JPH05298474A - Optical character reader - Google Patents

Abstract

PURPOSE:To provide an optical character reader where an erroneous reading character is efficiency and surely corrected. CONSTITUTION:This device is provided with a scanner 1, a character segmenting part 2, a recognizing part 3, a recognized result storing part 4, a correcting part 5, an input part 6 and a display part 7. The correcting part 5 displays a recognized result by the recognizing part 3 in the display part 7. An operator operated the input part 6 on the display screen of the display part 7 and designates the erroneous reading character in an object to be replaced, its attribution information, a correction character (right character) and a replacing mode (recognition replacement and forced replacement). The correcting part 5 retrieves a recognition character which is the character having the same character code as that of the erroneous reading character designated by the recognition result storing part 4 and also having the same attribution information as designated attribution information. The correcting part 5 replaces the retrieved recognition character (erroneous reading character) with the correction character in accordance with the designated replacing mode.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reader, and more particularly to an optical character reader having a function of correcting misread characters included in recognized characters.

[0002]

2. Description of the Related Art An optical character reading device reads a character written on a document as a character-recognized character (recognition result). However, under the present circumstances, it is difficult to obtain a complete character recognition result without misreading characters, and therefore correction work by an operator is required.

Conventionally, the correction work is performed by the operator comparing the recognition result with the original and correcting the misread characters of the optical character reading device one by one. This correction method includes a method of correcting using a general Kana-Kanji conversion and a method of selecting from a candidate character string output by the optical character reading device by an operator.

[0004]

However, in any of the correction methods, when the recognition result is short, the correction time is not so long, but when it is long as in a general document, the correction time is proportional to the number of characters. It was a burden to the operator.

Further, since the correction work is monotonous, there is a problem that it is easy to overlook a misread character.

Therefore, the present invention has been made in view of the above circumstances, and an object of the present invention is to provide an optical character reading device capable of efficiently and reliably correcting a misread character. ..

[0007]

In order to achieve the above object, the invention according to claim 1 is an optical character reading apparatus for reading a character written on a manuscript as a character-coded recognized character, wherein the recognized character is A recognition result storage unit to store,
Based on the designation of the misread character and the corrected character included in the recognized character, the recognition result storage unit is searched for a recognized character having the same character code as the designated misread character, and the found recognized character is the corrected character. And a correction unit for replacing

According to a second aspect of the present invention, in an optical character reading device for reading a character written on a manuscript together with its attribute information as a recognized character, the recognition character and its attribute information are stored. A result store,
A recognition character having the same character code as the specified misread character from the recognition result storage unit based on the specification of the misread character included in the recognized character, its attribute information, and the correction character, and the specified attribute It is characterized by including a correction unit that searches for a recognized character having the same attribute information as the information and replaces the searched recognized character with the corrected character.

According to a third aspect of the invention, in the first or second aspect of the invention, the replacement by the correction unit is performed according to the selected replacement mode.

[0010]

According to the first aspect of the invention, when a character written in a certain format is read by a certain reading device, the recognized character includes a substantially constant misread character. Therefore, by designating one misread character and a corrected character included in the recognized character, the correction unit searches all the misread characters (recognized characters). Then, the correction unit replaces the retrieved recognized character with the corrected character. This eliminates the need for the operator to correct the misread character one by one, and the misread character can be corrected efficiently and reliably.

According to the second aspect of the present invention, when a character written in a certain format is read by a certain reading device, the recognized character includes a substantially constant misread character. Therefore, by designating one misread character included in the recognized character, its attribute information, and the modified character, the modification unit searches all the misread characters (recognized characters). Then, the correction unit replaces the retrieved recognized character with the corrected character.
This eliminates the need for the operator to correct the misread character one by one, and the misread character can be corrected efficiently and reliably. In addition, the identification of the misread character to be replaced becomes more accurate.

According to the third aspect of the present invention, more efficient correction can be performed by selecting the replacement mode depending on whether or not the misread characters included in the recognized characters are constant.

[0013]

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

FIG. 1 is a schematic block diagram showing an embodiment of the optical character reading apparatus of the present invention.

