JPH07210632A - Character recognition device - Google Patents

Abstract

PURPOSE:To provide the character recognition device superior in reliability and workability which automatically discriminates connection relations between sentence areas to recognize the characters. CONSTITUTION:This device is provided with an image read part 1 which reads in a document or the like and stores image data in an image storage part 2, a layout storage part 3 which extracts individual sentence areas from image data to recognize the layout, a character recognition part 4 which recognizes characters in each sentence area to store them in a recognition result storage part 5, a word dictionary 6 where all words are stored, a connection discriminating part 7 which couples characters or a character string in the end part of each sentence area and characters or a character string in the head part of another sentence area to generate a character string for discrimination and collates it with the word dictionary 6 and discriminates connectability between sentence areas in the case of success of collation with the word dictionary 6, and a recognition result sort part 9 which sorts code data of each sentence area based on the connectability obtained by the connection discriminating part 7 and stores the result in a sentence data storage part 8.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for converting printed characters, dot characters and handwritten character patterns of newspapers, magazines, novels etc. into code data such as JIS code.

[0002]

2. Description of the Related Art In recent years, as computers have become more sophisticated, documents have been digitized for the purpose of reducing the storage space for documents and improving workability in grinding and the like. As an input device therefor, various character recognition devices have been developed, which can read a document or the like and directly convert it into code data such as JIS code instead of a keyboard.

A conventional character recognition device will be described below. 6 is a diagram showing an example of a document to be recognized, FIG. 7 is a diagram showing a general layout recognition result for FIG. 6, and FIG. 8 is a diagram showing a recognition result for a conventional character recognition device for FIG. is there. In FIG. 7, a is a first text area obtained by the general layout recognition processing for FIG. 6, b is a second text area similar to the first text area a, and c is
Is a third sentence area similar to the first sentence area a, and d is the first sentence area
It is a fourth text area similar to the text area a.

In the conventional character recognition, first, a general layout recognition process is performed on a document to be recognized based on the distribution of characters and the like. Here, it is assumed that the layout recognition result as shown in FIG. 7 is obtained from the recognition target document as shown in FIG. 6 and four sentence regions a to d are extracted. Next, according to the layout recognition result, character recognition is performed in each sentence area in a predetermined order, and a recognition result is obtained. Here, character recognition is performed from the upper left sentence area toward the lower right. Here, as shown in FIG.
Character recognition is performed in the order of the first sentence area a, the second sentence area b, the third sentence area c, and the fourth sentence area d.
As is apparent from FIG. 8, the conventional character recognition device erroneously recognizes the connection relation between the text areas.

[0005]

However, in the above-mentioned conventional configuration, since the order of character recognition of the respective text areas obtained by the layout recognition processing is determined only from these positional relationships, each text area is determined. There is a problem in that the connection relationship between them is erroneously recognized and reliability is poor, and correction of the recognition result is complicated and time-consuming, resulting in poor workability. In addition, if character recognition is performed for each text area and the user specifies these connection relationships, the documents to be recognized cannot be recognized all at once, which is cumbersome and troublesome and lacks workability. Had the problem.

The present invention solves the above-mentioned conventional problems, and provides a character recognition device excellent in reliability and workability, which is capable of automatically recognizing the connection relation between each text area and performing character recognition. The purpose is to provide.

[0007]

To achieve this object, a character recognition apparatus of the present invention comprises an image reading unit for reading a document or the like and converting it into image data, and an image data obtained by the image reading unit. An image storage unit for storing, a layout recognition unit for recognizing a layout by extracting individual text areas from the image data stored in the image storage unit, and an image data stored in the image storage unit. A character recognition unit for recognizing the inside of each sentence region extracted by the layout recognition unit, and the code data obtained by the character recognition unit is stored for each individual sentence region extracted by the layout recognition unit. A recognition result storage unit, a word dictionary, a character or character string at the end of each sentence area stored in the recognition result storage unit, and another sentence. When the characters or character strings at the beginning of the area are combined to create a character string for determination and the character string for determination is collated with the word dictionary and the collation succeeds, the connectivity between the two sentence areas is established. A connection determination unit that determines that there is,
A sentence data storage unit that stores a final recognition result, and code data for each sentence region stored in the recognition result storage unit is sorted based on the connectivity determined by the connection determination unit, and the sentence data And a recognition result sorting section to be stored in the storage section.

