JPH07262322A - Method and device for recognizing character - Google Patents

Abstract

PURPOSE:To provide the method for recognizing character with which even a word not registered on a dictionary yet can be properly recognized with a high recognition rate. CONSTITUTION:A character string image stored in a character string image storage part 101 is divided into minimum unit areas by a character string dividing part 102, all the combination to be recognized as one character at least is detected among the combination of adjacent unit areas, and a character pattern corresponding to the detected combination of unit areas and character patterns corresponding to the respective unit areas are generated. Next, a position range detecting part 104 detects a position range to be taken by each character pattern inside a word to be the recognized result and all the words provided with matched characters inside the position range to be taken by candidate characters are selected as candidate words. Afterwards, suitable words are selected out of the respective candidate words based on a prescribed evaluated value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition technique, and more particularly to a technique for correcting a recognition result. More specifically, when character recognition is performed by an optical character reader (OCR), etc., the recognition error of the character recognition result of the character string entered in various application documents, sales slips, delivery slips, and other forms is corrected. Technology.

[0002]

2. Description of the Related Art Character recognition techniques are required in many fields, and in particular, techniques for correcting character recognition results in order to improve recognition accuracy are being studied. For example, JP-A-6
3-103393 (word recognition device), JP-A-63-63
Japanese Patent Laid-Open No. 121989 (word reading method) and the like disclose a method of correcting a character recognition result for a document image whose character pitch is not constant.

In these methods, first, a character is cut out in units of one character based on a shape feature of a circumscribing rectangle of the character, and the most probable character pattern is selected to recognize the character.
Next, search for the corresponding word from the word dictionary by allowing the character position deviation and the character error in the candidate character string obtained as a result of the character recognition, and compare each searched word with the character recognition result candidate character. I do. In addition, the part where the character is cut out by mistake is detected, the character is cut out and the character is recognized again, the words that match the candidate characters are extracted, and the size of the value obtained by finally accumulating the evaluation value of the character recognition is calculated. The candidate word is decided according to.

[0004]

According to the above prior art, it is possible to obtain a certain degree of recognition accuracy. However, in such a method, in the first character segmentation, only one character pattern is created, character recognition and word search are performed, and the optimum word is selected from the results obtained by the word search. Therefore, if the first character segmentation is incorrect and the correct word is not included in the words obtained as a result of the word search, there was a correction error.

Regarding the character string to be read,
When the character strings can be covered in the word dictionary like address place names, it was possible to perform word matching by registering them in the word dictionary. On the other hand, when reading an unspecified number of company names, building names, etc., it is not possible to register all the character strings that appear, so when recognizing unregistered words in the dictionary, the conventional method There was a problem that the entire character string could not be collated or could be erroneously corrected.

The present invention has been made in view of the above background, and has a high recognition rate and is capable of performing correct character recognition even for a word that is not registered in a dictionary, and a character recognition that realizes this method. The purpose is to provide a device.

[0007]

The character recognition method of the first invention provided by the present invention is a method of dividing a character string image into unit areas and at least one character from a combination of the divided unit areas or adjacent areas. Detecting one or more area combinations that can be recognized as, and detecting character patterns corresponding to the detected area combinations and character patterns respectively corresponding to the unit areas included in the area combination; For each character pattern, a step of detecting at which character position each character pattern can be placed in the recognition target word, a step of identifying candidate characters that can be recognized for each character pattern, and a step of specifying each candidate character , A step of identifying a word having a character that matches the candidate character at a position that the original character pattern can take as a candidate word, and each candidate word according to a predetermined evaluation criterion A determining a word represented by al the character string image.

A character recognition method of a second invention provided by the present invention is the character recognition method of the first invention, wherein a character evaluation value representing a degree of approximation between each character pattern and the corresponding candidate character is generated. And a candidate character that matches the position of the character in the candidate word is detected by comparing the character forming the candidate word with the candidate character, and from the character evaluation value corresponding to the detected candidate character. The method includes the steps of generating respective word evaluation values, and evaluating each candidate word based on the generated word evaluation value to determine the word represented by the character string image. When the candidate word is evaluated, the optimum word is preferably determined by the number of matching characters with respect to the character string length and the cumulative value of the matching character evaluation values.

