JP3245241B2 - Character recognition apparatus and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition apparatus and method for recognizing characters of a document.

[0002]

2. Description of the Related Art In a conventional character recognition apparatus, a document image is read and input by an input unit, a pattern of one character is cut out from the input image by a preprocessing unit, and geometrical features are extracted from the cut out pattern. Is extracted, and is compared with a standard pattern of a dictionary stored in advance by a recognition unit to recognize a character portion of the document image, and the recognition result is displayed on a recognition result display unit.

[0003] In particular, the pre-processing unit cuts out one character pattern by taking a horizontal or vertical projection or the like one time at a time.

[0004]

However, in the above-mentioned conventional example, it is not possible to correctly cut out vertical and horizontal characters often used in newspapers and the like, that is, a plurality of horizontal characters and the like appearing in vertical writing sentences. For example, the vertical character pattern "58" of "580,000 per year" is recognized as "bright", so the user needs to correct "bright" to "58". Table 1 shows an example of the calculated similarity.
The smaller the calculated values shown in Table 1, the higher the similarity.

[0005]

[Table 1]

An object of the present invention is to solve the above problems and to provide a character recognition device and method capable of recognizing vertical, horizontal and horizontal characters.

[0007]

According to a first aspect of the present invention, there is provided a pre-processing unit for cutting out a character pattern from a document image, and character recognition of the character pattern cut out by the pre-processing unit. What is claimed is: 1. A character recognition apparatus comprising: first character recognition means for calculating a first similarity relating to a candidate character, wherein said determination means determines whether said document image is written vertically or horizontally, and said determination means When it is determined that the writing is vertical writing, a separating unit that separates the character pattern cut out by the pre-processing unit into two or more character patterns horizontally, and character recognition is performed on each character pattern separated by the separating unit. A second character recognizing means for calculating a second candidate character and a second similarity relating to the candidate character;
The first similarity calculated by the character recognition means and the second similarity
Determining means for determining an optimal candidate character by comparing the second similarity calculated by the character recognizing means with the second similarity.

In the first aspect, the separating means can separate the character pattern cut out by the preprocessing means into two or more character patterns by taking a vertical projection. In claim 1, the separating means can separate all the character patterns cut out by the preprocessing means. 2. The apparatus according to claim 1, further comprising a threshold value determination unit configured to determine whether a first similarity of the character pattern cut out by the preprocessing unit is lower than a predetermined threshold value, wherein the separation unit includes the threshold value. Character patterns for which the first similarity is determined to be lower than the predetermined threshold value by the value determining means can be separated.

In a preferred embodiment of the present invention, the apparatus further comprises connectivity determination means for determining whether the connectivity with the character pattern before and after the character pattern cut out by the preprocessing means is higher or lower. When the connectivity is determined to be low by the means, the character pattern cut out by the preprocessing means can be separated. Claim 1
In the above, the separating means can separate the character pattern cut out by the preprocessing means into two or more character patterns horizontally at a predetermined ratio.

In the first aspect, the separating means separates the character pattern cut out by the pre-processing means into two or more character patterns by taking a vertical projection.
A character pattern that could not be separated by vertical projection can be separated by dividing it into two or more character patterns horizontally at a predetermined ratio. In the first aspect, the determining unit can determine an optimal candidate character by comparing the first similarity with the average value of the second similarities of the character patterns separated by the separating unit.

[0011] In the first aspect, the determining means is such that the number of alphanumeric characters of the second candidate character recognized by the second character recognizing means for each of the character patterns separated by the separating means is a predetermined number or more. It is further provided with an alphanumeric number determining means for determining whether or not the second candidate character is an optimal candidate character when the alphanumeric number determining means determines that the number of alphanumeric characters is equal to or greater than a predetermined number. it can.

