Method of reducing background noise by tracking character skew
View Patent ↗A method of reducing background noise during optical character recognition includes conducting a first recognition pass over image data at a sequence of character positions to frame and recognize a first set of characters while a second set of characters remains unrecognized. Character skew is tracked based on the first recognition pass. The image data is modified to remove background noise based on the tracked character skew. A second recognition pass is conducted over the modified image data at the sequence of character positions to frame and recognize the second set of characters.
1. A method of reducing background noise during optical character recognition in which a scan band of image data is optically read from a document and the image data is passed to a recognition system, the image data including data at a plurality of character positions that form a sequence, the method comprising:
conducting a first recognition pass over the image data at the sequence of character positions to frame and recognize a first set of characters while a second set of characters remains unrecognized;
tracking character skew based on the first recognition pass, the tracked character skew being indicative of vertical drifting of framed characters over the sequence of character positions;
modifying the image data to remove background noise based on the tracked character skew; and
conducting a second recognition pass over the modified image data at the sequence of character positions to frame and recognize the second set of characters.
2. The method of claim 1 wherein tracking character skew further comprises:
determining a plurality of frame vertical centers based on the first recognition pass.
3. The method of claim 2 wherein tracking character skew further comprises:
determining frame vertical centers for character positions corresponding to the first set of characters based on the first recognition pass; and
determining an average vertical center for each character position wherein the average vertical center for a specific character position is based on any determined frame vertical center for that specific character position and for a limited range of character positions about that specific character position.
4. The method of claim 3 wherein modifying further comprises:
for each character position, retaining image data that falls within a limited distance in either direction from the corresponding average vertical center while removing image data that falls outside the limited distance in either direction from the corresponding average vertical center.
5. The method of claim 3 wherein modifying further comprises:
for each character position, retaining image data that falls within ¾ of a character height in either direction from the corresponding average vertical center while removing image data that falls outside ¾ of a character height in either direction from the average vertical center.
6. A system including a reader and a recognition system programmed to perform a method of reducing background noise during optical character recognition in which a scan band of image data is optically read from a document and the image data is passed to the recognition system, the image data including data at a plurality of character positions that form a sequence, the recognition system being programmed to:
conduct a first recognition pass over the image data at the sequence of character positions to frame and recognize a first set of characters while a second set of characters remains unrecognized;
track character skew based on the first recognition pass, the tracked character skew being indicative of vertical drifting of framed characters over the sequence of character positions;
modify the image data to remove background noise based on the tracked character skew; and
conduct a second recognition pass over the modified image data at the sequence of character positions to frame and recognize the second set of characters.
7. The system of claim 6 wherein tracking character skew further comprises:
determining a plurality of frame vertical centers based on the first recognition pass.
8. The system of claim 7 wherein tracking character skew further comprises:
determining frame vertical centers for character positions corresponding to the first set of characters based on the first recognition pass; and
determining an average vertical center for each character position wherein the average vertical center for a specific character position is based on any determined frame vertical center for that specific character position and for a limited range of character positions about that specific character position.
9. The system of claim 8 wherein modifying further comprises:
for each character position, retaining image data that falls within a limited distance in either direction from the corresponding average vertical center while removing image data that falls outside the limited distance in either direction from the corresponding average vertical center.
10. The system of claim 8 wherein modifying further comprises:
for each character position, retaining image data that falls within ¾ of a character height in either direction from the corresponding average vertical center while removing image data that falls outside ¾ of a character height in either direction from the average vertical center.