In the character recognition method of the third invention, a step of dividing the character string image into unit areas, and a step of cutting out a recognition target area at a predetermined position from the character string image using the unit area as a minimum unit, The step of performing character recognition processing for each of the cut-out recognition target areas to determine a candidate word, the step of detecting a corresponding area in the original character string image for each of the determined candidate words, Corresponding to the combination of candidate words, selecting a combination that minimizes the overlapping portion between the determined corresponding areas and the gap portion between the corresponding areas in the character string image among the combinations, and the selected combination Determining a combination of candidate words as a character represented by the character string image.

In the character recognition method of the fourth aspect of the invention, as a preferred example of the character recognition method of the third aspect of the invention, the step of cutting out the recognition target area at a predetermined position is such that each of the plurality of divided unit areas is set as the head. It is characterized in that a combined area of a plurality of adjacent unit areas is cut out as a recognition target area.

Further, the character recognition device of the fifth invention provided by the present invention comprises: a first memory storing a character string image to be recognized; a second memory storing a recognition candidate character group;
An image area dividing means for dividing the character string image stored in the first memory into unit areas, and a divided unit area or an adjacent area. Area feature detection means for detecting an area combination that can be recognized as at least one character from the combination of the character combinations, and a character pattern corresponding to the detected area combination and a character pattern corresponding to a unit area included in the area combination. A character pattern generation means for generating the character pattern, a character position range detection means for detecting which character position of the recognition target word the generated character pattern can take, and a candidate corresponding to each of the character patterns. Candidate character selecting means for selecting a character from the second memory, and all characters having matching characters within the possible position range of the selected candidate character A first candidate word selecting means for selecting a word from the third memory; and a word determining means for determining a word represented by the character string image from the selected candidate word according to a predetermined evaluation criterion. Characterize.

Further, the character recognition device according to the sixth aspect of the present invention includes at least a first memory storing a character string image to be recognized and a third memory storing a recognition candidate word group, and further, the third memory. An image area dividing unit that divides a character string image in one memory into unit areas, an image area cutting unit that cuts out a predetermined recognition target area from the character string image with the unit area as a minimum unit, and the cut-out unit. Second candidate word selecting means for performing character recognition processing for each recognition target area and selecting all candidate words from the second memory;
A region in which the corresponding region in the original character string image is detected for each of the selected candidate words to generate a plurality of region combinations, and the overlap between detected corresponding regions and the gap between the corresponding regions are minimized. It is characterized by comprising area combination specifying means for specifying a combination, and compound word determining means for determining a combination of candidate words corresponding to the specified combination as a combination of words represented by the character string image.

The region combination in which the overlapping portions of the corresponding regions and the gap portion between the corresponding regions are minimized is specified as follows, for example. First, the coordinates are set for the original character string image, and the distances between the words are derived by referring to the coordinate positions of the character patterns corresponding to the first character and the terminal character of each candidate word. Next, the distance of the overlapping portion and the distance of the gap portion of the character pattern corresponding to each word are examined, and if the value of this distance is within the allowable range, the word connection is recognized. The words that can be connected by this method are combined into a compound word. In particular, when performing recognition processing using a predetermined word evaluation value, the optimum compound word is determined by the number of matching characters with respect to the character string length of the compound word and the cumulative value of the matching character evaluation values.

[0014]

According to the character recognition method of the first aspect of the present invention, when a character pattern is cut out from a character string image, a character pattern corresponding to each unit area and a character pattern corresponding to a combination of unit areas that can be regarded as at least one character. Since and are generated, all the character patterns that can be generated from the original character string are generated, and the correct character pattern to be recognized is also included in this generated character pattern. Since the correct character pattern is not missed in this way, the probability of selecting a character to be recognized as a candidate character to be selected is increased, and therefore the character recognition accuracy is increased. In particular, as in the second invention, generating a character evaluation value indicating how close each character pattern and the corresponding candidate character are, and using this character evaluation value to generate an evaluation value of each candidate word. Then, the similarity between each candidate word and the original character string image is expressed as an objective numerical value.