[0008]

With this configuration, the layout recognition section extracts individual sentence areas from the document to be recognized, the character recognition section recognizes each character in each sentence area, and the recognition results for these are individually recognized. Stored in the storage unit, the connection determination unit creates a character string for determination by combining the character or character string at the end of each text area and the character or character string at the head of another text area. , The judgment character string is collated with the word dictionary, it is judged that there is connectivity between the sentence areas that have been successfully collated, and the recognition result sorting unit determines the code for each sentence region in the recognition result storage unit. The data is sorted so that the text areas that are judged to be connected by the connection judgment unit are connected,
By storing the result in the text data storage unit, the order of recognizing each text area in the recognition target document can be determined not only by the position of each text area in the recognition target document but also by the end portion of each text area. Since the connectivity as a sentence between the other sentence area and the beginning portion of the sentence area is determined, the recognition accuracy for the recognition target document can be improved. Moreover, since the recognition order can be automatically determined and the characters to be recognized can be collectively recognized, the user can perform the character recognition work very easily.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A character recognition device according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a functional block diagram of a character recognition device according to an embodiment of the present invention. Reference numeral 1 denotes an image reading unit configured by a scanner (not shown) or the like for reading a document to be recognized and converting it into image data, 2 an image storage unit for storing the image data obtained by the image reading unit 1, 3 an image storage unit A layout recognition unit for recognizing the layout by extracting individual text areas based on the distribution of characters in the image data in FIG.
Reference numeral 4 is a character recognition unit for performing character recognition on the image data in the image storage unit 2 for each text region extracted by the layout recognition unit 3 to obtain code data, and 5 is a code obtained by the character recognition unit 4. A recognition result storage unit that stores data for each sentence area extracted by the layout recognition unit 3, 6 is a word dictionary that stores correct spelling of words, and 7 is a layout recognition unit 3 that is stored in the recognition result storage unit 5. The character or character string at the end of each text area extracted in step 3 and the character or character string at the beginning of the other text area extracted by the layout recognition unit 3 are combined to create a judgment character string. A connection determination unit that determines that there is connectivity between the two sentence areas when the determination character string is compared with the word dictionary 6 and the matching is successful, and 8 is a sentence data storage that stores the final recognition result. Part 9 is recognized The code data for each text area stored in the result storage unit 5 is sorted based on the connectivity determined by the connection determination unit 7, and the resulting code data is used as a final recognition result as a text data storage unit. 8 is a recognition result sorting unit stored in FIG.

The operation of the character recognition device having the above-described structure according to the embodiment of the present invention will be described below. 2 to 4 are flowcharts of the character recognition device in the embodiment of the present invention, and FIG. 5 is a diagram showing the final recognition result of the character recognition device in the embodiment of the present invention with respect to FIG. In FIG. 5, a is the first text area,
b is the second sentence area, c is the third sentence area, and d is the fourth sentence area. Since these are the same as in the conventional example, the same reference numerals are given and the description thereof is omitted.

