Character recognition post-processing method

Claim 2

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; and (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b); wherein said step (a) comprises steps of: (1) extracting a plural number of characters forming candidates of a reference character; (2) correcting recognition results of said other characters in said character string by using each of said candidate characters as the reference character; and (3) selecting from said candidate characters, the character most suitable for said reference character, on the basis of correction results for each of the candidate characters.

Claim 3

The character recognition post-processing method as claimed in claim 1, wherein said step of the selection of a character that forms a suitable reference, from each of the candidate characters comprises the steps of, checking by using each of the correction results the reliability of each of the candidate characters, as the reference character, and selecting a candidate character having the highest reliability as the reference character.

Claim 4

The character recognition post-processing method as claimed in claim 2, wherein said step of checking the reliability of each of the candidate reference characters as the reference character comprises the steps of comparing the recognition results for the character string prior to correction with the recognition results for the character string after correction, and checking the number of different recognition results between the recognition results for the character string prior to correction and the recognition results for the character string after correction.

Claim 5

The character recognition post-processing method as claimed in claim 1, wherein said step of selecting from each of the candidate characters, a character that is most suitable as the reference character comprises the steps of, checking contradictions between the correction results of each candidate character, and selecting a most suitable candidate character from among those candidate characters producing contradictory results, as the reference character.

Claim 6

The character recognition post-processing method as claimed in claim 1, wherein said step of selecting from each of the candidate characters, a character that is most suitable as the reference character comprises the steps of, checking contradictions between the correction results of each candidate character, excluding those candidate characters producing contradictory results as being not suitable as a reference character, from the object candidate characters for forming a reference character, by using the correction results for the remaining candidate characters checking the reliability as the reference character of each of the candidate characters, and selecting the candidate character having the highest reliability as the reference character.

Claim 7

The character recognition post-processing method as claimed in claim 5, wherein said step checking the reliability as the reference characters of each of the candidate characters, comprises the steps of comparing the recognition results for the character string prior to correction with the recognition results for the character string after correction, and checking the number of different recognition results between the recognition results for the character string prior to correction and the recognition results for the character string after correction.

Claim 8

The character recognition post-processing method claimed in claim 1, wherein said step extracting a plural number of candidates characters for the reference character comprises the steps of, extracting as candidate characters from a recognized character string, those characters that satisfy predetermined conditions, and extracting as a final candidate characters from those extracted candidate characters, characters satisfying a further limited condition.

Claim 9

The character recognition post-processing method as claimed in claim 7, wherein said step of the selection of a character that forms a suitable reference, from each of the candidate characters comprises the steps of, checking by using each of the correction results, the reliability of each of the candidate reference characters, as the reference character and selecting a candidate character having the highest reliability as the reference character.

Claim 10

The character recognition post-processing method as claimed in claim in 7, wherein said step of selecting from each of the candidate characters, a character that is most suitable as the reference character comprises the steps of, checking contradictions between the correction results of each candidate character, and selecting a most suitable candidate character from among those candidate characters producing contradictory results, as the reference character.

Claim 11

The character recognition post-processing method as claimed in claim 7, wherein said step of selecting from each of the candidate characters, a character that is most suitable as the reference character comprises the steps of, checking contradictions between the correction results of each candidate character, excluding those candidate characters producing contradictory results as being not suitable as a reference character, from the object candidate characters for forming a reference character, checking by using the correction results for the remaining candidate characters the reliability as the reference characters of each of the candidate characters, and selecting the candidate character having the highest reliability as the reference character.

Claim 12

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b); and (d) when a character can not be extracted as the reference character from a character string for which the recognition results should be corrected, extracting a character as the reference character from another string in the document for which recognition results have been obtained.

Claim 13

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b); and (d) when a character can not be extracted as the reference character from a character string for which the recognition results should be corrected, extracting a character as the reference character from another string in the document for which recognition results have been obtained.

Claim 14

The character recognition post-processing method as claimed in claim 11, further comprising the steps of, correcting by using characteristic for the reference character which has been extracted from another character string, the recognition results of another character in a character string, judging by using the correction results, whether or not said extracted reference character is suitable, and if not, making said correction of said recognition results invalid.

Claim 15

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b); (d) presuming a character string type on the basis of corrected recognition results, and (e) further correcting said corrected recognition results so that the correction result is suitable for said presumed character string type obtained in said step (d).

Claim 16

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b); (d) presuming a character string type on the basis of corrected recognition results, and (e) further correcting said corrected recognition results so that the correction result is suitable for said presumed character string type obtained in said step (d).

Claim 17

The character recognition post-processing method as claimed in claim 13, further comprising: judging whether or not a position of a character in a character string is appropriate; and further correcting by using the judgment results, said recognition results.

Claim 18

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b), wherein the recognition results are corrected in the order of from the first reference character to successively adjacent characters; and (d) correcting a recognition result for each character, on the basis of comparison result for a reference position and the position of each character in a perpendicular direction with respect to a line, and (e) correcting the reference position on the basis of a position of each character in a perpendicular direction with respect to a line, and where a position of the reference character in a perpendicular direction with respect to a line is an initial value, said corrected reference position being applied to successively adjacent characters.