In the character recognition method of the third invention, a plurality of recognition target areas are generated from the original character string image, and the candidate words selected for each recognition target area are combined, so that the character string image is Recognized as a combination of multiple words. This allows the character string image to be recognized as a compound word composed of a plurality of words, whereas the character string image can only be recognized as a single word in the past.

In particular, as in the fourth aspect of the present invention, this character string image is constructed by cutting out a combined area of a plurality of adjacent unit areas each having a plurality of divided unit areas as a head area as a recognition target area. All possible words are detected. In addition, by checking the distance between overlapping portions and the distance between gaps in the character pattern, and if the value of this distance is within the allowable range, it is possible to connect words so that the same character is used in duplicate for two or more words. , Logically contradictory combinations are eliminated.

Further, in the character recognition device of the fifth invention, in order to carry out the character recognition method of the first invention, the character string image to be recognized is in one memory, the recognition candidate character group is in two memory, and the recognition candidate character group is recognized. The candidate word group is stored in the third memory,
The image dividing unit divides the character string image in the first memory into unit regions, and the region feature detecting unit detects a region combination that can be recognized as at least one character from the divided unit regions or combinations of adjacent regions. Then, a character pattern corresponding to each area combination and a character pattern corresponding to a unit area included in the area combination are generated based on each area feature, and each character pattern determines the position of the character of the recognition target word. The character position range detecting means detects whether or not to obtain. After that, the candidate character selecting means selects a candidate character corresponding to each character pattern from the second memory, and selects all candidate words having a matching character within the possible position range of the selected candidate character as the first candidate. The word selecting means selects from the third memory. Then, the word determining means determines the word represented by the character string image from these candidate words according to a predetermined evaluation criterion.

In order to carry out the character recognition method of the third invention, the character recognition device of the sixth invention stores the character string image to be recognized in the first memory and the recognition candidate word group in the second memory. The character string image in the first memory is divided into unit areas by the image area dividing means. Then, a predetermined recognition target area is cut out from the character string image using the unit area as a minimum unit, and further, the second candidate word selection means performs character recognition processing for each recognition target area to extract all candidate words from the second memory. select. The area specifying unit detects a corresponding area in the original character string image for each of the selected candidate words and generates a plurality of area combinations. Then, an area combination that minimizes the overlap between corresponding areas and the gap between corresponding areas is specified. The compound word determining means determines a combination of candidate words corresponding to the specified combination as a combination of words represented by the character string image.

[0019]

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

(First Embodiment) FIG. 1 is a functional block diagram of a character recognition apparatus according to an embodiment of the present invention. 10 in the figure
Reference numeral 1 is a character string image storage unit (first memory) that stores a character string image (document line image) input from an image scanner or the like, and 102 is a character string division that divides the character string image into minimum unit areas. A unit (image area dividing unit) 103 detects all combinations of adjacent unit areas that can be recognized as one character, and uses image features to detect a character pattern corresponding to the detected combination of unit areas. And a character pattern generation unit (region characteristic detection means, character pattern generation means) for generating a character pattern corresponding to each unit area. Position range detecting section (character position detecting means) for detecting all possible positions
Reference numeral 05 denotes a candidate character storage unit (second memory) for storing the recognition candidate character and its standard pattern.

Reference numeral 106 is a candidate character for comparing each character pattern in the position range detected by the position range detecting unit 104 with a standard pattern in the candidate character storage unit 105 and selecting a candidate character for each character pattern. The selection unit (candidate character selection means) generates a character evaluation value that is an index of how close the character pattern and the standard pattern are to each other for each selected candidate character. The candidate character selected in this way, the possible position range thereof, and the corresponding character evaluation value are output. In this embodiment, the character evaluation value is set to increase as the character pattern and the standard pattern become closer to each other.