First, the image reading section 1 is used for the operation shown in FIG.
The document to be recognized as shown in (1) is read and the obtained image data is stored in the image storage unit 2 (S1). Next, the layout recognition unit 3 performs layout recognition on the image data in the image storage unit 2 to obtain a layout recognition result as shown in FIG. 7, and for each text area obtained by this processing. Then, each temporary area number is added (S2). Here, the temporary area number 1 for the first sentence area a, 2 for the second sentence area b, 3 for the third sentence area c, and 4 for the fourth sentence area c. 3 and 4 are added to the fourth text area d. Next, 4 which is the maximum value of the temporary area number is substituted into the variable n (S3). Next, the variable i
1 is substituted for (S4), the inside of each text area is character-recognized by the character recognition unit 4, and the recognition result is stored in the recognition result storage unit 5 for each text area (S5-
S7). Next, in FIG. 3, 1 is substituted into the variable rc which is a counter of the number of processed text areas (S8). Next, 1 is first in the array variable r () indicating the connection order information of the text area.
Is assigned to specify that processing is first performed from the first text area a with the temporary area number 1 (S9). Next, 1 is substituted into the variable i which is the counter of the loop (S10). Next, the last character of the text area with the temporary area number i is substituted into the variable c1 (S11). Here, since i is 1, the variable c
The end character "finger" of the first sentence area a is substituted for 1.
Next, 1 is substituted for the variable j which is the counter of the loop (S12). Next, 1 is added to the variable k that is the loop counter.
Is substituted (S13). Next, it is checked whether the k-th value of the array variable r () storing the temporary area number of the processed text area is equal to j (S14). If yes,
The sentence area having the temporary area number j is judged to be processed and jumped to S21. If No, jumped to S15. Here, since r (1) = 1, go to S21.
ump. Next, the variable j is incremented by 1 (S
21). Here, j = 2. Next, in FIG. 4, it is checked whether the variable j exceeds the variable n (S22). No
If yes, jump to S13 to perform processing for the next text area, and if Yes, it is determined that the text area whose tentative area number is i has no connectivity and S28.
Jump to (S22). Here, the variable j is the variable n
Since it has not exceeded, jump to S13. next,
Similarly to the above, it is checked whether the first of the array variable r () is equal to 2 (S13 to S14). Here, r (1) = 1
Therefore, the variable k is incremented by one (S1
5). Next, it is checked whether the variable k is larger than the variable rc (S16). If No, jump to S14.
If Yes, the sentence area with the temporary area number j is judged to be unprocessed and jump to S17. Here, since the variable k is larger than the variable rc, jump to S17.
To do. Next, the first character of the text area with the temporary area number j is substituted into the variable c2 (S17). Here, since j is 2, the variable c2 has the first character "out" in the second sentence area b.
Is substituted. Next, the variables c1 and c2 are combined and substituted into the variable s to create a judgment character string (S18).
Here, the character string “pointing” is assigned to the variable s. Next, the variable s is collated with the word dictionary 6 (S19). Here, it is assumed that the character string “instruction” is not stored in the word dictionary 6. Next, it is checked whether the collation is successful in S19 (S20). If Yes, the temporary area number is i
If it is determined that there is connectivity between the text area of No. and the text area of the temporary area number j, jump to S23, and if No, jump to S21. Here, since the collation has failed, the process jumps to S21. Next, the variable j is incremented by 1 (S21). Here, j = 3. Next, in FIG. 4, it is checked whether the variable j exceeds the variable n (S22). Here, since the variable j does not exceed the variable n, jump to S13. Next, similarly to the above, the first character of the sentence area with the temporary area number of 3 is set to the variable c2.
(S13 to S17). Here, the character "constant"
Is obtained. Next, the c1 obtained in S11 and the variable c2 obtained in S17 are combined and substituted into the variable s (S1
8). Here, the character string "designation" is obtained. Next, collation by the word dictionary 6 is performed and it is checked whether or not this is successful (S19 to S20). Here, the character string "designation" is stored in the word dictionary 6, and since the collation is successful, the variable j is incremented by 1 and jumped to S22. Next, the variable rc is incremented by 1 (S23). Here, rc = 2. Next, at the rcth position of the array variable r (), a variable j indicating the provisional region number of the connected text region
Is substituted (S24). Here, 3 is substituted into r (2). Next, the variable j is substituted for the variable i (S25),
Next, specify the text area to check the connectivity. Where i =
It becomes 3. Next, it is checked whether the variable rc is equal to the variable n (S26). If No, jump to S11 to perform the process for the next sentence area, and if Yes, jump to S27. Here, rc = 2
Since n = 4, jump to S11. Next, similarly to the above, a text area having connectivity with the third text area c having the temporary area number 3 is searched (S11 to S2).
2). Here, the second text area b having the temporary area number 2 is obtained as the text area connected to the third text area c.
Next, the variable rc is decremented by 1 (S23). Here, rc = 3. Next, the variable r of the array variable r ()
The variable j is substituted for the c-th (S24). Where r
2 is substituted into (3). Next, the variable j is substituted for the variable i (S25). Here, i = 2. Then the variable r
It is checked whether c and the variable n are equal (S26). Where r
Since c = 3 and n = 4, jump to S11.
Then, similarly to the above, a text area having connectivity with the second text area b having the temporary area number 2 is searched (S11 to S11).
S22). Here, the fourth sentence region d having the temporary region number 4 is obtained as the sentence region connected to the second sentence region b. Next, the variable rc is incremented by 1 (S2
3). Here, rc = 4. Next, the variable j is substituted for the variable rcth of the array variable r () (S24).
Here, 4 is substituted into r (4). Next, the variable j is substituted for the variable i (S25). Here, i = 4. Next, it is checked whether the variable rc and the variable n are equal (S2
6). Since rc = 4 and n = 4 here, S27
Jump to. Next, the recognition result sorting unit 9
The recognition results stored for each sentence area in the recognition result storage unit 5 are sorted into each sentence area based on the array variable r () obtained in S24 or the like, and the final recognition result is stored in the sentence data storage unit 8 It is stored in (S27). Here, since the contents of the array variable r () are 1, 3, 2, and 4, the sorting is performed so that the temporary area numbers are in the order of 1, 3, 2, and 4.
An accurate recognition result as shown in FIG. 5 is obtained.

