Document recognition method and system

• US 5,717,794 A
• Filed: 10/04/1996
• Issued: 02/10/1998
• Est. Priority Date: 03/17/1993
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Koga, Masashi, Marukawa, Katsumi, Nakashima, Kazuki, Shima, Yoshihiro
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): SHALWALA, BIPIN H
• Application Number:
  US08/725,477
• Time in Patent Office: 494 Days
• Field of Search: 382/173, 382/229, 382/302, 382/276, 382/309, 382/310, 382/311, 382/317, 382/177, 382/203, 382/190, 364/737
• US Class Current: 382/309
• CPC Class Codes: G06V 10/987 with the intervention of an operator
• CPC Class Codes: G06V 30/40 Document-oriented image-based pattern recognition

Abstract

A document recognition system includes various recognitive steps for document recognition, and various correctional steps corresponding to the recognitive steps. The operation modes of the system are a "sequential mode" (sequential recognition operation) in which the respective recognitive steps are executed in "step by step" fashion, an "auto mode" (batch recognition operation) in which all the recognitive steps are collectively executed, and a "retry mode" (re-recognition operation) in which the execution of any of the correctional steps is automatically followed by the execution of a necessary one of the recognitive steps. In the "sequential mode", any of the recognitive steps having been executed can be shifted, not only to the correctional step corresponding to the executed recognitive step, but also to the correctional step preceding the execution. When an error is involved in a recognized result because the limits of a character line recognized by the system are incorrect, the user of the system corrects the limits of the character line. Then, the system can execute the necessary recognitive step again based on the corrected limits of the character line.

Claims

1. A document recognition system wherein characters on a document are recognized from an image of the document, and at least the recognized characters are displayed on a display unit, comprising:

input means for entering the document image which is an object to be recognized;

character line extraction means for obtaining character line data which indicate limits of a character line formed of a character string, including either of a frame and a block surrounding said character string, on said document image entered by said input means, wherein said frame and said block indicate a set character line displayed as a polygon in superposition of the document image by utilizing graphics;

character line correction means for presenting the limits of the character line obtained by said character line extraction means to a user of said document recognition system, superimposed on said document image on said display unit, and for correcting said limits of said character line in accordance with an instruction given by the user;

character segmentation means for deriving character patterns of the individual characters contained in said character line, from said document image on the basis of the character line data;

character recognition means for recognizing the character patterns derived by said character segmentation means, and for converting the recognized character patterns into respectively corresponding character codes; and

display means for causing said display unit to display the character codes of said characters recognized by said character recognition means.

2. A document recognition system as defined in claim 1, wherein said character line correction means makes the correction so that two character lines displayed adjacent to each other on said display unit may be connected into a single character line.

3. A document recognition system as defined in claim 1, wherein said character line correction means makes the correction so that a single character line displayed on said display unit may be separated into two character lines.

4. A document recognition system as defined in claim 1, further comprising skew alteration means for detecting a skew of said document image on the basis of said character line data extracted by said character line extraction means, and for subjecting said document image to rotation processing on the basis of the detected skew, so as to alter said skew.

5. A document recognition system as defined in claim 1, further comprising layout analysis means for recognizing blocks, each including a plurality of character lines, and for determining a sequence in which a the plurality of blocks are read, wherein said character segmentation means and said character recognition means execute the respectively corresponding processing for the individual character lines in the reading sequence of the plurality of blocks.

6. A document recognition system as defined in claim 5, further comprising layout correction means including first means for presenting limits of blocks recognized by said layout analysis means superimposed on said document image on said display unit, and second means for altering the presented limits of said blocks in accordance with an alteration instruction given by the user.

7. A document recognition system as defined in claim 6, wherein said layout correction means further includes third means for altering the reading sequence of the plurality of blocks.

8. A document recognition system as defined in claim 1, further comprising language processing means for recognizing any inappropriate phrase in text data which includes of said character codes of the character string recognized by said character recognition means, with reference to a language dictionary, and for presenting the inappropriate phrase to the user on said display unit.

9. A document recognition system as defined in claim 8, further comprising phrase correction means for correcting said inappropriate phrase in accordance with an instruction which is given by the user in response to the presentation of said inappropriate phrase by said language processing means.

10. A document recognition system as defined in claim 1, further comprising skew correction means for correcting skew angle of said document image displayed on said display unit which has been detected by said system, in accordance with an instruction given by the user.

11. A document recognition system as defined in claim 1, wherein said character segmentation means detects first character segmentation positions being the most likely to be actual boundaries of characters included in said character line and second character segmentation positions being next most likely to be actual boundaries of characters included in said character line, and said document recognition system further comprises character segmentation correction means for presenting the first and second character segmentation positions to the user superimposed on said document image on said display unit, for correcting the character segmentation.