This apparatus comprises a scanner 1 for detecting an image of an original document 10, a character cutout section 2 for cutting out a character pattern from the detected original image, and a character cutting section 2 for detecting attribute information of the cut out character pattern, and a character. A recognition unit 3 which obtains a character-coded recognition character by performing character recognition processing on the character pattern cut out by the cut-out unit 2; a recognition result storage unit 4 which stores a recognition result including the recognition character and attribute information;
Correction unit 5 for correcting misread characters included in recognized characters
And a display section 7 such as a CRT display and an input section 6 having a keyboard and a mouse for correcting misread characters.

Next, the details of each of the above parts will be described.

The scanner 1 includes a light source for irradiating the original 10 with light and a photoelectric conversion element for receiving the reflected light from the original 10 and converting it into an electric signal, and optically scans the entire original 10. The original image is detected.

The character cut-out unit 2 cuts out a character pattern for each character from the document image detected by the scanner 1, detects attribute information of the cut-out character pattern, and detects each character together with the document image before cutting. The pattern and its attribute information are output to the recognition unit 3.
The attribute information detected by the character cutout unit 2 includes, for example, the position (coordinates) of the character pattern, the size of the character pattern (horizontal,
Vertical), dictionary type (print, handwriting), feature vector (n
Dimensional vector) etc.

The recognition unit 3 has a candidate character memory for storing candidate character patterns, performs character recognition processing on the character patterns cut out by the character cutting unit 2, and outputs candidate character strings. The character recognition processing performed here is performed by, for example, a superposition method (pattern matching method). That is, the character pattern cut out from the original image is compared with the candidate character pattern stored in the candidate character memory to calculate the similarity value, and the first candidate character having the largest similarity value is sequentially selected from the nth candidate character. Up to a character, a plurality of candidate characters are determined. The character recognition process here is not limited to the pattern matching method, and other methods may be used.

The recognition result storage unit 4 stores a recognition file,
It is composed of a candidate file and an image file. In the recognition file, the recognition character (character code) by the recognition unit 3 and the attribute information detected by the character cutout unit 2 are stored in association with the recognition character. Further, the candidate file stores the candidate character strings determined by the recognition unit 3. Further, the image file stores an image of one document (including partial images such as drawings and photographs). Then, the recognition file and the candidate file are used to correct the misread character. A line image file may be provided and an image of each line of the document 10 may be stored in this line image file.

The correction section 5 controls each section of this apparatus, and also controls display described later, search processing for misread characters,
It is provided with a CPU 50 that performs replacement processing and the like, a misread character memory 51, a corrected character memory 52, and a program memory 53 connected to this CPU 50.

Misread character memory 51 and modified character memory 5
Under the control of the CPU 50, 2 stores the misread character or the corrected character (correct character) selected (input) by the input unit 6, respectively. Further, the program memory 53 stores an operation program for correcting misread characters. The CPU 50 operates according to the operation program.

The display control performed by the CPU 50 will be described.

The CPU 50 recognizes the recognition result (recognition character, original image, etc.) stored in each file of the recognition result storage unit 4.
A correction screen (described later) created in a predetermined format from is displayed on the display unit 7 by display control. At the stage of setting correction conditions (search and replacement conditions), a correction condition setting screen (described later) is displayed and controlled. Is displayed on the display unit 7.

An example of the correction screen is shown in FIG. The layout 70 of the manuscript 10 is displayed in the upper right of the screen of FIG.
At the center of the screen in the figure, the contents of the block 72a in which the cursor (for example, blue) 71 is currently displayed and the other blocks 72b following it are displayed, and the cursor 7 is displayed at the bottom of the screen.
A line image 73 in which 1 is currently displayed is displayed, a window is opened in the lower right part of the screen so as to overlap the block 72 displayed in the center of the screen, and a candidate character string 74 is displayed in an overlapping manner. The line image 73 is displayed by the CPU 50 by cutting out the corresponding line image from the document image stored in the image file of the recognition result storage unit 4. In the figure, specifically, the cursor 71 is the layout 70.
It is displayed below the character “ヰ” in the block 72a with a diagonal line, “2. Japanese test” is displayed as the line image 73, and the first candidate character is “ヰ”, 2 The candidate character is "ki", the third candidate character is "noon", the fourth candidate character is "medium", the fifth candidate character is "gi", and the sixth candidate character is "half". Shows.

