JPH0291785A - Character recognizing device - Google Patents

Abstract

PURPOSE:To attain correct character recognition by deciding whether or not block synthesis is executed based on two adjacent block widths, segmenting the block data based on the decided result and executing the character recognition based on the segmented block data. CONSTITUTION:Image data is inputted by an input means, and block data including character pattern data is extracted based on the input image data by a block extracting means 7. Next, by block synthesizing deciding means 8 and 9, it is decided whether or not the block synchronization is executed based on two adjacent block widths of the extracting block data. By character segmenting means 10 and 11, the block data are segmented based on the decided result by the block synthesizing deciding means 8 and 9, and by a recognizing means 12, the character recognition is executed based on the block data segmented with the character segmenting means 10 and 11. Thus, the character segmenting to list up the word candidates by the word collation can be correctly executed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character recognition device, and for example, to a character recognition device that recognizes separated characters.

[Prior Art] Conventionally, in this type of device, character lines in optical character recognition are first extracted by taking a vertical black pixel histogram, and then each extracted character is extracted. By taking a horizontal histogram for each line, a circumscribed rectangle (hereinafter referred to as a block) of a character is obtained. In this case, for example, for separated characters such as ``i'', ``su'', ``ru'', and ``butsu'', the block of each separated character is divided into two blocks on the left and right, but when the left and right blocks are combined, When the composite block has a width close to that of an average character, the composite block is cut out as one character.

Here, the character recognition operation will be explained based on the schematic block configuration of the conventional character recognition device shown in FIG.

First, image data on a document optically read by the reading unit 51 is stored in the memory 52 . Next, based on the image data, the block extraction section 53 performs row cutting and block extraction as described above. If there are blocks that can be combined, they are combined in the block combination unit 54 and transferred to the character buffer 5.
Store in 5. Then, the recognition unit 56 performs character recognition on the standard pattern of characters stored in the recognition dictionary unit 57 and the single block or composite block stored in the character buffer 55. The standard pattern character-recognized in this manner is used as a recognized character, and its character code is stored in a word buffer (not shown) of the word matching unit 58. When a predetermined recognized character is stored in the word buffer, the word is compared with the word dictionary stored in the word dictionary section 59.

[Problems to be Solved by the Invention] However, in the above conventional example, small characters such as periods, commas, bullets, and half-width numbers are mixed, so for example, as shown in FIG. ,” are synthesized to form “3. ”, which may result in incorrect block composition. As a result, correct character candidates are already eliminated at the character extraction stage. It is possible to compensate by finding word candidates by comparing the words, but - even if there are differences in letters, accuracy will be lost, and errors in block composition will lead to incorrect numbers of characters composing a word. Therefore, it is necessary to compare the degree of similarity between words of different lengths, which makes word matching complicated.

The present invention has been made in view of the above-mentioned drawbacks of the conventional examples, and its purpose is to provide a character recognition device that can accurately cut out characters to list word candidates in word matching. .

[Means for Solving the Problems] In order to solve the above-mentioned problems and achieve the purpose, a character recognition device according to the present invention is a character recognition device that performs character recognition based on image data. an input means; a block extracting means for extracting block data including character pattern data based on the input image data; and determining whether to perform block synthesis based on two adjacent blocks of the extracted block data. It includes a block composition determination means, a character extraction means for extracting block data based on the determination result by the block composition determination means, and a recognition means for performing character recognition based on the block data extracted by the character extraction means. It is characterized by

Preferably, the recognition means includes a word candidate forming means that forms word candidates based on the result of character recognition, and a word matching means that performs word matching on the word candidates.

Furthermore, preferably, the recognition means includes an identification means that uses a character pattern in the block data as a candidate for a recognition character and identifies the recognition character based on the degree of similarity to a standard pattern stored in advance.