[0012] In the first aspect, the determining means is such that the number of alphanumeric characters of the second candidate character recognized by the second character recognizing means for each of the character patterns separated by the separating means is a predetermined number or more. Further comprising an alphanumeric number judging means for judging whether or not the first similarity and the second similarity are determined when the number of alphanumeric characters is judged to be a predetermined number or more by the alphanumeric number judging means. Accordingly, the optimum candidate character is determined, and when the number of alphanumeric characters is determined to be less than the predetermined number by the alphanumeric character number determination means, the first candidate character can be determined as the optimal candidate character. 2. The method according to claim 1, wherein the determining unit further includes an alphanumeric determination unit configured to determine whether the character recognized by the second character recognition unit with respect to the character pattern separated by the separation unit is an alphanumeric character. A first operation value is obtained by multiplying the sum of the number of character patterns separated by the separation means and a predetermined value by the first similarity, and the second similarity and alphanumeric characters of all the separated character patterns are obtained. The optimum candidate character can be determined by adding a value determined according to the determination by the determining means to obtain a second calculated value, and comparing the first calculated value with the second calculated value. .

[0013]

According to a twelfth aspect of the invention, there is provided a pre-processing step of cutting out a character pattern from a document image, and character recognition of the character pattern cut out in the pre-processing step. First to calculate similarity
A character recognition method comprising: a character recognition step; a determination step of determining whether the document image is written vertically or horizontally; and a pre-processing step when the determination step determines that the document image is written vertically. Separating the character pattern cut out in step 2 into two or more character patterns in the horizontal direction, and character-recognizing each character pattern separated in the separating step to obtain a second candidate character and a second similarity related to the candidate character. A second character recognition step of calculating a degree, and comparing the first similarity calculated in the first character recognition step with the second similarity calculated in the second character recognition step to determine an optimal candidate character And a determining step.

In the twelfth aspect, in the separating step, the character pattern extracted in the preprocessing step can be vertically projected and separated into two or more character patterns horizontally. In the twelfth aspect, the separating step can separate all character patterns extracted in the preprocessing step. 13. The method according to claim 12, further comprising a threshold value determining step of determining whether the first similarity of the character pattern cut out in the preprocessing step is lower than a predetermined threshold value. A character pattern determined to have a first similarity lower than a predetermined threshold can be separated.

According to a twelfth aspect of the present invention, the image processing apparatus further comprises a connectivity judging step of judging a degree of connectivity with a character pattern before and after the character pattern cut out in the preprocessing step. Character patterns determined to be low can be separated. In the twelfth aspect, in the separating step, the character pattern cut out in the preprocessing step can be horizontally divided into two or more character patterns at a predetermined ratio and separated.

In the twelfth aspect, in the separating step, the character pattern cut out in the preprocessing step is vertically projected and separated into two or more character patterns horizontally, and the character that cannot be separated by the vertical projection is obtained. The pattern can be horizontally divided into two or more character patterns at a predetermined ratio and separated. In the twelfth aspect, in the determining step, an optimal candidate character can be determined by comparing an average value of the first similarity and the second similarity of the character pattern separated in the separating step.

In the twelfth aspect, in the deciding step, the number of alphanumeric characters among the second candidate characters recognized in the second character recognition step for each of the character patterns separated in the separation step is a predetermined number or more. The method further comprises an alphanumeric number determination step of determining whether or not the number of alphanumeric characters is greater than or equal to a predetermined number in the alphanumeric number determination step. it can.

In the twelfth aspect, in the determining step, the number of alphanumeric characters is equal to or greater than a predetermined number among the second candidate characters recognized in the second character recognition step for each character pattern separated in the separation step. The method further comprises an alphanumeric number determining step of determining whether or not the number of alphanumeric characters is equal to or greater than a predetermined number in the alphanumeric number determining step. If the number of alphanumeric characters is determined to be less than the predetermined number in the alphanumeric character determination step, the first candidate character can be determined as the optimal candidate character. 13. The separating step according to claim 12, wherein the determining step further includes an alphanumeric determining step of determining whether a character recognized in the second character recognizing step for the character pattern separated in the separating step is an alphanumeric character. Multiplying the sum of the number of character patterns separated by a predetermined value and the first similarity to obtain a first operation value, and determining the second similarity and the alphanumeric character of all the separated character patterns A second operation value is obtained by adding a value determined in accordance with the determination in step (1), and the first operation value and the second operation value are compared to determine the optimum candidate character.

