JPH0830717A - Character recognition method and device therefor - Google Patents

Abstract

PURPOSE:To provide an inexpensive character recognition method and a device therefor with a high character recognition rate. CONSTITUTION:A picture input part 1 reads characters and stores them in a picture storage part 4. A character recognition part 8 extracts a feature from input picture data, compares it with a prescribed feature stored in a character recognition dictionary part 7, outputs a matching character code and stores it in a character code storage part 9. By reading a prescribed character font stored in a character font storage part 6 corresponding to the character code, the character is displayed at a display part 3. When an operator corrects the displayed character, display character coordinates on a screen are successively inputted by a coordinate input part 2 and stored in a coordinate storage part 5. The character recognition part 8 extracts the feature from the input picture data and coordinate data, compares it with the prescribed feature stored in the character recognition dictionary part 7 and outputs the matching character code. The prescribed character font stored in the character font storage part 6 is displayed at the display part 3 corresponding to the character code.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for recognizing and coding characters in an image.

[0002]

2. Description of the Related Art Generally, a character recognition device includes an image input section for inputting an image, an OCR character recognition section for outputting a character code based on image information, and a character output section for outputting the recognized character as a font. It is composed of a character correction unit for selecting an erroneously recognized character and inputting a correct character.

There is a character correction unit that writes a character on a coordinate input surface with a stylus pen and obtains a character code by performing an online handwritten character recognition process.

[0004]

However, in the online character recognition for character correction, the corrected character is likely to be erroneously recognized when the habit of the character of the operator is remarkable. was there.

Further, if a plurality of different character recognition units are provided in order to deal with this problem, there is a problem that the dictionary capacity becomes large and the cost increases.

The present invention has been made in view of the above conventional example, and an object of the present invention is to provide a character recognition method having a high character recognition rate and a low cost, and an apparatus therefor.

To achieve the above object, an image recognition method and apparatus of the present invention have the following configurations. That is, a first image input means for inputting image information,
Based on the character feature data included in the image information,
Corresponding to the first character recognition means for generating the character code, the selection means for selecting the character code to be corrected from the character codes generated by the first character recognition means, and the character code selected by the selection means. Second image input means for inputting a corrected character stroke, second character recognition for generating a character code based on the characteristic data corresponding to the selected character code and the corrected character stroke input from the second image input means. And means.

Another aspect of the present invention is a first character inputting step of inputting image information from the first image inputting means, and a first character generating a character code based on characteristic data of characters included in the image information. The second step of inputting a recognition step, a selection step of selecting a character code to be corrected from the character codes generated in the first character recognition step, and a corrected character stroke corresponding to the character code selected in the selection step. A second image input step of inputting from the means, a second character recognition step of generating a character code based on the characteristic data corresponding to the selected character code and the corrected character stroke input in the second image input step. Equipped with.

[0008]

In the above structure, the first image inputting means inputs the image information, the first character recognizing means generates the character code based on the character feature data included in the image information, and the selecting means operates. A character code to be corrected is selected from the character codes generated by the first character recognition means, and the second image input means inputs a corrected character stroke corresponding to the character code selected by the selection means, The second character recognizing means generates a character code based on the characteristic data corresponding to the selected character code and the corrected character stroke input from the second image input means.

Another aspect of the present invention is to input image information from the first image input means, generate a character code based on characteristic data of a character included in the image information, and generate a character code from the generated character code. A character code to be corrected is selected, a corrected character stroke corresponding to the selected character code is input from the second image input means, and characteristic data corresponding to the selected character code and the input corrected character stroke are input. Based on this, a character code is generated.

DESCRIPTION OF THE PREFERRED EMBODIMENTS First, the basic structure of the character recognition apparatus of this embodiment according to the present invention will be summarized below.