Reference numeral 107 denotes a candidate word storage unit (third memory) for storing a word table in which character codes are associated with the recognition candidate word groups and character positions for words registered in a word dictionary (not shown). Reference numeral 108 denotes a candidate word selection unit (candidate word selection unit) that detects a character code corresponding to a candidate character obtained from the candidate character selection unit 106. Each candidate in the candidate word selection unit 107 using the word table. From the words, all words having matching characters within the possible position range of the candidate character are selected as candidate words.

109 detects a character matching each of the above candidate characters for each character in each selected candidate word, and whether the detected character is within the position range of the original candidate character. Is a word evaluation value calculation unit that detects a character satisfying these conditions for each candidate word, and adds the evaluation value of the corresponding candidate character as the score of the word, and obtains the total sum value. Is the word evaluation value. Reference numeral 110 denotes a suitable word determination unit that selects a suitable word from the selected candidate words based on the word evaluation value (above word determination means).

Here, since the word evaluation value calculation unit 109 only detects whether or not the detected character is within the position range of the original candidate character for the word candidate, 1
Multiple character candidates may be assigned to one character position. In addition, the character order of the candidate words may be reversed from the character order of the original cut-out pattern. As described above, since there is a possibility that the evaluation value of the candidate character having a logical contradiction may be added to the word evaluation value, the character position of the searched word candidate is again determined. Decide the combination of cut-out patterns that gives the largest number of matching characters without duplication or inversion. Then, the evaluation values of the matched candidate characters are summed to obtain a true word evaluation value, and a suitable word is selected based on that value. In this embodiment, one candidate word having the highest score is selected as a suitable word.

At this time, the true word evaluation values may be obtained for all the candidate words, but in this embodiment, in order to simplify the processing, the original word evaluation values having a high score are selected. A predetermined number was selected in order, and a true word evaluation value was obtained for the selected candidate words.

Next, a specific example of recognition processing of the character string image "Kawaguchi City Shiba Hinosame" will be described. Character string dividing unit 102
2 shows a state in which the character string image is divided into unit areas. In this example, the smallest circumscribed rectangular area is used as the unit area. The character pattern generation unit 103 generates a character pattern from the unit area shown in FIG. The position range of each character pattern is detected as described above, and the candidate character and the character evaluation value are detected. FIG. 3 is an example of each character pattern generated in this way, the corresponding candidate character, and its character evaluation value.

As shown in FIG. 3, each unit pattern overlaps each character pattern. For example, the “tree” (tree bias) in the “gutter” is the cutting pattern 13, 14, 15 in FIG.
Is used in duplicate. Then, according to the candidate character and its position range, the candidate word selection unit 108 and the word evaluation value calculation unit 109 obtain the candidate word and the word evaluation value.
An example of the obtained word and word evaluation value is shown in FIG.

FIG. 5 is an explanatory diagram of a process of generating a true word evaluation value in the suitable word determining unit 110. In this figure,
It shows a state in which a predetermined number of candidate words having a high word evaluation value are selected, and character patterns having candidate characters that match each character of each candidate word are indicated by numbers.

Regarding the candidate word “Kawaguchi City Shiba Hinotsume”,
A combination using pattern 18 and pattern 21 has been obtained. However, since the unit area corresponding to the pattern 18 is used for the pattern 21, both patterns overlap. Therefore, a combination of “3, 11, 12, 15, 18, 24” is obtained as a combination of cut-out patterns that avoids this. Next, the evaluation values of the candidate characters corresponding to each cutout pattern are summed.

Referring to FIG. 3, the character evaluation value of "kawa" in pattern 3 is 676, the character evaluation value of "city" in pattern 11 is 757, and the character evaluation value of "shiba" corresponding to pattern 12 is 639. Since the character evaluation value of “Grain” in pattern 15 is 628, the character evaluation value of “No” in pattern 18 is 629, and the character evaluation value of “nail” corresponding to pattern 24 is 622, the values of these character evaluation values are The word evaluation value 3996 is obtained by summing.

