JPH0981685A - Occidental character recognition device and occidental character recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an occidental character recognition device whose processing speed is fast, recognition accuracy is high and operability and operation efficiency are high when the recognition characters of low recognition likelihood are many. SOLUTION: This device is provided with a character processing method judgement means 15 for deciding the processing speed of the recognition character depending on at what ratio the recognition characters of the high recognition likelihood are present in a document, a contact character separation means 7 for judging the recognition characters of the low recognition likelihood as contact characters, obtaining a separating position from graphic features and separating the contact character at the separating position when the ratio of the recognition characters of the high recognition likelihood is high in the character processing method judgement means 15, a word candidate detection means 16 for collating a word string composed of the recognition characters with words stored in a word dictionary, selecting one or plural word candidates and displaying them at a display part when the ratio of the recognition characters of the high recognition likelihood is low in the character processing method judgement means 15 and a word candidate establishment means 17 for selecting an accurate word from the word candidates displayed in the word candidate detection means 16 by a user.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention converts image data of Western characters such as alphabets input by an image input device such as an image scanner into character data composed of character codes such as ASCII code and processed by a computer or the like. Regarding the Western character recognition device and the Western character recognition method to facilitate,
In particular, the present invention relates to a Western character recognition device and a Western character recognition method used for recognizing image data that has poor image quality and is difficult to distinguish between contact characters.

[0002]

2. Description of the Related Art In recent years, it has been noticed that a large amount of character data can be processed in a short time by taking a document in which characters are written into a computer by an image input device such as an image scanner and processing the characters. A character recognition device that converts image data into a character code or the like is required for processing in a computer. Especially, since Western characters have few character types and high conversion efficiency, they are being put to practical use. At this time, it is desired that accurate conversion be performed with high quality regardless of the quality of the input image data.

A conventional Western character recognition device will be described below. FIG. 4 is a functional block diagram of a Western character recognition device in a conventional example. In FIG. 4, 1 is an image input unit for inputting a recognition target document including characters as binary image data, 2 is an image storage unit for storing the image data input in the image input unit 1, and 3 is an image storage unit 2. A character slicing means 4 for obtaining a rectangle circumscribing a character in the image data stored in the slicing device and slicing the rectangle as a character region, 4 measures the distribution of black pixels in the horizontal direction in the character region cut out by the character slicing device 3, and determines the character. A line segmenting means for segmenting a region into a line segment, 5 is a word segmenting segment for segmenting a word region by measuring a distribution of black pixels in the vertical direction with respect to each line segment segmented by the line segmenting segment 4, and 6 is a character segmenting segment. Three
The distribution of the black pixels of the image in the character area obtained by is extracted as a graphic feature, and this graphic feature and the graphic feature of the standard character stored in the recognition dictionary described later are pattern-matched to determine the recognized character and at the same time. Character recognition means for measuring the recognition accuracy from the degree of collation, 7 is a character which has a low recognition accuracy among the recognized characters determined by the character recognition means 6 is used as a contact character, and the separation position is detected from the figure feature to detect the contact character. The contact character separating means for cutting off the character and again collating it with the recognition dictionary by the character recognizing means 6, 8 is a recognition dictionary for storing the graphic features of the entire recognition target character, and 9 is the character area cut out by the character cutting means 3. A character area storage unit for storing, 10 is a line area storage unit for storing the line area cut out by the line cutting means 4, and 11 is a word area for storing the word area cut out by the word cutting means 5. A storage unit, 12 is a recognition result storage unit that stores the recognition characters obtained by the character recognition unit 6, and 13 is a character cutting unit 3, a line cutting unit 4, a word cutting unit 5, a character recognition unit 6, and a contact character separating unit 7. The control unit 14 has a recognition result output unit that outputs the final recognition result stored in the recognition result storage unit 12.