An example of the correction condition setting screen is shown in FIG. In the frame of the screen shown in the figure, the replacement target character (misread character) "ヰ", the correction character (correct character) "ki", the attribute specification (valid, invalid), block (all blocks, current block) ), Pattern size (horizontal, vertical), dictionary type (yes / no, print, handwriting, print / handwriting), feature vector matching (yes / no), replacement mode (confirmation replacement, forced replacement), confirmation key (YES, NO) is displayed. The figure shows that the misread character "ヰ" is changed to the corrected character "ki". Also, in the figure, the black circle in the center of the double circle mark indicates that it is designated. Therefore, in the figure, the attribute designation is selected to be valid, all the blocks to be searched and replaced are selected, and the pattern size is 5 mm ± 1 in both horizontal and vertical directions.
A value within mm is input, the dictionary type is set to “Yes” and the type is selected, “Feature vector matching” is set to “Yes”, and the replacement mode is set to “confirmed replacement”.

Search process (search process) performed by the CPU 50
Will be described with reference to FIG.

On the screen shown in FIG. 3, the operator can operate the mouse or keyboard of the input unit 6 to select each attribute information. When the misread character or the modified character is designated as the search condition in the input unit 6 and the attribute designation is set to invalid, the CPU 50 stores the recognized character having the same character code as the misread character in the recognition result storage unit 4. It is a search from the recognition file. Also,
When the misread character or the modified character is specified as the search condition in the input unit 6 and the attribute specification is selected to be valid, it is a recognized character having the same character code as the specified misread character, The recognition file having the same attribute information as the generated attribute information (meaning the same within a certain range) is searched from the recognition file in the recognition result storage unit 4.

More specifically, the retrieval process of the CPU 50 will be described. As for the “block” to be retrieved and replaced, when all the blocks are selected, the recognition characters (misread characters) in all the blocks in the manuscript are selected. ) Is searched and if the current block is selected, the current cursor 7
Search within the block in which 1 is displayed. As for the “pattern size”, the recognized characters corresponding to the horizontal and vertical sizes of the input pattern are searched.
Therefore, basic character sizes such as double-width, full-width, and half-width can be handled by inputting this pattern size. The basic character size such as double-width, full-width, and half-width may be selectable. For “Dictionary type”, select “Dictionary type available”, and if either print type, handwriting, or print type / handwriting is selected,
When the recognition character having that type is searched and no dictionary type is selected, the search by type is not performed. For “feature vector matching”, if “Yes” is selected, the feature vector of the misread character is compared with the feature vector of another recognized character, for example, simple matching is performed, and the result is a certain value (eg, 80%)
It is determined that the misrecognized character and the feature vector are the same as those of the recognized characters described above, and if none is selected, the feature vector is not compared.

The replacement process performed by the CPU 50 will be described with reference to FIG.

There are two types of replacement modes: confirmation replacement and forced replacement. The CPU 50 performs the replacement process according to the replacement mode selected by the input unit 6 among these replacement modes. That is, when the confirmation replacement is selected on the correction condition setting screen shown in FIG. 3, the operator is asked whether the replacement is necessary each time the misread character is searched by the search processing, and the replacement is required. Sequentially replace only the ones with the corrected characters (search-confirm-replace). When the forced replacement is selected, the misread character retrieved by the retrieval process is forcibly replaced with the corrected character (search-replace) without confirming with the operator. Also in this case, similarly to the search processing, when the search area (block) is selected by the input unit 6, the CPU 50 performs the replacement processing only for the block.

Next, the operation of the present embodiment will be described with reference to the flowchart shown in FIG. 4, taking as an example the case where the misread character "ヰ" is replaced with the corrected character "ki".

It is assumed that the original 10 to be read by the apparatus has characters written in a fixed format (printing of characters, etc.).