In this way, for each candidate word, the optimum combination of character patterns is found and the score of the word is found again. FIG. 6 shows the true word evaluation for each candidate word. Finally, the candidate word with the highest score among them, "Kawaguchi City Shiba Hinotsume"
Is selected as the preferred word.

As described above, according to this embodiment, a plurality of character patterns are generated, so that the recognition accuracy is improved. Further, when the true word evaluation value is obtained for all the candidate words, the amount of processed data becomes very large, a large memory may be required, and the calculation time may be long. In this case, since a predetermined number of original word evaluation values are selected in order, and the true word evaluation value is calculated for the selected candidate words, character recognition is performed in a short time with a simple configuration. In addition to the above, it is possible to obtain almost the same result as when the true word evaluation value is obtained for all candidate words.

When the word evaluation value and the true word evaluation value are obtained as described above, a candidate word having a large number of characters tends to have a high word evaluation value even if the character evaluation value in the candidate word is low. . Therefore, a value obtained by dividing the sum of the character evaluation values by the number of characters of the candidate word may be used as the word evaluation value or the true word evaluation value. When such an evaluation is performed, the evaluation value of the word in which the ratio of the characters matching the candidate character is small among the characters in the candidate word is low.

(Second Embodiment) As in the first embodiment,
A word representing a place name or the like can be registered in a dictionary in advance, and a word corresponding to the entire character string image can be easily selected. However, when recognizing a compound word, the first word of the character string image can be selected. Sometimes you can only get words that correspond to a few letters. For example, if the words "2nd", "Kawashima" and "Bill" called "2nd Kawashima Building" are registered separately in the dictionary, the character string image of the compound word "2nd Kawashima Building" will be recognized. Even if the word “second” can be recognized in the row image, it cannot be recognized with respect to the character string image corresponding to the subsequent “Kawashima Building”. Second
In the embodiment, a character recognition device capable of recognizing such a compound word will be described.

FIG. 7 shows a functional block diagram of the character recognition apparatus in the second embodiment. In FIG. 7, the same functional blocks or corresponding portions as those in FIG. 1 are designated by the same reference numerals. The character string image storage unit 101, the character string dividing unit 102, and the character pattern generating unit 103 are blocks that perform the same processing as in the first embodiment.

A collation start position setting unit 701 sets a recognition target area in the character string image with each unit area as a start position, and performs character recognition processing on the original character string image and each recognition target area. Perform (image area cutting means).

More specifically, the processing of the position range detection unit 104 and thereafter is performed with the first unit area as the start position.
Therefore, the original character string image is directly processed for recognition. Next, the start position is set to the second cutout pattern from the top, and all cutout patterns other than the top cutout pattern are output to the position range detection unit 104. Therefore, the image obtained by removing the area corresponding to the first unit area from the original character string image becomes the recognition target area. Hereinafter, similarly, the collation start position is sequentially shifted, and all the cutout positions are set as the collation start positions.

After that, the position range detection unit 104 to the word evaluation value calculation unit 109 perform the same processing as that of the first embodiment on each recognition target area. The preferred word determination unit 110 performs the same processing as in the first embodiment to generate a true word evaluation value. However, although only one preferred word is selected in the first embodiment, this embodiment also allows a plurality of words to be selected as preferred words when a predetermined evaluation criterion is satisfied. Therefore, although one word may be selected as a suitable word, a plurality of words may be selected. In this way, a suitable word is determined for each matching start position.

Reference numeral 702 is a compound word determining unit which refers to the coordinate positions of the character patterns corresponding to the leading character and the terminal character of each of the selected preferred words and calculates the distance between the candidate words. Then, based on the result, the overlap and spacing of the character patterns corresponding to each word are investigated, and if the value of this distance is within the allowable range, the word connection is recognized. By combining words that can be connected by this method into a compound word, a suitable compound word is determined based on the number of matching characters with respect to the character string length of the compound word and the cumulative value of the evaluation values of the matching characters (compound word determining means).