[Operation] According to the above configuration, the input means inputs image data, the block extraction means extracts block data including character pattern data based on the input image data, and the block composition determination means extracts block data adjacent to the extracted block data. It is determined whether or not to perform block synthesis based on two blocks that match, and the block data is extracted by the character extraction means based on the determination result of the block composition determination means, and the block data is extracted by the character extraction means by the recognition means. Character recognition is performed based on block data.

Embodiments] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Description of the first embodiment> First, a first example will be described.

FIG. 1 is a block diagram showing a first embodiment of a character recognition device according to the present invention. In the figure, 1 indicates a character recognition device of the first embodiment. 2 indicates a CPU that controls the entire device. 3 is a control program for operating the CPU2.

Error processing program and FIG. 3(a), which will be described later.
A ROM stores programs according to the flowchart shown in (b), and 4 indicates a RAM used as a work area for various programs stored in the ROM 3 and a temporary save area during error processing. Reference numeral 5 indicates a reading section that optically reads the original image, and 6 indicates the reading section 5.
The figure shows the memory that stores the image data read by the .

Reference numeral 7 indicates a block extraction unit that extracts a block of one character after cutting out a character line from the image data stored in the memory 6, and 8 indicates the relationship between adjacent blocks in the block extracted by the block extraction unit 7. This figure shows a block synthesis unit that performs block synthesis when the inside of a block (in this case, the inside of two blocks) is approximately equal to the average inside of a character indicating the inside of one normal character.

Reference numeral 9 indicates a block accuracy determination unit that determines the accuracy of the composite block synthesized by the block composition unit 8, and 10
indicates a character buffer that stores a block output based on the determination result of the block accuracy determination unit 9 or a single block output without determination processing performed by the block accuracy determination unit 9. The number of blocks output from the block accuracy determination section 9 to the character buffer 10 is the number when character extraction is performed once. 11 is a block accuracy determination unit 9
3 shows a block counter that counts the total number of blocks in one character extraction output from the block counter. 12
indicates a recognition unit that performs character recognition by matching a standard pattern stored in a recognition dictionary unit 13 (to be described later) that corresponds to a character pattern of a block stored in the character buffer 10. Reference numeral 13 indicates a recognition dictionary section storing standard patterns for matching.

Further, 14 has a word buffer, which lists word candidates based on the recognition result in the recognition unit 12, and uses the word dictionary unit 15 (described later) to select word candidates.
The figure shows a word matching unit that matches words with the word dictionary stored in . Reference numeral 15 denotes a word dictionary section that stores a word dictionary for matching with word candidates of recognition results.

Next, the character recognition method of the first embodiment will be explained.

FIGS. 2(a) and 2(b) are flowcharts for explaining the operation of the CPU 2 in the first embodiment, and FIG. 3 is a diagram for explaining word matching in the first embodiment.

First, the document is optically read by the reading section 5, and
It is converted into individual images and stored in the memory 6 (step S
t). Then, the block extracting unit 7 takes a black pixel histogram of the image data stored in the memory 6 in the main scanning direction, that is, in the row direction, and cuts out the character line by using the valley position of the histogram as the cutting position of the character line ( Step S2). For the character line area cut out in this way, a histogram of black pixels is taken in the sub-scanning direction of the line, that is, in a direction perpendicular to the line direction, and the circumscribed rectangle of each character block (hereinafter referred to as block) is calculated. (Step S3).

Next, the following processing procedure is performed sequentially between blocks. First, calculate the total block width of the first block width and the block width of the next second block (Step S4), and compare it with a predetermined average character size set in advance. (Step S5) As a result, when it is determined that the blocks are almost equal (for example, if the difference is within an error of 20%, they are considered equal), it is determined that the two blocks should be combined, and the process proceeds to step S7. On the other hand, the error is 20
%, it is determined that the two blocks should not be combined, and the first block is output to the block accuracy determination section 9 as a single block. In the case of this single block output, the block accuracy determining section 9 stores the output in the character buffer 10 without performing any processing. At the same time, the contents of the block counter 11 are counted up by one and the process proceeds to step S12 (step S6). Note that the initial value of the block counter 11 is 0°.