[0020]

[0021]

Embodiments of the present invention will be described below in detail with reference to the drawings.

First Embodiment FIG. 1 shows a first embodiment of the present invention.

In FIG. 1, reference numeral 1 denotes a character recognition device.
An input unit 2 reads and inputs a document image. Reference numeral 3 denotes a central processing unit which controls each unit of the apparatus 1 according to a control program in the ROM 4. 4 is ROM
The control program and various data of a flowchart described later are stored. Reference numeral 5 denotes a memory, which is used as a work area of the central processing unit 3 and stores an area for storing the document image data input by the input unit 2;
Is provided with an area for storing the position and size of the character pattern cut out by.

Reference numeral 7 denotes a pre-processing unit for cutting out one character pattern of the document image stored in the memory 5.
Reference numeral 8 denotes a recognition unit that extracts a geometric feature from the pattern preprocessed by the preprocessing unit 7 and compares it with a standard pattern stored in the recognition dictionary RAM 9 in advance.
The similarity of the cut-out character pattern to the standard pattern is compared, a character pattern having a high similarity is determined as an optimal candidate character, and character recognition of a document image is performed.

A re-recognition unit 6 further separates and re-recognizes the character pattern cut out by the pre-processing unit 7, that is, compares the character patterns before and after the separation with each other and has a high similarity. The other character pattern is determined as the optimum candidate character, and it is determined whether or not to replace the character pattern with the previous recognition result. A display unit 10 displays an input document image, a recognition result, and the like.

Reference numeral 11 denotes a system bus, which includes a data bus from the central processing unit 3, an address bus, a control signal bus, and the like. Reference numeral 12 denotes an external output device having an interface unit 12 for performing interface control.

FIG. 2 is a flowchart showing an example of a control program stored in the ROM 4 shown in FIG.

In step S 41, the document image set by the operator is read and input by the input unit 2 and stored in the memory 5. Next, in step S42, preprocessing is performed by the preprocessing unit 7, character recognition is performed by the recognition unit 8, and the results, that is, the position and size of the character and the recognition result are stored in the memory 5. Further, the memory 5
Are re-recognized by the re-recognition unit 6 based on the data stored in the display unit 1 and the recognition result is displayed in the display unit 1 in step S44.
Display at 0.

FIG. 3 is a flowchart showing the details of step S43 shown in FIG.

In step S51, it is determined whether the text is written vertically or horizontally. If the result of determination is that the document is horizontally written, this step S43 is ended.
Move to step S52. In step S52, vertical projection is performed only once for each character to separate the characters. For example, "58" shown in FIG.
As shown in (5), it is separated into "5" and "8".

Next, in step S53, it is determined whether or not there are two or more separated characters. If the result of the determination is negative, step S43 is terminated.
On the other hand, if a positive determination is made, the process moves to step S54,
In step S54, the character separated into two or more characters is recognized by the recognizing unit 8, and in step S55, for each re-recognized character, the calculated value of the recognition result before character separation and the recognition result after character separation Is compared with the average calculated value of the above, and a character having a smaller calculated value, that is, a character having a higher similarity is adopted.

For example, the recognition result of the original image "58" before character separation is "bright", and the calculated value is 2863 as shown in Table 1. On the other hand, the recognition results of the character “58” after character separation are “5” and “8”, and their calculated values are 1520 and 1535, respectively, as shown in Table 2, and the average value is 1527. Therefore, the calculated value is small,
That is, "5" and "8" having high similarity are adopted.