The character recognition device of this embodiment according to the present invention is
An image input unit that inputs image information, a first character recognition unit that outputs a character code based on the input image information, a character output unit that outputs the character as a font, and a character that is erroneously recognized is selected. A pointing device for inputting coordinate information, and a second character recognition unit for outputting a character code based on the image information and the coordinate information. Prepare Here, the second character recognition unit selects an OCR character recognition unit that outputs a character code based on the image information, and a dictionary selection that selects a part of the dictionary used by the OCR character recognition unit based on the coordinate information. Section. Further, the OCR character recognition unit includes a dictionary classified according to the number of strokes, and the dictionary selection unit includes a stroke number calculation unit that obtains the number of strokes of the character based on the coordinate information. Furthermore, the second character recognition unit
An OCR feature extraction unit that extracts first line information that forms a character based on image information; an online feature extraction unit that extracts second line information that has a direction that forms a character based on coordinate information; And a stroke matching unit that gives information on the direction of the first line information by associating the first and second line information. Furthermore, the second character recognition unit performs recognition using the shape information of the character obtained from the image information, outputs one or more character codes and evaluation information corresponding to each character code, and an OCR character recognition unit, and coordinates. Recognition using the shape information of the character obtained from the information,
An online character recognition unit that outputs one or more character codes and evaluation information corresponding to each character code, and an OCR character recognition unit and a determination unit that outputs the character code based on the output information of the online character recognition unit . The character correction unit performs character recognition based on the image information input from the image input unit and the coordinate information input by the pen. [Embodiment 1] FIG. 2 is an external view of a character recognition apparatus according to an embodiment of the present invention. Here, the image scanner 51
Optically reads a document placed on the document table 52 and transmits image data to the main body 54 via the cable 53. The main body 54 performs the first character recognition from the image data and displays the character font on the screen 55. Screen 55
If the operator finds that the characters displayed on the screen are different from the characters on the document, the characters can be corrected by writing them on the screen 55 with the pen 56. Screen 55
Has a structure in which a transparent coordinate input device is overlaid on the display device, the second character recognition is performed based on the coordinate data of the character written by the pen 56, and the corrected character is displayed on the screen as a character font. 55 is displayed.

This embodiment according to the present invention relates to the second character recognition.

FIG. 1 is a block diagram of the character recognition apparatus of this embodiment.

The image input section 1 optically reads an image containing characters and stores it in the image storage section 4. The character recognition unit 8 extracts the characteristic information from the input image data and compares it with the characteristic information stored in advance in the character recognition dictionary unit 7 to output the character code information that best matches,
It is stored in the character code storage unit 9. The characters are displayed on the display unit 3 by reading out the character fonts stored in the character font storage unit 6 corresponding to the character code.

When the displayed characters are to be corrected by the operator, the coordinates of the characters written on the screen are sequentially changed to the coordinate input section 2.
Is input and stored in the coordinate storage unit 5. The character recognition unit 8 extracts a feature from the information of both the input image data and coordinate data, and compares it with the feature information stored in advance in the character recognition dictionary unit 7 to code the character with the best feature match. Output as information. A character font stored in advance in the character font storage unit 6 corresponding to the character code is read out to display the character on the display unit 3.

Specific processing of the character recognition unit 8 in the first embodiment of the present invention will be described in more detail with reference to the block diagram of FIG.

First, the image data stored in the image storage unit 4 is subjected to noise removal, binarization,
Perform processing such as thinning and cutting out characters. Next OC
The R feature extraction unit 12 extracts, as the character shape information, feature point information such as end points and branch points, and line information connecting the feature points.

In the first character recognition, the shape information is compared with the characteristic information stored in the character recognition dictionary unit 7, and the character having the best characteristic match is output as code information.

On the other hand, in the second character recognition, the coordinate data stored in the coordinate storage unit 5 is used together with the input image data. As the coordinate data, first, the coordinate value when the pen comes into contact with the screen is stored, and each time the pen moves, a new coordinate value is continuously stored. When the pen leaves the screen, a control code called a pen-up code is stored.

The user inputs the stroke of the corrected character by using the pen 56 on the character to be corrected in the character string displayed on the screen 55.

Based on the coordinate data of this stroke, the character selection section 13 first determines the character to be corrected.

The coordinate system of the screen 55 for displaying characters,
The coordinate system of the coordinate input device arranged thereon does not necessarily match, and the position of the character to be corrected and the position of the stroke of the corrected character may be different, and the process of selecting the character to be corrected is It may be an application such as another editor.

When a character is selected, the preprocessing unit 11 reads out the image data corresponding to the selected character from the image storage unit 4. Furthermore, the OCR feature extraction unit 12 compares the shape information of the image data with the feature information of the character recognition dictionary unit 7, and the feature information is classified by the number of strokes of the character,
Only the characteristic information for the specified number of strokes of characters is compared. The stroke count calculation unit 14 obtains the stroke count of a character from the coordinate data by counting the number of pen-up codes per character, and specifies stroke count information for the character recognition dictionary unit 7.