Further, when the process advances from step S5 to step S7, two adjacent blocks are combined to generate one combined block (step S6).

Then, the accuracy of block synthesis is determined based on the following formula.

If the block accuracy calculated based on the above formula is less than 10%, that is, if the difference between the composite block width and the average character size is smaller than 10% of the average character size, it is determined that the block composition is inaccurate. However, if the block accuracy is 10% or more, it is determined that the block combination is inaccurate. Therefore, when it is determined that the blocks have been combined correctly, each individual block is output to the character buffer 10 in the state of the block before the two blocks were combined (step 5IO), and then the combination of the two blocks is combined. The block is also output to the character buffer in step S12.
(Step 511). In the above step S10, the block counter 11 is counted up by two, and in the following step Sll, the block counter 11 is counted up by one. Therefore, the value of block counter 11 is “3”
becomes. If it is determined that the combination of blocks is accurate, the process directly proceeds to step Sll, where only the combined block is output, and the process proceeds to step Sll. In this case, the value of the block counter 11 becomes "1".

Here, in the example of FIG. 7 explained in the conventional example, “
Since the difference between ``ha'' and ``de'' is less than 10% from the average character middle, individual blocks are not output, and ``3. ” has a difference of 10% or more from the average character, and a single block is output.

Furthermore, in the case of blocks of characters other than "wa", "de", and "3.", the combination is not performed and the blocks are output to the character buffer 10 as individual blocks. In this way, each character is output to the character buffer 10 of only a single block (step S6), (2) to the character buffer 10 of only a composite block, depending on the presence or absence of block composition and the degree of certainty of block composition. The output of (
Character extraction is performed in three ways: (3) outputting the single block and composite block to the character buffer (step S10. step 511). In cases (1) and (2) above, one block is output to the character buffer 10. In the case of (3) above, three blocks (two individual blocks and one composite block) are output to the character buffer 10.

Next, in step Sll, blocks corresponding to the value of the block counter 11 are taken out one by one from the character buffer 10, and the recognition unit 12 performs character recognition using a known technique using the recognition dictionary 13 (step 512). ). The character code of the recognition result is then sent to the word matching section 14,
It is stored in the word buffer in the word matching section 15. however,
If the value of the block counter 11 is "3", step 51 sends the recognition result, that is, the control code to the word buffer in the word matching section 14 before sending the character code.
4), and then sends three character codes (step 515). Here, the recognition unit 12 sends out recognition results in accordance with each of the cases (1), (2), and (3) described above. First, in case (1), one character code of a single block is sent, in case (2), one character code of a composite block is sent, and in case (3), a control code is sent.

Four character codes are sent: the first single block character code, the second single block character code, and the composite block character code. Here, a JIS 2-byte code is used as the character code, and an unused 2-byte code in the JIS code system is used as the control code.

In this way, the above-described process is repeated until one word is accumulated in the word buffer that allows word matching. That is, when the recognition result is sent to the word buffer once in step S15, it is determined in step S16 whether word matching is possible at that point. If it is determined in step S16 that it is not possible, it is first determined whether or not it is necessary to cut out the next line (step 517), and if it is determined that it is necessary to cut out the next line, then While the process is not completed, the process returns to step S2 and repeats the above process (
step 518). If it is determined in step S17 that line cutting is not necessary, the process returns to step S4 and repeats the above-described process.

In this way, when a word with a character code (which may include a control code) is accumulated in the word buffer, the number of control codes in the word buffer is checked, and if it does not contain any control codes, outputs the contents of the word buffer as is as the final recognition result (step S19
．． step 520). Therefore, the contents of the word buffer are output to the outside as a recognition result (step 527). If it is determined in step S19 that the word buffer contains a control code, word matching is performed as follows. First, if there are three character codes following the control code, the first... The character code of the second block indicates the result of the first character recognition without block composition, and the character code of the composite block indicates the result of the second character recognition with block composition. Under these conditions, if the word buffer contains n control foods, there are 2 word candidates.
There are 0 combinations of character strings.