The operation of the Western character recognizing apparatus having the above-described structure will be described below. First, a document to be recognized containing Western characters is input as a binary image from the image input unit 1 into the Western character recognition device and stored in the image storage unit 2.
Next, in the character cut-out means 3, a rectangular area circumscribing a black pixel area which is considered to be a character pattern in the image data stored in the image storage section 2 is stored in the character area storage section 9 as a cut-out character area. Next, line cutting means 4
At, the line area is cut out from the distribution of black pixels in the horizontal direction of the character area stored in the character area storage unit 9 and stored in the line area storage unit 10. Next, in the word cutting means 5, the distribution of vertical black pixels for each line area is measured from the character area stored in the character area storage unit 9 and the line area stored in the line area storage unit 10 to determine the word area. The cut-out word area storage unit 11 stores it. Next, in the character recognition means 6,
The graphic features corresponding to the respective characters in the character area stored in the character area storage unit 9 are subjected to pattern matching with the graphic characteristics of the standard characters stored in the extraction recognition dictionary 8 to make similar patterns and recognized characters. Is calculated as the recognition accuracy, and the character code of the recognized character and the recognition accuracy are stored in the recognition result storage unit 12. Next, the contact character separating means 7
When the recognition accuracy stored in the recognition result storage unit 12 is low, the character area is determined to be a contact character, the separation position of the contact character is detected from the graphic feature, the character is separated at the separation position, and then the character recognition unit 6 again recognizes the character. To recognize. Next, the recognition result output unit 14 outputs the recognition result stored in the recognition result storage unit 12 to complete the recognition process for Western characters.

[0005]

However, in the above-mentioned conventional Western character recognition apparatus, the frequency of re-recognition increases and the time required for recognition increases as the print quality of the recognition target document is low and the number of contact portions between characters increases. However, there is a problem in that the work efficiency becomes low and the work efficiency decreases. As a result, there is a problem in that post-processing after the recognition processing is troublesome and workability and work efficiency are poor.

The present invention solves the above-mentioned conventional problems, and when there are many recognized characters with low recognition accuracy, word recognition is performed using the recognized character with high recognition accuracy as a key, and the processing speed is increased by selecting word candidates. The object of the present invention is to provide a Western character recognition device that is fast and has high recognition accuracy and high workability and work efficiency, and a Western character recognition method that has both high recognition speed and recognition accuracy and excellent workability and reliability. .

[0007]

In order to achieve this object, a Western character recognition apparatus according to claim 1 of the present invention recognizes a recognition character having a high recognition accuracy by the proportion of the recognition character in a document. A character processing method determining means for determining a character processing method, and a character recognition method having a high recognition accuracy in the character processing method determining means determines a recognition character having a low recognition accuracy as a contact character and separates it from a graphic feature. Contact character separating means for separating contact characters at the obtained separation position, and matching with a word string composed of recognized characters and words stored in a word dictionary when the proportion of recognized characters with high recognition accuracy in the character processing method determination means is low The word candidate detection means for selecting one or a plurality of word candidates and displaying them on the display unit, and the word candidate confirmation means for selecting an accurate word from the word candidates displayed by the word candidate detection means by the user. Consisting configuration having a means.

The Western character recognition method according to claim 2 of the present invention comprises a character processing method determining step for determining the processing method of the recognized character according to the proportion of the recognized character having high recognition accuracy in the document. In the character processing method determination step, when the proportion of recognized characters with high recognition accuracy is high, the recognized character with low recognition accuracy is determined to be a contact character, the separation position is obtained from the graphic feature, and the contact character is separated at the separation position. In the character processing method determining step, when the proportion of recognized characters having high recognition accuracy is low, the word string made up of the recognized characters is collated with the words stored in the word dictionary, and one or more word candidates are selected and displayed on the display unit. A word candidate detection step displayed on the screen, and a word candidate confirmation step in which the user selects an accurate word from the word candidates displayed in the word candidate detection step. Consisting of configuration that.

Here, the determination value of the recognition accuracy in the character processing method determination means may be a fixed value in advance, but it may be variable so that the user can set it while confirming the recognition result.