Next, an example of the recognition processing of the character string image "2nd Kawashima Building" will be described. FIG. 8 shows a state in which the character string dividing unit 102 divides the character string image into unit areas. In this example, the smallest circumscribed rectangular area is used as the unit area.

The character pattern generator 103 generates a character pattern from the unit area shown in FIG. 8 as in the first embodiment. FIG. 8 shows each of the generated character patterns, the corresponding candidate characters, and their character evaluation values. The position range of each character pattern is detected as described above, and the candidate character and the character evaluation value are detected.

In the collation start position setting unit 701, the position range detecting unit 10 first sets the leading cutout pattern 801 as the start position, that is, with respect to the entire cutout pattern.
4 to the word evaluation value calculation unit 109 and the suitable word determination unit 110. As a result, the suitable words “second” and “Yoshikawa” are obtained. Next, the collation start position is set in the second cutout pattern 802 from the beginning, and the same processing is performed to obtain "Kawashima", "Kawacho", and "Kobashi". Less than,
Similarly, the collation start position is advanced to other cutout patterns 803 to 806 to obtain a plurality of candidate words.

The compound word determining unit 702 refers to the coordinate positions of the character patterns corresponding to the first character and the last character of the candidate word to find the distance between the words, checks the connectivity of the words, and connects the words. Decide whether or not In this example, the conditions under which a certain word A and a certain word B can be connected were obtained as follows. First, the words A and B are selected so that the leftmost x coordinate of A ≦ the leftmost x coordinate of B, and Gap is determined as follows. The left end x-coordinate of Gap = B-the right end x-coordinate of A Then, when the following formula is materialized, the word A and the word B shall be connected. -(Limit of overlap of cutout pattern) <Gap <Remaining limit of cutout pattern

For example, in FIG. 10, "second" and "yoshikawa" cannot be connected because the overlapping portion is too large. Further, the “second” and “building” cannot be connected because the gap is too wide, but the “second” and “Kawashima” can be connected. The "Kawashima" part has two suitable words, but "Kawashima" is selected according to the score. As a result, “second”, “Kawashima” and “building” are selected. In this way, a compound word is created by combining the connectable words, and a suitable compound word is obtained from the number of matching characters and the cumulative value of the evaluation values.

[0045]

As is apparent from the above description, according to the present invention, since all the character patterns that can be generated from the original character string are generated, the character patterns to be recognized are included therein. The probability increases. Therefore, the character recognition accuracy is increased.

In particular, by generating the evaluation value of each candidate word based on the character evaluation value showing how close each character pattern and the corresponding candidate character are, it is possible to obtain each candidate word and the original character string image. Can be expressed as an objective numerical value. Furthermore, by selecting the word with the highest degree of similarity, it is possible to increase the recognition accuracy.

Since the character string image can be recognized as a compound word formed by combining a plurality of words,
Even if the entire compound word is not registered in the dictionary, if the individual constituent words are registered in the dictionary, the entire compound word can be recognized. All words that may make up the string image are detected. Further, the written compound word can be recognized even if the boundaries between the words forming the compound word are unknown and the cut-out position of each character is not clear.

[Brief description of drawings]

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

FIG. 2 is an explanatory diagram showing a state in which a character string image is divided into unit areas.

FIG. 3 is an explanatory diagram of generated character patterns.

FIG. 4 is an explanatory diagram of candidate words and their evaluation values.

FIG. 5 is an explanatory diagram of a method of calculating a true word evaluation value.

FIG. 6 is an explanatory diagram of a true evaluation value as a candidate word.

FIG. 7 is a functional block diagram of a character recognition device in a second embodiment.

FIG. 8 is an explanatory diagram of a state in which a character string image is divided into unit areas.

FIG. 9 is an explanatory diagram of character patterns.