Here, taking the character string "alcohol" as an example, let's use the third
This will be explained using figures. In this case, the above-mentioned (3) applies to the character "ru" as a result of character extraction. That is,
“A rule is represented by a control code and three character codes.

Therefore, since the character string "alcohol" contains two "ru" characters, there are four word candidates. Namely, (a) 7/I, t :l=/1.z(
There are four word candidates: b) 7/L/:1 = ru (c) 7 uni/L/ (d) 7 ru yu = ru. Therefore, all word candidates are matched with the contents of the word dictionary section 15, and the matching word candidates in the word dictionary section 15 are taken as the final recognition result. Therefore, in the case of "alcohol" above, the word candidates are (a) and (b).

(C) is rejected because no matching word exists in the word dictionary section 15. Since the last word candidate (d) has a matching word in the word dictionary section 15, the correct recognition result is "7) Ls 3: ru" in (d). Therefore, as a process, word candidates are listed as many as the number of words to be matched (2''') in step S21, and word matching is performed while checking each match and mismatch (step S22.
Step 523). Here, if the next word candidate is cut off without a matching word, all word candidates are treated as recognition results and step S24. step 525)
, and outputs this recognition result to the outside (step 527).

Further, when a matching word candidate is found in step S23, the matching word candidate is set as the recognition result (step S23).
26). Then, output the recognition results to the outside (step 5)
27).

After outputting the recognition result to the outside in this way, the process returns to step S17 and the above-described process is repeated.

As described above, according to the first embodiment, accurate character recognition can be performed by collating a plurality of characters and eliminating omissions in character recognition candidates.

Description of the second embodiment> Next, a second example will be described.

FIG. 4 is a block diagram showing a second embodiment of the character recognition device according to the present invention, and FIG. 5 is a block diagram showing the CPU of the first embodiment.
2 is a flowchart illustrating the operation of step 2.

In FIG. 4, numeral 21 indicates a character recognition device of the second embodiment. 22 indicates a CPU that controls the entire device. Reference numeral 23 indicates a ROM that stores a control program for operating the CPU 22, an error processing program, and a program according to a flowchart shown in FIG. It shows a RAM used as an area and a temporary save area during error processing. here,
The functions of the respective parts with reference numbers 25 to 31 are the same as those in the first embodiment described above, and therefore the description thereof will be omitted.

32 determines the degree of similarity by comparing the character pattern in the block stored in the character buffer 30 with a standard pattern stored in the recognition dictionary section 33 (described later), and recognizes the character based on the standard pattern with the maximum degree of similarity. This figure shows a similarity calculation unit that lists candidates. 33 is a similarity calculation unit 32
2 shows a recognition dictionary section that stores standard patterns for listing candidates for recognition characters. 34 is the similarity calculation unit 3
2 shows an identification unit that identifies the final recognition result from the recognized character candidates listed in 2.

Here, a character recognition method according to a second embodiment will be explained.

In the second embodiment as well, steps 81 to 81 of the first embodiment
The same process as that up to step Sll is performed at step 81.
- Step s11' is performed. Therefore, explanation of the processing during that time will be omitted.