[0010]

With this configuration, the character processing method determining means accurately determines at what rate the recognized characters having a high recognition accuracy are present in the document, so that the recognition speed and the recognition accuracy can be improved in the subsequent steps. Can be selected. The contact character separating means can accurately identify the separation position of the recognized character recognized as the contact character, separate the recognized character from the separated position, and perform the recognition process again to improve the recognition accuracy. The recognition accuracy can be further improved by the word candidate detecting means collating the recognized characters identified for each character with the word dictionary on a word-by-word basis.
By making the user determine whether or not the word candidate is correct by the word candidate determining means, the recognition accuracy can be increased and at the same time the correction in the post-processing can be omitted.

[0011]

【Example】

(Embodiment 1) Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a functional block diagram of the Western character recognition device in the first embodiment, and FIG. 2 is a device block diagram of the Western character recognition device in the first embodiment.
In FIG. 1, 1 is an image input unit, 2 is an image storage unit, 3 is character cutting means, 4 is line cutting means, 5 is word cutting means, 6 is character recognition means, 7 is contact character separation means,
8 is a recognition dictionary, 9 is a character area storage unit, 10 is a line area storage unit, 11 is a word area storage unit, 12 is a recognition result storage unit, 1
Reference numeral 3 is a control unit and 14 is a recognition result output unit. These are the same as those in the conventional example, and the same reference numerals are given to omit the description.

Reference numeral 15 is a character processing method determining means for determining whether to recognize the recognized character or to collate it with a word dictionary in word units at a later step based on the recognition accuracy of the recognized character stored in the recognition result storage unit 12. Reference numeral 16 matches one or a plurality of similar words by matching a word area stored in the word area storage unit 11 or a recognition character stored in the recognition result storage unit 12 with a word stored in a word dictionary described later. A word candidate detecting means for selecting as a word candidate, and 17 is a word candidate determining means for determining an accurate word from the word candidate selected by the word candidate detecting means 16 displayed on the display unit, which will be described later, and determining the recognition result. , 18 is a word candidate storage section for storing the word candidates detected by the word candidate detecting means 16 and the coordinates thereof,
19 is a word dictionary that stores the spelling of all words, 20
Is a display unit for displaying the positions of the word candidates stored in the word candidate storage unit 18 and the word candidates in the image data stored in the image storage unit 2, and 21 is the word candidate displayed by the user in the display unit 20. It is a recognition result input unit that selects an accurate word from among them and inputs it to a Western character recognition device to obtain a recognition result.

In FIG. 2, reference numeral 22 is an image input device composed of an image scanner, a camera or the like for capturing a document to be recognized as image data into a Western character recognition device, and 23 is a user's input of a system start / end command or a recognition result. An input device including a keyboard, a trackball, a pointing device, and the like for issuing a command, 24 is a CR that displays the operating status of the system or image data, word candidates, and the like.
T is a display device including a liquid crystal display, 25 is a central processing unit (abbreviated as CPU) for performing overall control, 26
Is a control program including each control command to be loaded and used by the CPU, 27 is a read-only read-only memory (ROM) in which the control program 26 is stored, 28 is image data stored in the image storage unit 2, Reference numeral 29 is character area data stored in the character area storage unit 9, 30 is line area data stored in the line area storage unit 10, 31 is word area data stored in the word area storage unit 11, and 32 is recognition result storage. The recognition result data stored in the unit 12, 33 is the word candidate data stored in the word candidate storage unit, 34 is a write-erasable random access memory (RAM) that stores each data in the middle of calculation, 35 Is an output device such as a printer that outputs the recognition result, and 36 is an internal bus that connects the constituent devices and exchanges control signals and data.