First, the scanner 1 detects the image of the original 10. Next, the character cutout unit 2 cuts out a character pattern for each character from the document image detected by the scanner 1, detects attribute information of the cutout character pattern, and outputs the attribute information to the recognition unit 3. The recognition unit 3 includes the character cutting unit 2
Character recognition processing is performed on the character pattern output from the recognition pattern storage unit 4, and the recognition result (recognition character, candidate character string, attribute information, document image, etc.) is output to the recognition result storage unit 4. The recognition result storage unit 4 stores the output recognition result (recognition character, candidate character string, attribute information, document image, etc.) in each corresponding file in the storage unit 4.

The CPU 50 of the correction section 5 executes display control, search processing and replacement processing according to the operation program stored in the program memory 53. First, the CPU 50
Displays a correction screen as shown in FIG. 2 created in a predetermined format from the recognition result stored in each file of the recognition result storage unit 4 on the display unit 7 by display control.

Here, the operator compares the recognition result displayed on the display unit 7 with the original 10 and finds the first misread character "" (S1).

The operator puts the pointer on the found misread character "ヰ" and clicks. CP
Based on the click operation, U50 displays a cursor 71 under the misread character "ヰ" displayed on the display unit 7 as shown in FIG. 2, and the misread character is selected from the candidate files in the recognition result storage unit 4. The candidate character string 74 corresponding to “ヰ” is searched and displayed on the display unit 7 in an overlapping manner.

Next, the operator determines whether or not there is a correction character (correct character) in the candidate character string 74. If there is no correction character in the first to sixth candidate characters displayed on the screen, scroll display, page turning display, etc.
To the nth candidate character are displayed. In this example, since there is a correction character (correct answer character) in the candidate character string 74, the operator operates the mouse of the input unit 6 to move the pointer to the correction character "ki" and click. CPU50
Displays the selected candidate character “ki” in black and white based on the click operation. If there is no correction character in the candidate character string 74, the operator inputs the correction character “ki” by a general correction method (kana-kanji conversion).

Next, when the correction character "ki" is selected or input, the operator selects the candidate character string 74 on the display screen.
Move the pointer to "Replace" displayed at the bottom of and click. Based on the click operation, the CPU 50 corrects only one character of the misread character "ヰ" indicated by the cursor 71 with the correction character "ki" and corrects it. When the correction character “ki” is selected from the candidate character string 74, the double-click operation may serve as both the selection of the correction character “ki” and the replacement instruction (S2).

Next, the operator presses the PF key on the keyboard of the input unit 6.

The CPU 50 determines whether or not the PF key is pressed and the condition α is true (S3). The “condition α” means that the recognized character (including the corrected character) indicated by the cursor 71 is different from the first candidate character of the candidate character string 74 at that time.

When the PF key is pressed and the condition α is true, the CPU 50 stores the misread character (character code) "ヰ" in the misread character memory 51, and the corrected character memory 5
The correction character (character code) "ki" is stored in 2 (S
4).

Next, the CPU 50 displays the correction condition setting screen as shown in FIG. 3 based on the depression of the PF key.
The operator operates the mouse or keyboard of the input unit 6 to select attribute designation, search, block to be replaced, pattern size input, dictionary type selection, feature vector matching on the screen shown in FIG. And the replacement mode are selected. Further, when it is desired to change the correction character here, the correction character displayed by operating the mouse or keyboard of the input unit 6 is changed to another correction character. Here, as shown in FIG. 3, the attribute designation is selected to be valid, all the blocks to be searched and replaced are selected, the pattern size is input within 5 mm ± 1 mm in both horizontal and vertical directions, and there is a dictionary type. It is assumed that the print type is selected, the feature vector matching is selected as “Yes”, and the confirmation mode is selected as the replacement mode.

After the correction conditions have been set, the operator operates the mouse of the input unit 6 to move the pointer to the position of "YES" shown in FIG. 3 and clicks (S).
5).

Based on the click operation, the CPU 50 redisplays the correction screen as shown in FIG. 2 by the display control of the display unit 7, and performs the search process and the replacement process according to the correction condition set in FIG.

