JPH06309503A - English character recognizing device - Google Patents

Abstract

PURPOSE:To provide an English character recognizing device which can eliminate the integrating errors of broken characters and can recognize the English characters with high accuracy. CONSTITUTION:An English character recognizing device is provided with a character rectangle detecting means 3 which decides a rectangle that is virtually circumscribed to characters as a character rectangle based on the continuation of black picture elements in a document image, a broken character integrating means 11 which performs the integration processing by deciding one character when a certain space between character rectangles is smaller than other spaces of character rectangular, and a character recognizing means 9 which recognizes the characters in the character rectangles and the characters in the character rectangles that have been integrally processed through the means 11. Furthermore, a recognition result deciding means 15 is added to confirm and decide the results of characters in the character rectangle recognized by the means 9. Then, only the character rectangles that are not decided by the means 15 are integrated by the means 11. The character images of the character rectangles to which the recognition results are decided are excluded in regard of decision of the broken characters. Thus, it is possible to prevent such a case where the characters other than the broken characters are wrong integrated even with a document of small character pitches.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an English character recognizing device for recognizing English characters in a document image, and in particular, it is constructed so that "cut characters" which are faint in the middle can be correctly recognized.

[0002]

2. Description of the Related Art In recent years, a character recognition device has been used widely as an input device for inputting characters to a computer or the like, and its field of use has expanded greatly, and high-precision character recognition ability is required.

As shown in FIG. 7, a conventional English character recognition apparatus includes an image input unit 1 for inputting a document to be recognized as a binarized document image, an image storage unit 2 for storing the input document image, and an image storage unit 2. A character rectangle means 3 for obtaining a rectangle virtually circumscribing a character based on a series of black images in a document image,
Character rectangle storage means 4 for storing the coordinate data of the obtained character rectangle on the document image and the serial number given to each character rectangle.
And a word segmentation unit that obtains a horizontal interval between the rectangles based on the coordinate data stored in the character rectangle storage unit 4, detects a portion having a wide interval as a word delimiter, and obtains a character rectangle number forming a word. 5, the word character rectangle storage means 6 for collectively storing the numbers of the character rectangles forming the obtained word, and the coordinate data of the character rectangle corresponding to the character rectangle number are extracted from the character rectangle storage means 4 and A graphic feature extracting means 7 for obtaining a graphic feature by extracting a character image from the image storing means 2 based on the image data, and extracting the distribution of the black pixels.
Recognition dictionary means 8 for storing the graphic features of the characters forming the document image, and graphic feature and recognition dictionary means 8 in the character rectangle.
The character recognition means 9 for comparing the graphic features stored in the table 1 to obtain the difference between them, and the recognition result storing means 10 for storing the recognition result of each character rectangle. When the space between adjacent rectangles is narrower than other character spaces, it is determined as one character and integrated, and the graphic feature extraction means 7 obtains the graphic feature in the integrated rectangle, and then the character recognition means 9 To integrate the broken character requesting the recognition result
11 and display processing means 12 for displaying the recognition result in units of words based on the word information.

The operation of this English character recognition apparatus will be described with reference to FIG. First, the recognition target document is input as a binary image by the image input means 1 and stored in the image storage means 2.
An example of the document image 30 is shown in FIG. The stored data is composed of binary image data indicating whether the intersection point on the X coordinate and the Y coordinate in the document 30 is white or black.

Next, the character rectangle detecting means 3 obtains a rectangle (31 to 49) virtually circumscribing a character based on the series of black pixels in the document image, and the character rectangle storing means 4 stores the character rectangle. Store the coordinates of (b).

Further, based on the coordinates of the obtained character rectangles, the "cut character integrating means" 11 integrates the cut characters (one character "h" is displayed separately in the rectangle 44 and the rectangle 45). And stores the result in the character rectangle storage means 4.

The integration of broken characters is performed by paying attention to the space between the character rectangles in the x-coordinate direction, which is narrower than the normal space (for example, space 50 between the rectangles 41 and 42) (space between the rectangles 44 and 45).
51) is integrated as the interval of break characters.

Based on the stored coordinates of the character rectangle storage means 4, the word cut-out means 5 finds a horizontal interval between the rectangles, and detects a wide interval as a word break,
The character rectangle numbers forming the words are obtained, and the numbers of the character rectangles forming the obtained words are collected and stored in the word character rectangle storage unit 6. The form at the time of storage is shown in FIG.

After that, based on the coordinate data stored in the character rectangle storage means 4, all the character images are taken out from the image storage means 2 and sent to the figure feature extraction means 7 to extract the figure features and the character recognition processing means. The recognition results for all the characters are obtained by 9 and stored in the recognition result storage means 10.