The operation of the Western character recognition apparatus of the first embodiment constructed as described above will be described below with reference to the drawings. FIG. 3 is a flow chart showing the operation of the Western character recognition apparatus in the first embodiment. First, in the image input unit 1, the image of the document to be recognized including Western characters is
It is converted into value image data and stored in the image storage unit 2 (image input process, S1). Next, in the character cut-out means 3, a group of black pixels connected at eight adjacent points in the image data stored in the image storage unit 2 is regarded as one character pattern, and a rectangle circumscribing this character pattern is set as a character area. cut. At that time, when the size of the extracted rectangle is a minute rectangle and a rectangle exists in the immediate vertical direction of the rectangle, it is regarded as a separated character such as i and j, and the two rectangles are integrated into one character area. In other cases, the extracted circumscribed rectangle is used as one character area, and the coordinates of the upper left corner and the width and height of the rectangle are stored in the character area storage unit 9 for each character rectangle (character cutout process, S2). Next, the line cutting means 4 checks the horizontal overlap of the character rectangles in the character area stored in the character area storage unit 9 to find the line area of each line, and determines the coordinates of the upper left corner and the lower right corner of each line. The coordinates are stored in the line area storage unit 10 (line cutout process, S3). next,
In the word cutout means 5, a histogram of the horizontal intervals of the character regions to which the lines belong is obtained for each line region stored in the line region storage unit 10 and a threshold value for dividing the peaks of this histogram into two is obtained. If the interval is wide, the word is divided into words, and the coordinates of the upper left corner and the coordinates of the lower right corner are stored in the word area storage unit 11 for each word (word cutting process, S4). Next, in the character recognition means 6, the image data in all the character areas stored in the character area storage unit 9 is taken out from the image storage unit 2 and the distribution of black pixels is extracted as the graphic feature to obtain the graphic feature in the recognition dictionary 8. The recognition accuracy is calculated from similar characters by pattern matching, the character with the highest recognition accuracy is used as the recognition character, and the character code is stored as a recognition result together with the recognition accuracy in the recognition result storage unit 12 (character recognition processing, S5). Next, the character processing method determination means 15 determines whether or not 85% or more of the recognized characters whose recognition accuracy stored in the recognition result storage unit 12 is a predetermined value or more (character processing method determination processing, S6). In the case of YES, the contact character separating means 7 has a low recognition accuracy in the graphic features of the recognized characters, where the graphic connection is weak, for example, the distribution of black pixels in the vertical direction is small and the connection position is one place. Cut the image with (contact character separation process, S
7). Next, the image data cut by the contact character separating means 7 is again subjected to character recognition processing by the character recognizing means 6 and the one having the highest recognition accuracy is recognized as the recognition result.
2 (character recognition process, S8). Next, step 11 is executed. If step 6 is NO, the word candidate detection means 16 separates the recognized character from the recognized character stored in the recognition result storage section 12 and the word area stored in the word area storage section 11 into a word string and this word string. The word stored in the word dictionary 19 is collated, and a word having a certain degree of collation accuracy is displayed on the display unit 20 as a word candidate (word candidate detection process, S9). Next, the word candidate determination means 17
In, the user stores an accurate word from the word candidates displayed on the display unit 20 in the recognition result storage unit 12 as a recognition result (word candidate determination process, S10). Next, the recognition result is output to the recognition result output unit (recognition result output process, S11).
Complete the operation.

According to the present embodiment, the character processing method determining means for determining the processing method of the recognized character based on the distribution state of the recognition accuracy, and the word collation of the word dictionary when the ratio of the recognized character having the high recognition accuracy is small. Since the word candidate detection means for selecting one or a plurality of word candidates and the word candidate determination means for allowing the user to determine the recognition result from the word candidates are provided, the recognition character distribution with high recognition accuracy is small and recognized. The user can perform accurate character recognition by selecting a word candidate even for image data of poor quality in which the accuracy is widely distributed. As a result, there is no need for complicated post-processing, rapid character recognition is possible, and work efficiency can be improved. Further, even an unskilled person can easily work and the workability is improved.