The CPU 50 stores the misread character "ヰ" stored in the misread character memory 51 from the current position of the cursor 71 to the file end in the recognition file of the recognition result storage unit 4.
All the blocks are searched for a recognized character having the same character code as the recognized character having the same attribute information as the attribute information set on the correction condition setting screen (S6).
When the CPU 50 searches for the misread character "ヰ" based on the selected replacement mode (confirmation replacement), the CPU 50 confirms with the operator whether or not replacement is necessary (for example, the fact is displayed on the display screen). When the operator determines that the replacement is necessary, the operator performs a replacement request operation (for example, pressing the return key). Based on the replacement request operation, the CPU 50 replaces the retrieved misread character "ヰ" with the corrected character "ki" stored in the corrected character memory 52 (S7). In this way, when the search, confirmation, and replacement are sequentially executed up to the file end (S8), the process returns to the start and the cursor 71 returns to the original position (the position shown in FIG. 2). On the display screen of the display unit 7,
By the display control by the CPU 50, the movement of the cursor 71 and the replacement of the erroneously read character are displayed so that the operator can understand the process of the search and replacement processing.
Then, when the search, confirmation, and replacement are completed up to the file end, all the misread characters "ヰ" in the designated block 72a are replaced with the corrected characters "KI" and displayed on the display unit 7.

According to the optical character reader of the above-described embodiment, one misread character among the recognized characters can be found and corrected, and other misread characters can be confirmed by just setting the correction condition. Alternatively, since the correction character can be forcibly replaced, the misread character can be corrected efficiently and reliably.

The present invention is not limited to the above embodiment,
Various modifications can be made without departing from the spirit of the invention.
For example, the attribute information that is the search information may use a font, a basic character size (double-byte, full-width, half-width) or the like. For example, if the Gothic type logo is more likely to be erroneous than the Mincho type logo, by specifying such a font and limiting the replacement target, efficient search and replacement processing can be performed.
The effect that the burden on the operator is reduced is obtained. Also, the replacement target is not limited to one character, the character range when searching for the replacement target is specified as "Higashi no Taro", and the position of the replacement target character is after "East" and "Taro". It is also possible to replace the erroneously read character "no" with the corrected character "shiba" by designating as before. Further, "Toshiba" may be replaced in units of the word "bunshiba" including the misread character "bundle". By doing in this way, the misread character to be replaced can be accurately specified, the chances of forced replacement are increased, and efficient correction can be performed. Also, the replacement target may be specified by the type of printing device such as a dot printer or a laser printer.

[0049]

According to the invention described in claim 1 described above in detail, the operator does not need to correct the misread characters one by one, so that the misread characters can be corrected efficiently and surely. It is possible to provide a simple optical character reader.

According to the second aspect of the present invention, the correction unit searches and replaces the misread character in consideration of the attribute information of the misread character. Therefore, in addition to the effect of the first claim, It enables more accurate identification of misread characters.

Further, according to the invention of claim 3, since the replacement mode can be selected depending on whether or not the misread character included in the recognized character is constant, in addition to the effect of claim 1 or 2, More efficient correction is possible.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing an embodiment of an optical character reading device of the present invention.

FIG. 2 is a diagram showing an example of a correction screen on a display unit of the present embodiment.

FIG. 3 is a diagram showing an example of a correction condition setting screen on the display unit of the present embodiment.

FIG. 4 is a flowchart for explaining the operation of this embodiment.

[Explanation of symbols]

 1 Scanner 2 Character Extraction Section 3 Recognition Section 4 Recognition Result Storage Section 5 Correction Section 6 Input Section 7 Display Section 10 Original

Claims (3)

[Claims] 1. An optical character reading device for reading a character entered on a manuscript as a character-coded recognized character,
A recognition result storage unit that stores the recognition character, and a recognition character having the same character code as the designated misread character from the recognition result storage unit based on the designation of the misread character and the correction character included in the recognition character. An optical character reading device comprising: a correction unit that searches and replaces the searched recognition character with the correction character. 2. An optical character reading device for reading a character written on a manuscript together with its attribute information as a character coded recognition character, and a recognition result storage unit for storing the recognition character and its attribute information, and the recognition character. A recognition character having the same character code as the designated misread character from the recognition result storage unit based on the designation of the misread character, its attribute information, and the corrected character included in the designated attribute information. And a correction unit that replaces the searched recognition character with the correction character. 3. The optical character reader according to claim 1, wherein the replacement by the correction unit is performed according to the selected replacement mode.