The display processing means 12 displays the recognition result of each character image stored in the recognition result storage means 10 in units of words based on the word information stored in the word character rectangle storage means 6.

By the above processing, the document image including the broken character is recognized.

[0012]

However, in the conventional English character recognizing device, when the broken characters are integrated, all the character rectangles having a narrow gap with the adjacent character rectangles are integrated, so that the pitch between the characters is reduced. When a narrow document is targeted for character recognition, there is a problem that characters other than broken characters are integrated.

The present invention solves these conventional problems, and an object of the present invention is to provide an English character recognizing device capable of recognizing an English character with high accuracy by eliminating an error in integrating broken characters. I am trying.

[0014]

In view of the above, according to the present invention, a character rectangle detecting means for obtaining a rectangle which virtually circumscribes a character as a character rectangle based on a series of black pixels in a document image, and an interval between the character rectangles. Character is smaller than the space between other character rectangles, the character is recognized as one character and integrated processing is performed, and the character in the character rectangle and the character recognition of the character in the rectangle integrated by the cut character integrating means are performed. In the English character recognizing device including a character recognizing means for performing the above, a recognition result deciding means for confirming and deciding a result recognized by the character recognizing means for a character within the character rectangle is provided, and a target of the integrating process in the broken character integrating means is It is limited to only the character rectangles that have not been confirmed by the recognition result confirmation means.

Further, in this English character recognizing device, a character pattern for classifying a character rectangle in a document image into the same character type by superposing a character image in a character rectangle on a pixel unit in another character rectangle. A classifying means and a word dictionary means for storing spelling of English words are provided, and a word character string obtained based on the recognition result of the character recognizing means is a word character string stored in the word dictionary means and a character pattern classification. When a plurality of words including character rectangles classified by the means into the same character type are matched, the recognition result determination means is configured to determine the recognition result of the character rectangles classified into the same character type.

[0016]

Therefore, the determination as to whether the character is a broken character is excluded for the character rectangular character image for which the recognition result has been determined. Therefore, even in the case of a document having a narrow pitch between characters, it is possible to avoid a situation in which characters other than broken characters are mistakenly integrated.

Further, the recognition of the recognition result is confirmed as the correct character recognition when it is determined that a plurality of words including the characters recognized as the same character in the document image are spelled correctly.

[0018]

EXAMPLE An English character recognition apparatus in an example of the present invention is
As shown in FIG. 1, a character pattern classifying means 13 for classifying character images included in a document by superposing them and classifying whether or not they are of the same character type, and a word dictionary means 14 for storing spelling of English words. A recognition result determining means for determining whether or not the spelling of the word is correct by collating the recognized word character string with the word character string stored in the word dictionary means 14, and determining the recognition result if they match. It has 15 and.

The other functional blocks are the same as in the conventional device (FIG. 7).

The character pattern classification means 13 retrieves the coordinate data of the character rectangle from the character rectangle storage means 4 and
A character image is taken out from the image storage means 2 based on it, and all the character images in the document are overlapped with each other in units of pixels constituting the image, and pixels having the same character type are compared. And the number of pixels in the character rectangle, and when the ratio is large, it is classified as the same character type.

Further, the recognition result determining means 15 retrieves the recognition result of the character rectangle number forming the word and the rectangle group to which the character rectangle belongs from the word character rectangle storing means 6 and the recognition result storing means 10, and obtains the result. The word character string stored in the word dictionary means 14 and the word character string stored in the word dictionary means 14 are compared to determine whether or not the word is spelled correctly, and the recognition results of the character rectangles belonging to the same rectangle group are correctly spelled. If it matches the word, the recognition result of the rectangle group is confirmed, and the fact is stored in the recognition result confirmation information of the recognition result storage means 10.

The recognition result storing means 10 which receives information from the character pattern classifying means 13 and the recognition result determining means 15 as well as the character recognizing means 9 is a character rectangle of the same character type group classified by the character pattern classifying means 13. A number, a serial number given to each group of the same character type, a result of the recognition result of the representative rectangle obtained by the character recognition means 9, and recognition result confirmation information are stored.

Further, the broken character integration means 11 pays attention to the character rectangle which is not fixed by the recognition result fixing means 15,
1 if the space between adjacent rectangles is smaller than the space between other characters
It is determined that the character is a character, the characters are integrated, and the figure feature in the integrated rectangle is obtained by the figure feature extraction means 7 and then sent to the character recognition means 9 to obtain the recognition result.

As shown in FIG. 2, the device block representing the hardware structure of the English character recognition device includes an image input device 16 such as a scanner for reading a document to be recognized as a binarized document image and an overall control. Central processing unit (C
PU) 17 and read-on
A memory (ROM) 19, a random access memory (RAM) 23 for temporarily storing processing data, and a CP
A keyboard 18 for giving commands such as start and end to U17 from the outside, a display device 28 for displaying a recognition result,
And an internal signal transmission bus line 29 connecting each means.