According to this embodiment, characters having a similar shape but different stroke numbers can be excluded from the candidates, so that there is a high possibility that a correct character will be recognized. For example, “Government (8 strokes)”
The effect is great for identifying characters such as "Miya (10 strokes)".

Further, since the amount of data for comparison can be limited depending on the number of strokes, it can be executed at high speed.

Further, in this embodiment, since the online character recognition unit is not provided and only the OCR character recognition unit performs recognition, the capacity of the dictionary can be saved. [Embodiment 2] Specific processing of the character recognition unit 8 in the second embodiment of the present invention will be described in detail with reference to the block diagram of FIG.

The processes of the preprocessing unit 11, the OCR feature extraction unit 12, and the character selection unit 13 are the same as those in the first embodiment.

The online feature extraction unit 21 extracts feature point information such as end points and branch points, and line information connecting the feature points based on the coordinate data stored in the coordinate storage unit 5.

The stroke matching section 22 is provided with the OC.
Feature point information and line information obtained from the R feature extraction unit 12,
The corresponding feature point information and line information are found by matching with the feature point information and line information obtained from the online feature extraction unit 21. Although the line information obtained from the OCR feature unit 12 does not have information that specifies the line direction, the line direction information can be obtained from the corresponding line information obtained from the online feature extraction unit 21.

As described above, the shape information of the line information obtained from the OCR feature extraction unit 12 and the direction information of the line information obtained from the online feature extraction unit are used as the feature information stored in the character recognition dictionary unit 7. It compares and outputs the character with the best feature match as code information.

Although the character recognition dictionary unit of the conventional OCR does not include the information of the direction of the line, in the present embodiment, when the line information is stored in the character recognition dictionary unit 7, the storage address of the coordinate value of the starting point is used. The direction information is stored by determining the order of the end point coordinate value and the storage address so that the start point address is smaller.

FIG. 6 shows an example of the character shape of the dictionary data of the character recognition dictionary unit 7, where (a) is the character "thousand" and (b) is the character.
Represents each stroke of the character "Dai". Here, a stroke is a line connecting characteristic points such as end points and branch points.

FIG. 9 shows an example of the data structure of the character recognition dictionary unit 7, where (a) shows the data structure of the character "thousand" and (b) shows the data structure of the character "dry."

Here, it is assumed that the character having the shape shown in FIG. 7 is input from the image scanner. Further, it is assumed that the character is written in the direction shown by the arrow in FIG. 8 in the correction with the pen. The data shown in FIG. 10 is obtained by referring to the data input from the character image scanner and by referring to the data input by the pen for the stroke direction.

When the data of FIG. 10 and (a) and (b) of FIG. 9 are matched by, for example, the method of least squares, (a) is better matched by considering the stroke direction. Therefore, it can be determined that this character is “thousand”.

In the case of this example, if the recognition is performed only by the image data of the image scanner as in the conventional case, it is likely to be erroneously recognized as "dry", and if the recognition is performed only by the coordinate data of the pen, the katakana " Although there is a high possibility that it will be erroneously recognized as “h”, the possibility of correct recognition is increased by using this embodiment.

As described above, in the present embodiment, when comparing the input data and the dictionary data, it is considered that a match occurs when the line directions match, so that a correct character is recognized as compared with the conventional OCR. Is more likely to occur. For example,
It has a great effect on recognition of characters such as “water” and “tree”. [Third Embodiment] The specific processing of the character recognition unit 8 in the third embodiment of the present invention will be described in detail with reference to the block diagram of FIG.

The processes of the preprocessing unit 11, the OCR feature extraction unit 12, the character selection unit 13, and the online feature extraction unit 21 are the same as those in the first or second embodiment.

The OCR dictionary unit 31 is a dictionary similar to that of the first or second embodiment, but the scores (confidence) corresponding to a plurality of candidate characters are determined according to the degree of feature matching. Output.

On the other hand, in this embodiment, the online dictionary unit 32 is used.
Based on the coordinate data stored in the coordinate storage unit 5, character recognition is performed by a method conventionally known as online character recognition (see, for example, Japanese Patent Publication No. 62-39460), and the degree of feature matching Depending on, a plurality of candidate characters and scores are output.

In the judgment unit 33, the OCR dictionary unit 31
And, for each character output as a candidate in the online dictionary unit 32, it is output as the character code information with the highest total score.