[0016]

As described above, according to the present invention, the character processing method determining means and the character processing method determining means for determining the processing method of the recognized character depending on the proportion of the recognized character having high recognition accuracy in the document. When the ratio of recognition characters with high recognition accuracy in the means is high, the recognition character with low recognition accuracy is determined to be a contact character, the separation position is obtained from the graphic features, and the contact character is separated at the separation position, and a character processing method. When the ratio of recognition characters with low recognition accuracy is high in the determination means, a word string made up of recognition characters is compared with a word stored in a word dictionary to select one or more word candidates and display them on the display unit. Since the detection means and the word candidate confirmation means for selecting an accurate word from the word candidates displayed by the word candidate detection means are included, the proportion of recognized characters with high recognition accuracy is small. Image data with deteriorated quality is recognized for each word by using recognized characters with high recognition accuracy as a key, and one or more word candidates are selected to let the operator decide the final word. The calibration work of is easy and the workability is improved as a result. Therefore, it is possible to realize an excellent Western character recognition device that reliably performs highly reliable character recognition even on a recognition target document having many contact characters with poor print quality and a wide distribution of recognition accuracy.

Further, according to the present invention, the character processing method determining step of determining the processing method of the recognized character depending on the ratio of the recognized characters having high recognition accuracy in the document, and the recognition accuracy in the character processing method determining step. When the ratio of recognized characters is high, the recognized character with low recognition accuracy is recognized as a contact character, the separation position is obtained from the figure feature, and the contact character is separated at the separation position. A word candidate detecting step of selecting one or a plurality of word candidates by matching with a word string made up of the recognized characters and a word stored in the word dictionary when the proportion of the recognized characters having low accuracy is high; Since the user has a word candidate determination step of selecting an accurate word from the word candidates displayed in the word candidate detection step,
It is possible to realize an excellent character recognition method in the United States and America, which has excellent recognition speed and recognition accuracy, and has high work efficiency, and can also recognize characters with high reliability, even for a recognition target document having many contact characters with poor print quality.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a Western character recognition device according to a first embodiment.

FIG. 2 is a device block diagram of a Western character recognition device in the first embodiment.

FIG. 3 is a flowchart showing the operation of the Western character recognition device in the first embodiment.

FIG. 4 is a functional block diagram of a conventional Western character recognition device.

[Explanation of symbols]

 1 image input unit 2 image storage unit 3 character cutout unit 4 line cutout unit 5 word cutout unit 6 character recognition unit 7 contact character separation unit 8 recognition dictionary 9 character area storage unit 10 line area storage unit 11 word area storage unit 12 recognition result Storage unit 13 Control unit 14 Recognition result output unit 15 Character processing method determination unit 16 Word candidate detection unit 17 Word candidate determination unit 18 Word candidate storage unit 19 Word dictionary 20 Display unit 21 Recognition result input unit 22 Image input device 23 Input device 24 Display device 25 Central processing unit 26 Control program 27 Read only memory 28 Image data 29 Character area data 30 Line area data 31 Word area data 32 Recognition result data 33 Word candidate data 34 Random access memory 35 Output device 36 Internal bus

Claims (2)

[Claims] 1. A character processing method determining means for determining a processing method of the recognized character according to a ratio of recognition characters having high recognition accuracy in a document, and a high recognition accuracy in the character processing method determining means. When the ratio of the recognized characters is high, the recognized character having a low recognition accuracy is determined as a contact character, a separation position is obtained from a graphic feature, and the contact character is separated at the separation position, and the character processing method determination is performed. When the ratio of the recognized characters having high recognition accuracy is low in the means, the word string made up of the recognized characters is collated with the word stored in the word dictionary, and one or more word candidates are selected and displayed on the display unit. A Western sentence characterized by having a candidate detecting means and a word candidate determining means for selecting an accurate word from the word candidates displayed by the word candidate detecting means by a user. Recognition device. 2. A character processing method determining step of determining a processing method of the recognized character according to a ratio of recognition characters having high recognition accuracy in a document, and a high recognition accuracy in the character processing method determining step. When the ratio of the recognized characters is high, the recognized character having a low recognition accuracy is determined as a contact character, a separation position is obtained from a graphic feature, and the contact character is separated at the separation position, and the character processing method is determined. In the step, when the proportion of the recognized characters having high recognition accuracy is low, the word string composed of the recognized characters is compared with the word stored in the word dictionary, and one or more word candidates are selected and displayed on the display unit. A candidate detecting step, and a word candidate determining step of selecting an accurate word from the word candidates displayed by the user in the word candidate detecting step, Western character recognition method which is characterized in that.