Fast character segmentation of skewed text lines for optical character recognition

• US 5,172,422 A
• Filed: 05/13/1991
• Issued: 12/15/1992
• Est. Priority Date: 05/13/1991
• Status: Expired due to Term

• Current Assignee: Eastman Kodak Company
• Original Assignee: Eastman Kodak Company
• Inventors: Tan, Hin-Leong
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): COUSO, YON JUNG
• Application Number:
  US07/699,029
• Time in Patent Office: 582 Days
• Field of Search: 382/9, 382/45, 382/48, 382/18, 382/46, 358/488, 271/265
• US Class Current: 382/174
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/146 Aligning or centring of the image pick-up or image-field

Abstract

A character segmentation process for skewed documents divides the document bit map into vertical blocks and within each vertical block estimates initial top and bottom bounds of each character in a character row from horizontal projections of the vertical block, thereby avoiding the effects of skew. Operations in the document bit map to compute the exact top, bottom, left and right boundaries of the character are confined within the narrow strip between the initial top and bottom bounds of the character so as to greatly reduce the amount of data fetched from memory during such operations and thereby speed up the segmentation process.

Claims

1. A method for segmenting individual characters in a bit map of a document comprising successive rows of characters, said bit map comprising plural horizontal rows and vertical columns of "ON" and "OFF" pixels, in preparation for performing optical character recognition, said method comprising:

dividing said bit map into plural vertical blocks;

determining from horizontal projects of each of said vertical blocks the initial top and bottom bounds by;

a) constructing for each of said vertical blocks a top horizontal projection array listing for each character row in the vertical block the top pixel row thereof;

b) constructing for each of said vertical blocks a bottom horizontal projection array listing for each character row in the vertical block the bottom pixel row thereof; and

c) deriving from said top and bottom projection arrays the top and bottom pixel row of each character row in said vertical block;

determining from horizontal projections of reach of said vertical blocks initial top and bottom bound of one character in a successive one of said character rows in a current one of said vertical blocks;

finding a left boundary of said on character by searching from left to right within a restricted portion of said bit map between the initial top and bottom bounds of said one character for a first pixel column having a t least a first threshold number of "ON" pixels and defining said left boundary to be said first pixel column having at least said threshold number of "ON" pixels;

finding a fright boundary of said one character by searching from left to right from said left boundary of said character within said restricted portion of said bit map between the initial top and bottom bounds of said on character for a first pixel column having no more than a second threshold number of "ON" pixels and defining said right boundary to be said first pixel column having no more than said second threshold number of "ON" pixels;

finding a top boundary of said one character by searching down from said initial top bound of said one character within a restricted portion of said bit map between said left and right boundaries of said one character for a first pixel row containing at least said first threshold number of"ON" pixels; and

defining as said initial top bound of said one character the highest of the top pixel rows of the corresponding character rows in adjacent vertical blocks to the left and right of si current vertical block containing said character; and

finding a bottom boundary of said one character by searching up from said initial bottom bound of said one character within said restricted portion of said bit map between said left and right boundaries of said one character for a first pixel row containing at least said first threshold number of "ON" pixies.

2. The method of claim 1 further comprising:

defining as the initial bottom bound of said one character the lowest of the bottom pixel rows of the corresponding character rows in adjacent vertical blocks to the left and right of said current vertical block containing said character.

3. The method of claim 1 further comprising:

searching said top projection array for an isolated run of "ON" pixels near and above said top bound of said one character and re-defining said initial top bound as the top pixel row of said isolated run of "ON" pixels.

4. The method of claim 1 further comprising:

repeating the foregoing steps for the next character to the right in said one character row;

when the last character in said one character row has been segmented, finding the corresponding one character row in an adjacent vertical block to the right of said current vertical block and repeating said determining step and each of said finding steps beginning at a left end of said corresponding one character row in said adjacent vertical block.

5. The method of claim 1 wherein the step of finding the corresponding one character row in an adjacent vertical block to the right comprises finding a horizontal projection of a character row in said adjacent vertical block which spans a center portion of the last character segmented in said one character row of said current vertical block.

6. The method of claim 1 wherein said first threshold number is one and said second threshold number is zero.

7. Apparatus for segmenting individual characters in a bit map of a document comprising successive rows of characters, said bit map comprising plural horizontal rows and vertical columns of "ON" and "OFF" pixels and being stored in a bit map memory, said apparatus comprising:

means for dividing said bit map into plural vertical blocks;

means for determining from horizontal projections of each of said vertical blocks initial top and bottom bounds of one character in a successive one of said character rows in a current one of said vertical blocks by;

a) means for constructing for each of said vertical blocks a top horizontal projection array listing for each character row in the vertical block the top pixel row thereof

b) means for constructing for each of said vertical blocks a broom horizontal projection array listing for each character row in the vertical bloc the bottom pixel row thereof; and

c) means for deriving from said top and bottom projection arrays the top and bottom pixel row of each character row in said vertical block.

means for finding a left boundary of said one character by searching from left to right within a restricted portion of said bit map between the initial top and bottom bounds of said one character for a first pixel column having at lest a first threshold number of "ON" pixels and defining said left boundary to be said first pixel column having at least said threshold number of "ON" pixels;

means for finding a right boundary of said one character by searching from left to right from said left boundary of said character within said restricted portion of said bit map image between the initial top and bottom bounds of said one character for a first pixel column having no more tan a second threshold number of "ON" pixels and defining said right boundary to be said first pixel column having no more than said second threshold number of "ON" pixels;

means for finding a top boundary of said one character by searching down from said initial top bound of said one character within a restricted portion of said bit map image between said left and right boundaries of said on character for a first pixel row containing at lest said first threshold number of "OPN" pixels; and

means for defining as said initial top bound of said one character the highest of the top pixel rows of the corresponding character rows in adjacent vertical block to the left and right of said current vertical block containing said character; and

means for finding a bottom boundary of said one character by searching up from said restricted portion of said bit map image between said left and right boundaries of said one character for a first pixel row containing at least said first threshold number of "ON" pixels.

8. The apparatus of claim 7 further comprising:

means for defining as the initial bottom bound of said one character the lowest of the bottom pixel rows of the corresponding character rows in adjacent vertical blocks to the left and right of is current vertical block containing said character.

9. The apparatus of claim 7 further comprising

means for searching said top projection array for an isolated run of"ON" pixels near and above said top pixel row of said one character row and re-defining said initial top bound as the top pixel row of said isolated run of "ON" pixels.

10. The apparatus of claim 7 further comprising:

concatenation means responsive upon the last character in said one character row being segmented, for finding the corresponding one character row in an adjacent vertical block to the right of said current vertical.

11. The apparatus of claim 10 wherein said concatenation means for finding the corresponding one character row in an adjacent vertical block to the right comprises means for finding a horizontal projection of a character row in said adjacent vertical block which spans a center portion of the last character segmented in said one character row of said current vertical block.

12. The apparatus of claim 7 wherein said first threshold number is one and said second threshold number is zero.