In this embodiment, the recognition result is output by the integrated cooperation of OCR character recognition and online character recognition. Therefore, the character recognition rate can be increased.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of one device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

As described above, the image input unit for inputting an image, the first character recognition unit for outputting a character code based on the image information, and the character output unit for outputting the character as a font are erroneously set. In a character recognition device configured by a character correction unit that performs operation of selecting a recognized character and inputting a correct character, a coordinate input unit that inputs coordinates by a pointing device, and a character code based on the image information and the coordinate information. By providing the second character recognition unit for outputting, there is an effect that the defects of the OCR character recognition and the online character recognition can be compensated for and a higher recognition rate can be obtained.

[0044]

As described above, according to the present invention, a high character recognition rate can be obtained with an inexpensive structure.

[0045]

[Brief description of drawings]

FIG. 1 is a block diagram of a character recognition device of this embodiment.

FIG. 2 is an external view of a character recognition device of this embodiment.

FIG. 3 is a detailed block diagram of a character recognition unit according to the first embodiment.

FIG. 4 is a detailed block diagram of a character recognition unit according to a second embodiment.

FIG. 5 is a detailed block diagram of a character recognition unit according to a third embodiment.

FIG. 6 is a diagram showing an example of a character shape of dictionary data in the second embodiment.

FIG. 7 is a diagram showing an example of a character shape input by an image scanner in the second embodiment.

FIG. 8 is a diagram showing an example of character strokes input by a pen in the second embodiment.

FIG. 9 is a diagram showing an example of a data structure of a dictionary in the second embodiment.

FIG. 10 is a diagram showing a data structure of input data in the second embodiment.

[Explanation of symbols]

 1 Image Input Section 2 Coordinate Input Section 3 Display Section 4 Image Storage Section 5 Coordinate Storage Section 6 Character Font Storage Section 7 Character Recognition Dictionary Section 8 Character Recognition Section 9 Character Code Storage Section

Claims (14)

[Claims] 1. A first image input means for inputting image information, and based on character feature data included in the image information,
A first character recognition means for generating a character code; a selection means for selecting a character code to be corrected from the character codes generated by the first character recognition means; and a character code selected by the selection means. Second image input means for inputting a corrected character stroke; second character recognition for generating a character code based on the characteristic data corresponding to the selected character code and the corrected character stroke input from the second image input means. And a character recognition device. 2. The character recognition device according to claim 1, wherein the character feature data is position data of character end points and branch points, and line information connecting the end points and the branch points. . 3. The character recognition device according to claim 1, wherein the second image input means generates start position and end position data of each stroke from the input corrected character stroke. 4. The second character used by the second character recognition means.
4. The character recognition device according to claim 3, wherein the corrected character stroke input from the image input means is start position and end point position data of each stroke generated by the second image input means. 5. The character recognition device according to claim 1, wherein the first image input means is an optical image scanner. 6. The character recognition device according to claim 1, wherein the second image input means includes a stylus pen and a coordinate input surface. 7. The feature corresponding to the character code selected from the predetermined feature dictionary data corresponding to the stroke number of the corrected character stroke input from the second image input device, the second character recognizing device. The character recognition device according to claim 1, wherein the character dictionary data that is close to the data is searched to generate a corresponding character code. 8. A first image inputting step of inputting image information to a first image inputting means, and based on characteristic data of characters included in the image information,
A first character recognition step of generating a character code; a selection step of selecting a character code to be modified from the character codes generated in the first character recognition step; and a character code selected in the selection step. A second image input step of inputting the corrected character stroke to the second image input means; a character code is generated based on the characteristic data corresponding to the selected character code and the corrected character stroke input in the second image input step. And a second character recognition step of generating the character recognition method. 9. The character recognition method according to claim 8, wherein the character feature data is position data of character end points and branch points, and line information connecting the end points and the branch points. . 10. The character recognition method according to claim 8, wherein the second image input step generates start position and end position data of each stroke from the input corrected character stroke. 11. The corrected character stroke input from the second image input means used in the second character recognition step is start position and end point position data of each stroke generated in the second image input step. 11. The method according to claim 10,
Character recognition method described in. 12. The character recognition method according to claim 8, wherein the first image input means is an optical image scanner. 13. The character recognition method according to claim 8, wherein the second image input means includes a stylus pen and a coordinate input surface. 14. The feature corresponding to the selected character code from the predetermined feature dictionary data corresponding to the stroke number of the corrected character stroke input from the second image input means in the second character recognition step. The character recognition method according to claim 8, wherein the characteristic dictionary data close to the data is searched to generate a corresponding character code.