Therefore, after the block is stored in the character buffer 3o in steps 81 to s11', the value of the block counter 31 is checked (step 530). If the value is "1", the block is a single block or a composite block, so the similarity calculation unit 32 calculates the similarity with the standard pattern in the recognition dictionary unit 33 (step 531). Then, the standard pattern with the highest degree of similarity is set as the recognition result (step 532). The recognition result obtained in this way is obtained by the identification unit 34.
Then, it is output to the outside without any processing (step 5)
42). Further, when it is confirmed in step S30 that the value of the block counter 31 is "3", the similarity calculation unit 32
Now, each block in the character buffer 30 (first single block, second single block, and composite block)
The degree of similarity between the reference pattern and the standard pattern in the recognition dictionary section 33 is calculated. Then, at the time of each calculation, the maximum similarity is determined, and the maximum similarity of the first single block and the maximum similarity of the second single block are combined.

The maximum similarity of the blocks is set to mI and m2°m3, respectively (step S33 to step 838).

Next, the identification unit 34 identifies a standard pattern to be finally output as a recognized character based on the maximum similarity m r + m 2 + m 3. The following formula is used as this identification method. That is, (ml+mz)÷2≦m.

If so, the character code of the standard pattern with the greatest degree of similarity to the single block is set as the final recognition result. In this case, standard patterns corresponding to the degree of similarity m + + m2 are identified as recognition results (steps S39 and 540). Also, if (m+ +m2)÷2〉m3, then the standard pattern with the highest degree of similarity to the composite block is identified as the recognition result. The character code of the large standard pattern is used as the final recognition result. In this case, the similarity is m.

A standard pattern that corresponds to is identified as a recognition result (
step 541).

Next, step S32. Step S40. Then, the recognition results identified by the respective identification processes in step S41 are output to the outside (step 542). After outputting the recognized recognition result to the outside in this way, if the next line does not need to be cut out in order to perform the next character recognition, the process returns to the corresponding step S4 as in the first embodiment (step 543 ), and if the next line needs to be cut out, the process returns to the corresponding step S2 and repeats the process as in the first embodiment until the process for one page is completed (step 544).
.

As described above, according to the second embodiment, even if character recognition is performed based on similarity determination, accurate character recognition can be performed with no omission of candidates for character recognition.

[Effects of the Invention] As described above, according to the present invention, by accurately cutting out characters, accurate character recognition can be performed in a state where there are no omissions in candidates for character recognition.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a first embodiment of the character recognition device according to the present invention, FIGS. 2(a) and (b) are flowcharts explaining the operation of the CPU 2 of the first embodiment, and FIG. 3 4 is a block diagram showing the second embodiment of the character recognition device according to the present invention. FIG. 5 is a diagram illustrating the operation of the CPU 2 in the first embodiment. FIG. 6 is a block diagram showing a conventional character recognition device, and FIG. 7 is a diagram explaining conventional word matching. In the figure, 1.21...Character recognition device, 2,22...C
PU, 3.23...ROM, 4.24...RAM. 5.25.51...Reading unit, 6,26.52...Memory, 7,27.53...Block extraction unit, 8.28
．． 54...Block synthesis unit, 9.29...Block accuracy determination unit, 10,30.55...Character buffer,
11.31...Block counter, 12°56...
Recognition unit, 13.33.57... Recognition dictionary unit, 14.5
8...Word matching section, 15.59...Word dictionary section, 3
2... Similarity calculation unit, 34... Identification unit.

Claims (3)

[Claims] (1) A character recognition device that performs character recognition based on image data, comprising: input means for inputting image data; block extraction means for extracting block data including character pattern data based on the input image data; and the extraction means for extracting block data including character pattern data based on the input image data. a block composition determining means for determining whether or not to perform block composition based on the widths of two adjacent blocks of block data; and a character cutting means for extracting block data based on the determination result of the block composition determining means. , and recognition means for performing character recognition based on the block data cut out by the character cutout means. (2) The recognition means includes a word candidate forming means that forms word candidates based on the result of character recognition, and a word matching means that performs word matching with the word candidates. The character recognition device described. (3) The recognition means includes an identification means that uses a character pattern in the block data as a candidate for a recognized character and identifies the recognized character based on the degree of similarity to a standard pattern stored in advance. Character recognition device according to item 1.