[0033]

[Table 2]

Further, as an example of "ka", the recognition result of the original image "ka" before character separation is "ka", and its similarity is 1487 as shown in Table 5. on the other hand,
The document image after the character separation of “K” is separated into the document image shown in FIG. 7, and the recognition results are “A” and “H”.
The similarities are 1654 and 1836, respectively, as shown in Table 6, and the average value is 1745. Therefore, “ka” having a small degree of similarity is adopted.

[0035]

[Table 3]

[0036]

[Table 4]

Second Embodiment This embodiment differs from the first embodiment in the step S43 shown in FIG. That is, the present embodiment will be described with reference to a flowchart (FIG. 8) showing step S43 shown in FIG. 2 in detail. If the result of determination in step S51 is vertical writing, all characters are separated in step S61. Not the calculated value of the recognition result (similarity)
Is vertical-projected only once for characters that are larger than a predetermined threshold to separate the characters in the horizontal direction.
At 3, it is determined whether or not there are two or more separated characters. By doing so, the amount of computation is reduced,
The processing time until the result is displayed is shortened, and it becomes easier to use.

Third Embodiment The present embodiment differs from the first embodiment in that step S43 shown in FIG. 2 is different. That is, a flowchart (FIG. 1) showing the details of step S43 of the present embodiment shown in FIG.
In the case of vertical writing as a result of the determination in step S51, not all characters are separated in step S71 but one or more characters before and after the character of the recognition result are provided in advance. In comparison with the dictionary, only characters having low connectivity are vertically projected only once to perform character separation, and then, in step S53, it is determined whether there are two or more separated characters. I made it.

FIG. 9 shows a part of a dictionary for checking the connectivity of characters. Here, [number] represents a number such as 0-9, zero-9. According to this dictionary, the letters "akira" are followed by "rui", "star", "dark", etc., a number is placed before the letter "million", and after the letter "million" It can be seen that "people", "words", numbers and the like continue.

For example, since the character string before and after "Akira" in the recognition result of "58" for "580,000 a year" does not match the character string in the dictionary, it is determined that the connectivity is low, and the character "58" Will be separated. By doing so, the amount of calculation is reduced, the processing time until the result is displayed is shortened, and the device becomes easier to use.

Fourth Embodiment The present embodiment differs from the first embodiment in the step S43 shown in FIG. That is, this embodiment is different from FIG.
Step S43 shown in FIG. 11 will be described in detail with reference to a flowchart (FIG. 11) showing details. If the result of determination in step S51 is vertical writing, in step S52, a vertical projection is performed only once for each character, and characters are read out. To separate. Next, in step S53, it is determined whether or not there are two or more separated characters. If the determination result is negative, the process proceeds to step S54. If a positive determination is made, step S
Go to 111. In step S111, the character before separation is equally divided in the horizontal direction by α (α is an integer of 2 or more).
At 54, the character separated into two or more characters is recognized by the recognition unit 8, and then the process proceeds to step S55. FIG. 12 shows an example in which the character shown in FIG.
It is shown in (b).

As described above, for example, FIG.
Due to the presence of noise 151 as shown in (a), even if the character cannot be divided into two or more characters, if the character is small, or if the resolution of the input unit 2 shown in FIG. be able to.

For example, the recognition result of the original image "bright" before character separation is "bright", and its similarity is 1586 as shown in Table 3. On the other hand, the original image after the character separation of “Min” is separated into the original image shown in FIG.
The recognition results are "day" and "month". As shown in Table 4, the similarities are 2691 and 1847, respectively, and the average value is 2269. Therefore, “bright” having a small similarity is adopted.

[0044]

[Table 5]

[0045]

[Table 6]

Fifth Embodiment The present embodiment differs from the first embodiment in the step S43 shown in FIG. That is, this embodiment is different from FIG.
Step S43 shown in FIG. 13 will be described with reference to a flowchart (FIG. 13) showing details. In step S54, the character separated into two or more is re-recognized by the recognition unit 8, and in step S121, It is determined whether or not n (n is an integer equal to or greater than 1; here, suppose n = 2) is an alphanumeric character, and as a result, n of the separated characters is determined.
If more than one character is alphanumeric, use separated characters.