12. A document recognition system wherein characters on a document are recognized from an image of the document, and at least the recognized characters are displayed on a display unit, comprising:

input means for entering the document image which is an object to be recognized;

character line extraction means for obtaining character line data which indicate limits of a character line formed of a character string, including either of a frame and a block surrounding said character string, on said document image entered by said input means, wherein said frame and block indicate a set character line displayed as a polygon in superposition of the document image by utilizing graphics;

layout analysis means for recognizing blocks including a plurality of character lines, and for determining a sequence in which the plurality of blocks are read;

character segmentation means for detecting first character segmentation positions being the most likely to be actual boundaries of characters included in said character line and second character segmentation positions being next most likely to be actual boundaries of characters included in said character line, and for successively deriving character patterns of the individual characters contained in said character lines, divided at the first character segmentation positions, from said document image, on the basis of the sequence in which the plurality of blocks are read;

character recognition means for recognizing the character patterns derived by said character segmentation means, and for converting the recognized character patterns into respectively corresponding character codes;

language processing means for recognizing any inappropriate phrase in text data which includes the character codes of the character string recognized by said character recognition means, with reference to a language dictionary;

display means for causing said display unit to display a processed result of said language processing means;

character line correction means for presenting the limits of the character line obtained by said character line extraction means to a user of said document recognition system, superimposed on said document image on said display unit, and for correcting said limits of said character line in accordance with an instruction given by the user;

skew correction means for correcting a skew angle of said document image displayed on said display unit which has been detected by said system, in accordance with an instruction given by said user;

layout correction means for presenting limits of the blocks recognized by said layout analysis means, to said user superimposed on said document image on said display unit, and for correcting the presented limits of said blocks in accordance with an instruction given by said user;

character segmentation correction means for presenting said first and second character segmentation positions to said user superimposed on said document image on said display unit, for correcting the character segmentation;

character correction means for presenting said characters corresponding to said character code obtained by said character recognition means, to said user on said display unit, and for correcting said character codes in accordance with an instruction given by said user;

phrase correction means for presenting the inappropriate phrase recognized by said language processing means to said user, and for correcting said inappropriate phrase in accordance with an instruction given by said user;

means for selectively starting any selectable one of said character line correction means, said layout correction means, said character segmentation correction means, said character correction means and said phrase correction means, immediately after any of the processing steps of said character line extraction means, said layout analysis means, said character segmentation means, said character recognition means and said language processing means; and

retry control means which includes;

a control table for defining, for each of said character line correction means, said layout correction means, said character segmentation correction means, said character correction means and said phrase correction means, combinations of a plurality of means to perform processing after an operation of said each of said correction means, each of said combinations being selected from a group including said character line extraction means, said layout analysis means, said character segmentation means, said character recognition means and said language processing means, and

start means for automatically activating, after an operation of one of said correction means, said means included in one of the combinations selected with respect to said one of said correction means, with reference to said control table.

13. A document recognition system as defined in claim 12, further comprising batch control means which includes:

a control table for defining a combination selected from any of said character line extraction means, said layout analysis means, said character segmentation means, said character recognition means, said language processing means, said character line correction means, said layout correction means, said character segmentation correction means, said character correction means and said phrase correction means, along with a sequence of activation of said means defined in said control table, and

means for automatically activating said means defined in said control table, in the sequence of activation.

16. A document recognition system as defined in claim 12, wherein said character line correction means alters the limits of said character line on said display unit so that two character lines displayed adjacent to each other on said display unit may be connected into a single character line.

17. A document recognition system as defined in claim 12, wherein said character line correction means alters the limits of said character line on said display unit so that a single character line displayed on said display unit may be separated into two character lines.

18. A document recognition system as defined in claim 12, wherein said skew correction means detects a skew of said document image on the basis of said character line data extracted by said character line extraction means, and subjects said document image to rotation processing on the basis of the detected skew, so as to alter said skew.

21. A document recognition system as defined in claim 12, wherein said layout correction means further alters the reading sequence of the plurality of blocks.

14. A document recognition method wherein characters on a document are recognized from an image of the document, and at least the recognized characters are displayed on a display unit, said method comprising the steps of:

inputting the document image which is an object to be recognized;

gaining character line data which indicate limits of a character line formed of a character string, including either of a frame and a block surrounding said character string, on the input document image, thereby extracting the character line, wherein said frame and block indicate a set character line displayed as a polygon in superposition of the document image by utilizing graphics;

displaying the limits of the extracted character line superimposed on said document image on said display unit;

altering said limits of said character line displayed on said display unit, thereby correcting the character line data;

deriving character patterns of the individual characters contained in said character line, from said document image based on the corrected character line data;

recognizing the derived character patterns, and converting the recognized character patterns into respectively corresponding character codes; and

display the character codes on said display unit,

wherein the step of correcting said character line data alters the limits of said character line on said display unit so that two extracted character lines may be connected into a single character line, thereby correcting said character line data.

15. A document recognition method as defined in claim 14, wherein the step of correcting said character line data alters said limits of said character line on said display unit so that a single extracted character line may be separated into two character lines, thereby correcting said character line data.

19. A document recognition method as defined in claim 14, wherein the step of correcting said character line data alters said limits of said character line on said display unit so that two character lines displayed adjacent to each other on said display unit may be connected into a single character line.

20. A document recognition method as defined in claim 14, further including the steps of:

detecting a skew of said document image on the basis of the extracted character line data; and

subjecting the document image to rotation processing on the basis of the detected skew, so as to alter the skew.