On the other hand, if Yes in S22, it is determined that the sentence area having the temporary area number i is not connected, and the array variable r () is arranged at the variable rc + 1 of the array variable r (). The temporary area number stored in the variable rc-th of is added with 1 (S28). Next, the variable rc
Is incremented by 1 (S29). Next, the temporary area number stored in the variable rcth of the array variable r () is
The variable i is substituted and jumped to S26 (S30).

In one embodiment of the present invention, one character at the end of a sentence area and one character at the beginning of another sentence area are combined to create a character string for determination. If the character string for determination is created from the character strings at the end and the beginning, the recognition accuracy can be improved, which is preferable in terms of reliability. Further, if not only the independent word but also the auxiliary word is used in the word dictionary 6, the recognition accuracy can be similarly improved, which is preferable in terms of reliability. Furthermore, although a Japanese document is used as the recognition target document, this may be another document such as English.

[0014]

As described above, according to the present invention, the layout recognition section extracts individual sentence areas from the document to be recognized, the character recognition section recognizes each character in each sentence area, and recognizes them. The results are individually stored in the recognition result storage unit, and the connection determination unit determines by combining the character or character string at the end of each text area and the character or character string at the beginning of another text area. Create a character string for comparison, compare this character string for determination with the word dictionary, determine that there is connectivity between the sentence areas that have been successfully matched, and the recognition result sorting unit stores the result in the recognition result storage unit. By sorting the code data for each text area so that the text areas that are determined to be connected by the connection determination unit are linked and storing the result in the text data storage unit, each text in the recognition target document is sorted. Determine the order of recognizing areas by recognizing each text area. Not only the position in the target document,
Since the connectivity as a sentence between the end of each sentence area and the beginning of another sentence area is determined, the recognition accuracy for the recognition target document can be improved, and the reliability is excellent. Further, since the recognition order can be automatically determined and the characters to be recognized can be collectively recognized, the user can perform the character recognition work extremely easily and improve the workability. It is possible to realize a remarkably excellent character recognition device.

[Brief description of drawings]

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

FIG. 2 is a flowchart of a character recognition device according to an embodiment of the present invention.

FIG. 3 is a flowchart of a character recognition device according to an embodiment of the present invention.

FIG. 4 is a flowchart of a character recognition device according to an embodiment of the present invention.

FIG. 5 is a diagram showing a final recognition result of the character recognition device in the embodiment of the present invention with respect to FIG. 6;

FIG. 6 is a diagram showing an example of a recognition target document.

FIG. 7 is a diagram showing a general layout recognition result for FIG. 6;

8 is a diagram showing a conventional character recognition result for FIG.

[Explanation of symbols]

 1 image reading unit 2 image storage unit 3 layout recognition unit 4 character recognition unit 5 recognition result storage unit 6 word dictionary 7 connection determination unit 8 sentence data storage unit 9 recognition result sorting unit

Claims (1)

[Claims] 1. An image reading unit that reads a document or the like and converts it into image data, an image storage unit that stores the image data obtained by the image reading unit, and an image data stored in the image storage unit. A layout recognition unit for recognizing a layout by extracting individual text regions from the character recognition unit for recognizing the layout, and a character for recognizing the inside of each text region extracted by the layout recognition unit among the image data stored in the image storage unit. A recognition unit, a recognition result storage unit that stores the code data obtained by the character recognition unit for each sentence area extracted by the layout recognition unit, a word dictionary, and a recognition result storage unit that stores the code data. The character or character string at the end of each text area and the character or character string at the beginning of another text area are combined to create a judgment character string. And a sentence data storage unit for storing a final recognition result, which determines that there is connectivity between the two sentence areas when the determination character string is compared with the word dictionary and the collation is successful. And a recognition result sorting unit that sorts the code data for each text area stored in the recognition result storage unit based on the connectivity determined by the connection determination unit and stores the result in the text data storage unit. A character recognition device characterized by being provided.