Since there are many alphanumeric characters in characters appearing as tate-chu-yoko characters, recognition accuracy is improved.

Sixth Embodiment This embodiment differs from the first embodiment in the step S43 shown in FIG. That is, this embodiment is different from FIG.
Step S43 shown in FIG. 14 will be described in detail with reference to a flowchart (FIG. 14) showing details. In step S54, the character separated into two or more is re-recognized by the recognition unit 8, and in step S131, It is determined whether or not n (n is an integer of 1 or more, here, suppose n = 2) characters or more are alphanumeric characters. If at least n of the separated characters are alphanumeric characters, separation is performed. Compare the average value of the calculated value after and the calculated value before separation, and adopt the character with the smaller calculated value.On the other hand, if the separated characters include non-alphanumeric characters, Adopt letters.

With this configuration, recognition accuracy is further improved.

Seventh Embodiment This embodiment is different from the first embodiment in that step S43 shown in FIG. 2 is different. That is, this embodiment is different from FIG.
Step S43 shown in FIG. 11 will be described in detail with reference to a flowchart (FIG. 15) showing details. In step S54, the character separated into two or more characters is re-recognized by the recognition unit 8. Then, in step S141, it is determined whether or not the separated character is an alphanumeric character. As a result of the determination, when both the separated characters are alphanumeric, θ = x (x is a negative value). ,
If the separated characters include non-alphanumeric characters, θ
= Y (y is a positive value),

[0051]

Α = (calculated value before re-recognition) × (separated number + Z) (Z is a certain value, here, Z = φ) β = (addition of calculated value after re-recognition) + θ . Next, it is determined whether or not α> β. α
If> β, the calculated value decreases as the degree of similarity increases, so the divided characters are used. In a processing system having a higher similarity as the calculated value of the recognition result is larger, x,
Since the sign of y is reversed, the divided characters are used when α <β.

[0052]

As described above, according to the present invention,
With the above configuration, there is an effect that vertical, horizontal and horizontal characters can be recognized.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a flowchart showing an example of a control program stored in a ROM 4 shown in FIG.

FIG. 3 is a flowchart showing in detail a step S43 shown in FIG. 2 of a control program stored in a ROM 4 in the first embodiment.

FIG. 4 is an explanatory diagram illustrating recognition in the first embodiment.

FIG. 5 is an explanatory diagram illustrating character recognition in the first embodiment.

FIG. 6 is an explanatory diagram illustrating character recognition in the first embodiment.

FIG. 7 is an explanatory diagram illustrating character recognition in the first embodiment.

FIG. 8 is a flowchart showing details of step S43 shown in FIG. 2 of the control program stored in the ROM 4 in the second embodiment.

FIG. 9 is a diagram illustrating an example of a dictionary according to the second embodiment.

FIG. 10 is a flowchart showing details of step S43 shown in FIG. 2 of the control program stored in the ROM 4 in the third embodiment.

FIG. 11 is a flowchart showing details of step S43 shown in FIG. 2 of the control program stored in the ROM 4 in the fourth embodiment.

FIG. 12 is an explanatory diagram illustrating character recognition in a fourth embodiment.

FIG. 13 is an explanatory diagram illustrating character recognition in a fifth embodiment.

FIG. 14 is an explanatory diagram illustrating character recognition in a sixth embodiment.

FIG. 15 is an explanatory diagram illustrating character recognition in a seventh embodiment.