Of these, the ROM 19 includes the control program 20 for the CPU 17, the recognition dictionary data 21, and the word dictionary data 22, and the RAM 23 includes the image storage data 24.
, Character rectangle storage data 25, word character rectangle storage data
26 and recognition result storage data 27.

The operation of this English character recognition apparatus will be described with reference to the flowchart of FIG. Step 1: Input the document to be recognized by the image input means 1 and store it in the image storage means 2 as a binarized document image. Step 2: The character rectangle detecting means 3 detects a series of black pixels in the document image, and determines a rectangle virtually circumscribing the character by using a series of black pixels as a character. As a result, the obtained rectangle coordinate data and the serial number given to each character rectangle are stored in the character rectangle storage means 4 as character rectangle information. FIG. 5A shows the character rectangles (31 to 49) obtained in step 2, and FIG. 5B shows the storage form 52 of the character rectangle information in the character rectangle storage means 4. There is.

Step 3; Based on the coordinates of the stored character rectangles, the word slicing means 5 finds the horizontal interval between the rectangles, and detects a portion having a larger interval than the average character interval as a word segment, Find the character rectangles that make up one word. The character rectangles that compose a word are stored in the word character rectangle storage unit 6 with their numbers collected together for each word. FIG. 6 shows a form 61 when the word character rectangle is stored.

Step 4; Next, character pattern classification means
Reference numeral 13 retrieves the coordinate data of the character rectangle from the character rectangle storage means 4, extracts the character image from the image storage means 2 based on the coordinate data, and targets all the character images in the document in units of pixels forming the image. By superimposing them, it is determined whether or not they have the same character type by the ratio of the number of matching pixels and the number of pixels in the character rectangle, and when the ratio is large, they are classified as the same character type.

The process will be described with reference to FIG.
Reference numeral 53 represents a state in which the character image “A” 55 and the character image “B” 54 are overlapped, and the pixel where 55 and 54 overlap
The number of 56 is counted in areas 57, 58, and 59 in which the height of the character is divided into three, and the number of matching pixels in area 57, area 58, and area 59 is determined. After that, divide the number of matching pixels in each area 57-59 by the area of areas 57-59, and
If the value is sufficiently close to 1 in 57 to 59, it is considered that the character images match.

Character rectangles 31 to 49 in the document image 30 are compared with each other and classified as follows.

Character rectangle group serial number Character rectangle number 1 31, 36, 41, 46 2 32, 33, 37, 38, 43 3 34, 39, 42 4 35, 48 5 40 6 44 7 45 8 47 9 9 Step 5 The recognition process is performed by using one of the classified character rectangle groups as a representative pattern. In the recognition processing, the graphic feature extraction unit 7 extracts the graphic feature from the character image in the character rectangle, the character recognition unit 9 compares the graphic feature with the graphic feature stored in the recognition dictionary unit 8, and the difference between the two is obtained. A character having a characteristic with a small difference is set as a recognition result.

The recognition result of each character rectangle group is shown below.

Character rectangle group serial number recognition result 1'a '2'p'3'l'4'e'5'y'6'I'7'l'8'b'9't'Step 4 and Step 5 The classification result and the recognition result obtained in step 3 are stored in the recognition result storage means 10. The form at the time of storage is shown in FIG.

Step 6; Next, a recognition result confirmation process is performed. The recognition result confirmation means 15 retrieves the character rectangle numbers forming the word from the word character rectangle storage means 6, and the recognition result of the corresponding character rectangle is stored in the recognition result storage means 10.
The word character string obtained as a character string by the recognition result is compared with the word character string stored in the word dictionary means 14 to determine whether or not the spelling of the word is correct, and belongs to the same rectangular group. If the recognition result of the character rectangle matches in a plurality of correctly spelled words, the recognition result of the character rectangle group is confirmed, and the fact is stored in the recognition result confirmation information of the recognition result storage means 10.

The process will be specifically described with reference to FIG.

The character rectangle of the word starting with the character rectangle 31, the character rectangle group to which the rectangle belongs, and the recognition result are: The spelling'apple 'is the correct spelling.

The character rectangle of the word starting with the character rectangle 36, the character rectangle group to which the rectangle belongs, and the recognition result are The spelling'apply 'is also a correct spelling.

Therefore, the recognition results of the character rectangle groups 1, 2, and 3 are determined to be'a ',' p ', and'l', respectively, because correct spelling was obtained in a plurality of words.