Claim 19

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a reference character serving as a reference with respect to correcting the recognition results from a predetermined character string obtained by recognizing a character in the document; (b) comparing a characteristic of said reference character and the same type of characteristic of each of other characters in said character string; (c) correcting recognition results of characters other than said reference character and in said character string based on results obtained in said step (b), wherein the recognition results are corrected in the order of from the first reference character to successively adjacent characters; and (d) correcting a recognition result for each character, on the basis of comparison result for a reference position and the position of each character in a perpendicular direction with respect to a line, and (e) correcting the reference position on the basis of a position of each character in a perpendicular direction with respect to a line, and where a position of the reference character in a perpendicular direction with respect to a line is an initial value, said corrected reference position being applied to successively adjacent characters.

Claim 20

The character recognition post-processing method as claimed in claim 15, wherein said reference position correction step comprises: determining a difference between an applied reference position and a position of a character in a perpendicular direction with respect to a line, determining a new reference position as a position of that character in a perpendicular direction with respect to a line when that difference is not more than a predetermined threshold value, and not correcting the reference position used until then, and proceeding to the next character when that difference is more than the predetermined threshold value.

Claim 21

The character recognition post-processing method as claimed in claim 15, wherein said step of correcting a reference position comprises: determining a difference between an applied reference position and a position of a character in a perpendicular direction with respect to a line, determining a new reference position as a position of that character in a perpendicular direction with respect to a line when that difference is not more than a predetermined threshold value, and determining a position obtained from a predetermined calculation on the basis of the height of a character and the position of that character in the direction perpendicular with respect to a line of that character, as a new reference value when that difference is more than the predetermined threshold value.

Claim 22

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a first reference character serving as a reference with respect to correction of the recognition results from a predetermined character string obtained by recognizing characters in the document, said first reference character having a substantially constant relative height without regard to the fonts which can be used in documents; (b) comparing a height of said first reference character and a height of each of other characters in said character string; and (c) correcting the recognition results of characters other than said first reference character and in said character string based on results obtained in said step (b).

Claim 23

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a first reference character serving as a reference with respect to correction of the recognition results from a predetermined character string obtained by recognizing characters in the document, said first reference character having a substantially constant relative height without regard to the fonts which can be used in documents; (b) comparing a height of said first reference character and a height of each of other characters in said character string; and (c) correcting the recognition results of characters other than said first reference character and in said character string based on results obtained in said step (b).

Claim 24

The character recognition post-processing method as claimed in claim 18, wherein: said step (b) comprises using said comparison to judge an attribute of each of said other characters on the basis of a character height; and said step (c) comprises correcting said recognition results for each of said other characters so that the correction result is suitable for said judged attribute.

Claim 25

The character recognition post-processing method as claimed in claim 18, wherein the recognition results are corrected in the order of from the first reference character to successively adjacent characters.

Claim 26

The character recognition post-processing method as claimed in claim 18, further comprising the steps of: (d) extracting a second reference character serving as a reference with respect to correction of recognition results from a character string obtained in said step (c), said second reference character having a substantially constant relative position perpendicular with respect to a line of said second reference character without regard to the fonts which can be used in documents; (e) comparing a position perpendicular with respect to a line of said second reference character and a position perpendicular with respect to a line of each of other characters in said character string obtained in said step (c); and (f) correcting the recognition results of characters other than said second reference character and in said character string obtained in said step (c) based on results obtained in said step (e).

Claim 27

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a second reference character serving as a reference with respect to correction of recognition results from a predetermined character string obtained by recognizing a character in the document, said second reference character having a substantially constant relative position perpendicular to a line of said second reference character without regard to the fonts which can be used in documents; (b) comparing a position perpendicular to a line of said second reference character and a position perpendicular to a line of each of other characters in said character string; and (c) correcting the recognition results of characters in said character string other than said second reference character based on results obtained in said step (b).

Claim 28

A character recognition post-processing method to be performed after recognition of characters written in a document, the method comprising: (a) extracting a second reference character serving as a reference with respect to correction of recognition results from a predetermined character string obtained by recognizing a character in the document, said second reference character having a substantially constant relative position perpendicular to a line of said second reference character without regard to the fonts which can be used in documents; (b) comparing a position perpendicular to a line of said second reference character and a position perpendicular to a line of each of other characters in said character string; and (c) correcting the recognition results of characters in said character string other than said second reference character based on results obtained in said step (b).

Claim 29

The character recognition post-processing method as claimed in claim 22, wherein: said step (b) comprises using said comparison to judge an attribute of each of said other characters on the basis of the position of a character; and said step (c) comprises correcting said recognition results for each of said other characters so that the correction result is suitable for said judged attribute.

Claim 30

The character recognition post-processing method as claimed in claim 22, further comprising: (d) extracting a first reference character serving as a reference with respect to correction of the recognition results from a character string obtained in said step (c), said first reference character having a substantially constant relative height without regard to the fonts which can be used in documents; (e) comparing a height of said first reference character and a height of each of other characters in said character string obtained in said step (c); and (f) correcting the recognition results of characters other than said first reference character and in said character string obtained in said step (c) based on results obtained in said step (e).

Claim 31

The character recognition post-processing method as claimed in claim 22, wherein: the recognition results are corrected in an order starting from the second reference character to successively adjacent characters.