[Explanation of symbols]

 Reference Signs List 1 character recognition device 2 input unit 3 central processing unit 4 ROM 5 memory 6 re-recognition unit 7 pre-processing unit 8 recognition unit 9 recognition dictionary RAM 10 display unit 11 system bus 12 interface 13 output device

────────────────────────────────────────────────── ─── Continued on the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06K 9/20-9/62 JICST file (JOIS)

Claims (22)

(57) [Claims] 1. A pre-processing unit for extracting a character pattern from a document image, and character recognition is performed on the character pattern extracted by the pre-processing unit to calculate a first candidate character and a first similarity relating to the candidate character. A character recognition device comprising: a first character recognition unit; a determination unit configured to determine whether the document image is written vertically or horizontally; and when the determination unit determines that the document image is written vertically, The character pattern cut out by the pre-processing means
Separating means for separating into two or more character patterns, and character recognition of each character pattern separated by the separating means,
A second character recognition unit that calculates a second candidate character and a second similarity relating to the candidate character; a first similarity calculated by the first character recognition unit;
A character recognition apparatus comprising: a determination unit that determines an optimum candidate character by comparing the second similarity calculated by the second character recognition unit. 2. The method according to claim 1, wherein the separating unit is configured to:
A character recognizing device, wherein two or more character patterns are horizontally separated by taking a vertical projection. 3. The character recognition apparatus according to claim 1, wherein the separation unit separates all character patterns cut out by the preprocessing unit. 4. The first character pattern extracted by the pre-processing means according to claim 1,
The apparatus further includes threshold value determination means for determining whether the degree of similarity is lower than a predetermined threshold value, and the separation means includes a first threshold value determination means for determining whether the first threshold value is equal to the first threshold value.
A character recognition device for separating a character pattern whose similarity is determined to be lower than the predetermined threshold value. 5. The apparatus according to claim 1, further comprising: connectivity determining means for determining whether the connectivity with a character pattern before and after the character pattern cut out by the preprocessing means is higher or lower. A character recognition device that separates the character pattern cut out by the pre-processing means when the connectivity determination section determines that the connectivity is low. 6. The character recognition apparatus according to claim 1, wherein the separation unit separates the character pattern cut out by the preprocessing unit into two or more character patterns at a predetermined ratio. 7. The method according to claim 1, wherein the separation unit performs a conversion on the character pattern cut out by the preprocessing unit.
By separating vertically into two or more character patterns by taking a vertical projection, by dividing a character pattern that could not be separated by taking a vertical projection into two or more character patterns at a predetermined ratio, A character recognition device characterized by separating. 8. The method according to claim 1, wherein the determining unit compares the first similarity with an average value of the second similarities of the character patterns separated by the separating unit.
A character recognition device, wherein the optimum candidate character is determined. 9. The method according to claim 1, wherein the determining unit determines, for each of the character patterns separated by the separating unit, the number of alphanumeric characters of the second candidate character recognized by the second character recognizing unit. Further comprising an alphanumeric number determining means for determining whether or not the number is equal to or greater than the number. If the alphanumeric number determining means determines that the number of alphanumeric characters is equal to or greater than a predetermined number, the second candidate character is optimized. A character recognition device characterized in that the character is determined as a candidate character. 10. The method according to claim 1, wherein the determining unit determines a predetermined number of alphanumeric characters of the second candidate character recognized by the second character recognizing unit for each of the character patterns separated by the separating unit. Further comprising an alphanumeric number determining means for determining whether or not the number is equal to or more than the first similarity and the second similarity when the number of alphanumeric characters is determined to be a predetermined number or more by the alphanumeric number determining means. By comparing with similarity,
Determining the optimal candidate character, and determining the first candidate character as the optimal candidate character when the number of alphanumeric characters is determined to be less than a predetermined number by the alphanumeric character determination means. Character recognition device. 11. The method according to claim 1, wherein the determination unit determines whether or not the character recognized by the second character recognition unit is alphanumeric for the character pattern separated by the separation unit. An alphanumeric character judging means, wherein the first similarity is multiplied by a value obtained by adding the number of character patterns separated by the separating means and a predetermined value to the first similarity;
An operation value is obtained, and the second value of all the separated character patterns is obtained.