Step 7: After that, the process of integrating the broken characters is performed on the character image that is not determined in Step 6. Paying attention to the positional relationship of the character rectangles that have not been determined by the broken character integration means 11, if the interval between adjacent rectangles is narrower than the other character intervals, it is determined to be one character and integrated, step 8; After the figure feature extraction means 7 obtains the figure feature in the integrated rectangle, it is sent to the character recognition means 9,
The recognition result is obtained and stored in the recognition result storage means 10.

The process will be specifically described with reference to FIG.

Character rectangle in a word starting with character rectangle 41
As described above, the recognition results of 41, 42, 43, and 46 are fixed, there is no possibility that they are broken characters, and they are excluded from the target of integration of broken characters. The gaps that are the target of integration of broken characters are limited to three positions of character rectangles 44 and 45, 47 and 48, and 48 and 49. Also, among them, only 51 between 44 and 45 is much narrower than the average character spacing,
It is highly possible that the character rectangles 44 and 45 are one character, these characters are integrated and recognized, and as a result, 'h' is obtained.

Step 9: The recognition result obtained by the above processing is displayed in units of words based on the word character rectangle storage means 6 and the recognition result storage means 10.

As described above, in the English character recognition device of the embodiment,
Character types are classified by overlapping characters in a document image, and a recognition result of the classified character group is determined when a plurality of words are spelled correctly. Then, the character image for which the recognition result has been confirmed is not subjected to the cut character integration processing, but the cut character integration processing is performed for the character images excluding it, and the integrated image recognition processing is performed.

As a method for determining the recognition result,
In addition to the method shown in the embodiment, it is also possible to adopt another simplified method such as determining as correct if there are a certain number or more of character images of the same character type in the document image.

[0045]

As is apparent from the above description of the embodiment, the English character recognition apparatus of the present invention performs the recognition result confirmation processing, and excludes the confirmed character from the target of the cut character integration to perform the integration processing. By performing the above, even when performing character recognition for a document with a narrow character pitch, it is possible to improve erroneous execution of broken character integration, and it is possible to recognize English characters with high accuracy.

[Brief description of drawings]

FIG. 1 is a functional block diagram of an English character recognition device according to an embodiment of the present invention,

FIG. 2 is a device block diagram of an English character recognition device according to an embodiment of the present invention,

FIG. 3 is a flowchart showing an operation procedure in the English character recognition device according to the embodiment,

FIG. 4 is a diagram showing a character rectangle (a) of a document image and a storage form (b) of a rectangle number in a character rectangle storage means;

FIG. 5 is a diagram showing a character image (a) superimposed for determining the same character type and a storage form (b) of recognition results;

FIG. 6 is a diagram showing a storage form of word character rectangles;

FIG. 7 is a diagram showing a functional block diagram of a conventional English character recognition device.

[Explanation of symbols]

 1 image input means 2 image storage means 3 character rectangle detection means 4 character rectangle storage means 5 word cutout means 6 word character rectangle storage means 7 graphic feature extraction means 8 recognition dictionary means 9 character recognition means 10 recognition result storage means 11 cut character integration Means 12 Display processing means 13 Character pattern classifying means 14 Word dictionary means 15 Recognition result determining means 16 Image input device 17 CPU 18 Keyboard 19 ROM 20 Control program 21 Recognition dictionary data 22 Word dictionary data 23 RAM 24 Image storage data 25 Character rectangle storage Data 26 Word character rectangle storage data 27 Recognition result storage data 28 Display 29 Bus 30 Document image 31 to 49 Character rectangle 50 Character rectangle spacing 51 Broken character rectangle spacing 53 Overlaid character images 54, 55 Individual character images 56 Overlapping pixels 57-59 area

Claims (2)

[Claims] 1. A character rectangle detecting means for determining a rectangle that virtually circumscribes a character as a character rectangle based on a series of black pixels in a document image, and an interval between the character rectangles is greater than an interval between other character rectangles. In the case where the character is narrow, it is determined that the character is one character and integrated processing is performed, and character recognition means that performs character recognition of the character in the character rectangle and the character in the rectangle integrated by the cut character integration means. In an English character recognizing device, the character recognizing means includes a recognition result deciding means for confirming and deciding a result of recognizing a character in the character rectangle
An English character recognition device, wherein the target of the integration processing in the cut character integration means is limited to only the character rectangles that have not been determined by the recognition result determination means. 2. A character pattern classifying unit for classifying a character rectangle in the document image for each same character type by superimposing a character image in the character rectangle on a character image in another character rectangle on a pixel-by-pixel basis. Provided with a word dictionary means for storing the spelling of words, the word character string obtained based on the recognition result of the character recognition means, the word character string stored in the word dictionary means,
When a plurality of words including character rectangles classified into the same character type by the character pattern classification unit are matched, the recognition result determination unit determines the recognition result of the character rectangle classified into the same character type. The English character recognition device according to claim 1.