FIG. 10 is an explanatory diagram of processing in a compound word determination unit.

[Explanation of symbols]

 101 character string image storage unit 102 character string division unit 103 character pattern generation unit 104 position range detection unit 105 candidate character storage unit 106 candidate character selection unit 107 candidate word storage unit 108 candidate word selection unit 109 word evaluation value calculation unit 110 suitable words Determining unit 701 Matching start position setting unit 702 Compound word determining unit

Claims (6)

[Claims] 1. A step of dividing a character string image into unit areas, and one or more area combinations that can be recognized as at least one character are detected from a combination of the divided unit areas or adjacent areas, and further detected. The step of detecting the character pattern corresponding to the area combination and the character patterns corresponding to the unit areas included in the area combination, and the character number of each character pattern in the recognition target word for each detected character pattern. , The step of detecting the candidate character that can be recognized for each character pattern, and the step of identifying the candidate character that can be recognized for each character pattern and the character that matches the candidate character at the position that the original character pattern can take. Is specified as a candidate word, and a word represented by the character string image is determined from each candidate word according to a predetermined evaluation criterion. Character recognition method which is characterized the door. 2. The character recognition method according to claim 1, wherein each character pattern and a character evaluation value representing a degree of approximation of the candidate character corresponding to each character pattern are generated; and a character forming the candidate word. Generating a word evaluation value from a character evaluation value corresponding to the detected candidate character by comparing the candidate character with a candidate character matching the position of the character in the candidate word; And evaluating each candidate word based on the evaluated word evaluation value to determine the word represented by the character string image. 3. A step of dividing a character string image into unit areas, a step of cutting out a recognition target area at a predetermined position from the character string image with the unit area as a minimum unit, and a step of cutting out each of the cut recognition target areas. Performing a character recognition process to determine each candidate word; detecting each corresponding candidate region in the original character string image for each of the determined candidate words; combining the determined candidate words; Selecting a combination that minimizes the overlap between the determined corresponding areas and the gap between the corresponding areas in the character string image, and the combination of candidate words corresponding to the selected combination is the character string image. And a step of determining as a character represented by. 4. The character recognition method according to claim 3, wherein the step of cutting out the recognition target area at the predetermined position is a combined area of a plurality of adjacent unit areas each having a plurality of divided unit areas as a head area. A character recognition method characterized by cutting out each as a recognition target area. 5. A first memory storing a character string image to be recognized, a second memory storing a recognition candidate character group, and a third memory storing a recognition candidate word group,
Further, an image area dividing unit that divides the character string image stored in the first memory into unit areas, and an area combination that can be recognized as at least one character is detected from a combination of the divided unit areas or adjacent areas. Area characteristic detection means, character pattern generation means for generating a character pattern corresponding to the detected area combination and a character pattern corresponding to a unit area included in the area combination based on the respective area characteristics, and A character position range detecting means for detecting which character position of the recognition target word the character pattern can take, and a candidate character selecting means for selecting a candidate character corresponding to each of the character patterns from the second memory. , Selecting, from the third memory, all candidate words having a matching character within the possible position range of the selected candidate character A candidate word selection means, a character recognition apparatus characterized by having a word determining means for determining a word represented by the character string image from the candidate words said selected according to a predetermined criterion. 6. A first memory that stores a character string image to be recognized, and a third memory that stores a recognition candidate word group, and further, a character string image in the first memory. An image area dividing unit that divides each unit area, an image area cutting unit that cuts out a predetermined recognition target area from the character string image with the unit area as a minimum unit, and a character recognition process for each cut recognition target area. Second to go and select all candidate words from the third memory
Of the candidate words and the corresponding regions in the original character string image for each of the selected candidate words are detected to generate a plurality of region combinations, and the overlapping portions of the detected corresponding regions and the corresponding regions Area combination specifying means for specifying an area combination with a minimum gap, and compound word determining means for determining a combination of candidate words corresponding to the specified combination as a combination of words represented by the character string image. A character recognition device characterized by the above.