
As shown in (C), the document width is divided into two. That is, in this case, the unit length of the document width after division (correction width) becomes W _3/2. Then, the inclination correction unit 126
Sets the temporary step position 802 as shown in FIG. 8C based on the correction width obtained by this equal division.

Further, the inclination correction unit 126 searches the above-described black point peripheral distribution and detects an area closest to the tentative step position 802 in an area having a small number of black points. And
As shown in FIG. 9A, a final step position 901 is set in an area obtained by this detection.

Thereafter, the inclination correction unit 126 reads the image information of the reading area 801 horizontally from the upper left end.
When the reading reaches the step position 901, the image information is read horizontally by shifting the reading position by one pixel in the vertical direction. Reading of the image information of the second and subsequent lines is executed in the same manner. The read image information is sequentially written to the image buffer 122. Thus, a read area image 902 after correction as shown in FIG. 9B is obtained.

Subsequently, the character cutout unit 124 outputs the character height H ₃ and the character pitch P from the format buffer 123.
And the number N of characters, and a character frame 910 of the read area image 902 after correction is obtained using the format information. Further, the character cutout unit 124
A character pattern 912 is cut out from the character area 911 in the character frame 910 and sent to the character recognition unit 125. The character recognizing unit 125 recognizes the character of the character pattern and sends a character code as a recognition result to the control unit 121. This character code is sent from the control unit 121 to the operation unit 110, and is displayed, saved, recorded, and the like.

As described above, also in the character recognition apparatus according to this embodiment, since the character pattern does not overlap with the step position, there is no possibility that the character pattern will be distorted due to the inclination correction. Can be reduced.

Further, the character recognition apparatus according to the present embodiment assumes the image frame 320 using the distribution around the black point. Therefore, even if the error of the format information or the error of the reading area specifying information is large, the character can be accurately recognized. Recognition can be performed.

[0085]

As described above in detail, according to the present invention, it is possible to provide a character recognition apparatus which does not cause recognition errors or incompatibility even if the inclination of a document is corrected.

[Brief description of the drawings]

FIG. 1A is a conceptual diagram showing a system configuration of a character recognition device according to a first embodiment, and FIG. 1B is an internal configuration of a CPU provided in the character recognition device according to the first embodiment; FIG.

FIG. 2 is a conceptual diagram for explaining the operation principle of the character recognition device according to the first embodiment.

FIG. 3 is a conceptual diagram illustrating the operation principle of the character recognition device according to the first embodiment.

FIG. 4 is a block diagram illustrating an internal configuration of a CPU provided in a character recognition device according to a second embodiment.

FIG. 5 is a conceptual diagram for explaining the operation principle of the character recognition device according to the second embodiment.

FIG. 6 is a conceptual diagram for explaining the operation principle of the character recognition device according to the second embodiment.

FIG. 7 is a block diagram illustrating an internal configuration of a CPU provided in a character recognition device according to a third embodiment.

FIG. 8 is a conceptual diagram for explaining the operation principle of the character recognition device according to the third embodiment.

FIG. 9 is a conceptual diagram illustrating the operation principle of a character recognition device according to a third embodiment.

10A is a conceptual diagram illustrating a system configuration of a conventional character recognition device, and FIG. 10B is a block diagram illustrating an internal configuration of a CPU provided in the conventional character recognition device.

FIG. 11 is a conceptual diagram for explaining the operation principle of a conventional character recognition device.

FIG. 12 is a conceptual diagram for explaining the operation principle of a conventional character recognition device.

FIG. 13 is a conceptual diagram for explaining the operation principle of a conventional character recognition device.

[Explanation of symbols]

 110 operation unit 120 CPU 121 control unit 122 image buffer 123 format buffer 124 character cutout unit 125 character recognition unit 126 inclination correction unit 127 reception buffer 401 frame detection unit 701 character interval detection unit

Claims (4)

[Claims] 1. A height determining means for determining a tilt height from a width and a tilt angle of an image, and a temporary step position is determined by equally dividing the width according to the tilt height. Determining means, setting an assumed position of a character frame of the image, a character frame setting means, determining a step position by selecting a position closest to the provisional step position from the assumed position of the character frame, A character, comprising: a step position determining unit; and a tilt correcting unit that corrects the tilt of the image by performing reading in the width direction while shifting one pixel in the height direction at the step position. Recognition device. 2. The character recognition device according to claim 1, wherein the character frame setting means sets an assumed position of the character frame using format information created in advance. 3. The character recognition according to claim 1, wherein the character frame setting means sets the assumed position of the character frame at a position where the distribution around the black point has a local maximum value or at a position in the vicinity thereof. apparatus. 4. The character recognition device according to claim 1, wherein the character frame setting means sets an assumed position of the character frame in an area having a small black point peripheral distribution.