Image analysis – Pattern recognition – On-line recognition of handwritten characters
Reexamination Certificate
1995-07-13
2001-08-07
Patel, Jayanti K. (Department: 2623)
Image analysis
Pattern recognition
On-line recognition of handwritten characters
C704S010000
Reexamination Certificate
active
06272242
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to the field of document image processing systems, and more particularly relates to optical character recognition (OCR) systems. The invention further relates to a method and system which groups similar character patterns when performing the character recognition process.
2. Discussion of the Background
Conventional OCR systems extract features from each character pattern. A pattern matching process is performed using the extracted character patterns and a recognition dictionary containing reference information. Differences between sizes of the characters, the type of fonts used, and noise included in the original image often affect the performance of OCR systems.
A typical image on which character recognition is performed includes alphabetic and numeric patterns which are used many times. The same character should be represented by the same bit-mapped image but in reality, there are some differences attributable to noise including quantization error or sampling (scanning) error. Further, the recognition process performed on each character pattern is complicated and time consuming, as demonstrated by known character recognition systems such as the publication “On the Recognition of Printed Characters of Any Font and Size,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-9, No. 2, March 1987, by Simon Kahan et al, pp. 274-288, which is incorporated herein by reference.
SUMMARY OF THE INVENTION
Accordingly, it is an object of the invention to use a grouping of similar character patterns to achieve an efficient and accurate character recognition process. It is a further object of the invention to use a grouping of similar character patterns to achieve a character recognition process which is more accurate and/or faster than character recognition processes which do not use character grouping.
It is another object of the invention to use a probability of appearance of a letter, in addition to using pattern grouping, in order to obtain a more accurate result.
It is yet another object of the invention to use a probability of appearance of n-grams (e.g., consecutive characters such as a digram or trigram) in order to achieve a more accurate recognition result.
These and other objects are accomplished by a novel character recognition process and apparatus which performs a grouping of similar character patterns. According to one embodiment, similar character patterns are grouped and a character recognition process is performed for every character which is a member of the group. Then the results of the character recognition process are analyzed and a single character code is assigned to each character pattern of the group based on the recognition result of each character. This process allows accurate recognition results to be obtained based on the accuracy of the grouping, even if the character recognition process for a specific member of the group does not result in an accurate recognition determination.
As an alternative to the above process, after the similar character patterns are placed into groups, a representative pattern is generated for each group. Then a character recognition process is performed on each representative pattern in order to obtain a recognition result. This recognition process results in especially fast processing, as compared to conventional methods of performing character recognition for an image.
It is possible that some representative patterns of groups which are generated may not result in a clear character recognition answer but there may be a plurality of character recognition results which may have a similar probability of being accurate. In cases where there is not a clear choice for the character recognition result, the probability of occurrence of each of the possible character recognition results is compared to characteristics of typical documents. For example, there are known probability tables for the appearance of characters within writings. These probability tables can be used along with the grouping process to obtain more accurate results. Also, there are known probabilities of the appearances of n-grams which are consecutive letters. Two consecutive letters are called a digram and three consecutive letters are called a trigram, etc. Using these probabilities of digrams, trigrams, or other n-grams, it is possible to obtain more accurate recognition results.
REFERENCES:
patent: 5075896 (1991-12-01), Wilcox et al.
patent: 5187751 (1993-02-01), Tanaka
patent: 5237628 (1993-08-01), Levitan
patent: 5239594 (1993-08-01), Yoda
patent: 5497432 (1996-03-01), Nishida
patent: 5526447 (1996-06-01), Shepard
patent: 5596657 (1997-01-01), Choi
patent: 5852685 (1998-12-01), Shepard
patent: 5875263 (1999-02-01), Froessl
patent: 5881172 (1999-03-01), Pintsov
Saitoh Takashi
Takatsu Kazunori
Oblon & Spivak, McClelland, Maier & Neustadt P.C.
Patel Jayanti K.
Ricoh & Company, Ltd.
LandOfFree
Character recognition method and apparatus which groups... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Character recognition method and apparatus which groups..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Character recognition method and apparatus which groups... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2505590