By adding the similarity and a value determined according to the determination by the alphanumeric determination means to obtain a second operation value, and comparing the first operation value and the second operation value, the optimum candidate character is obtained. A character recognition device characterized by determining: 12. A pre-processing step of cutting out a character pattern from a document image, and character recognition is performed on the character pattern cut out in the pre-processing step to calculate a first candidate character and a first similarity relating to the candidate character. A character recognition method comprising: a first character recognition step; a determination step of determining whether the document image is written vertically or horizontally; and if the determination step determines that the document image is written vertically, A separating step of horizontally separating the character pattern cut out in the processing step into two or more character patterns; and character recognition of each of the character patterns separated in the separating step to form a second candidate character and a second character pattern related to the candidate character. A second character recognition step for calculating a second similarity; a first similarity calculated in the first character recognition step; and a second similarity calculated in the second character recognition step. A determination step of determining an optimal candidate character by comparing the degree with the degree. 13. The character recognition method according to claim 12, wherein in the separating step, the character pattern cut out in the preprocessing step is vertically projected and separated into two or more character patterns horizontally. Method. 14. The character recognition method according to claim 12, wherein in the separating step, all character patterns cut out in the preprocessing step are separated. 15. The separating step according to claim 12, further comprising a threshold value determining step of determining whether the first similarity of the character pattern cut out in the preprocessing step is lower than a predetermined threshold value. In the character recognition method, a character pattern determined as having the first similarity lower than the predetermined threshold value in the threshold value determination step is separated. 16. The connectivity determination method according to claim 12, further comprising a connectivity determination step of determining the level of connectivity with a character pattern before and after the character pattern extracted in the preprocessing step. A character recognition method characterized in that a character pattern determined to have low connectivity in a step is separated. 17. The character recognition method according to claim 12, wherein in the separating step, the character pattern cut out in the preprocessing step is horizontally divided into two or more character patterns at a predetermined ratio and separated. . 18. The character pattern extraction method according to claim 12, wherein the separating step takes a vertical projection of the character pattern cut out in the preprocessing step, separates the character pattern into two or more character patterns horizontally, and takes a vertical projection to separate the character pattern. A character recognition method characterized in that a character pattern that cannot be obtained is divided into two or more character patterns horizontally at a predetermined ratio and separated. 19. The optimal candidate character according to claim 12, wherein the determining step compares the average value of the first similarity and a second similarity of the character pattern separated in the separating step. Character recognition method. 20. The method according to claim 12, wherein in the determining step, the number of alphanumeric characters is predetermined for each of the character patterns separated in the separating step in the second candidate characters recognized in the second character recognition step. The method further comprises an alphanumeric number determining step of determining whether or not the number is equal to or greater than the number. If the alphanumeric number determining step determines that the number of alphanumeric characters is equal to or greater than a predetermined number, A character recognition method characterized by determining a candidate character. 21. The method according to claim 12, wherein, in the determining step, the number of alphanumeric characters is predetermined among the second candidate characters recognized in the second character recognition step for each of the character patterns separated in the separation step. The method further comprises an alphanumeric number determining step of determining whether or not the number is equal to or greater than the number. When the alphanumeric number determining step determines that the number of alphanumeric characters is equal to or greater than a predetermined number, the first similarity and the second 2 to determine the optimal candidate character, and if the number of alphanumeric characters is determined to be less than a predetermined number in the alphanumeric character determination step, the first candidate character is converted to the optimal candidate character. A character recognition method characterized in that it is determined as a candidate character. 22. The alphanumeric character according to claim 12, wherein the determining step determines whether or not the character recognized in the second character recognition step for the character pattern separated in the separation step is an alphanumeric character. A determining step of multiplying the first similarity by a value obtained by adding the number of character patterns separated in the separating step and a predetermined value to the first similarity;
An operation value is obtained, and the second value of all the separated character patterns is obtained.
A second operation value is obtained by adding the degree of similarity and a value determined according to the judgment in the alphanumeric judgment step, and the first operation value and the second operation value are compared. Character recognition method.