US4959870A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Character recognition apparatus having means for compressing feature data
Assignee: Ricoh Co Ltd
•
Filing: 1988-05-23
•
Granted: 1990-09-25
•
Expires: 2008-05-23
Abstract
Abstract A character recognition apparatus comprises an extracting part for extracting a feature of a character within an image data related to a document in a form of a feature vector having components which are histogram values of n degree peripheral pattern describing the feature of the character by distances from an arbitrary side of a character frame to each of first through nth detected contour of the character within the frame, a compression part for compressing the feature vector into a compressed feature vector having components which are quantized data, a dictionary which stores compressed feature vectors of standard characters, and a matching part for matching the compressed feature vector obtained from the compression part with each of the compressed feature vectors stored in the dictionary so as to output at least one candidate character corresponding to one of the standard characters described by a compressed feature vector having a minimum difference with the compressed feature vector obtained from the compression part out of the compressed feature vectors stored in the dictionary.
Inventors
- 👤 Michiyoshi Tachikawa
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-26 13:33
CPC Classifications
G; G06; G06T; G06T9/00; G06T9/008; G06V; G06V30/00; G06V30/10; G06V30/18; G06V30/182
Claims Analysis
Total Claims: 12
Claim 1
Independent Claim
303 words
A character recognition apparatus comprising: scanning means for scanning a document and for producing image data related to the document; character extracting means for extracting from the output image data of said scanning means a character contained in a frame; feature extracting means for extracting a feature of the character extracted in said character extracting means in a form of a feature vector, said feature vector having components which are histogram values of n degree peripheral pattern describing the feature of the character by distances from an arbitrary side of the frame to each of first through nth detected contour of the character within the frame; compression means for compressing the feature vector from said feature extracting means into a compressed feature vector having components which are quantized data, said compression means comprising rearranging means for grouping the components of the feature vector into a predetermined number of groups for every predetermined number of degrees by uniformly distributing the components of the feature vector to each group depending on a sum of an average value and standard deviation of the components of the feature vector for each degree, and quantizer means for quantizing the components in each group with an independent quantization level for each group so as to obtain the quantized data having a predetermined number of bits for each group; a dictionary which stores compressed feature vectors of standard characters; and matching means for matching the compressed feature vector obtained from said compression means with each of the compressed feature vectors stored in said dictionary so as to output at least one candidate character corresponding to one of the standard characters described by a compressed feature vector having a minimum difference with the compressed feature vector obtained from said compression means out of the compressed feature vectors stored in said dictionary.
Claim 2
Dependent Claim
303 words
A character recognition apparatus comprising: scanning means for scanning a document and for producing image data related to the document; character extracting means for extracting from the output image data of said scanning means a character contained in a frame; feature extracting means for extracting a feature of the character extracted in said character extracting means in a form of a feature vector, said feature vector having components which are histogram values of n degree peripheral pattern describing the feature of the character by distances from an arbitrary side of the frame to each of first through nth detected contour of the character within the frame; compression means for compressing the feature vector from said feature extracting means into a compressed feature vector having components which are quantized data, said compression means comprising rearranging means for grouping the components of the feature vector into a predetermined number of groups for every predetermined number of degrees by uniformly distributing the components of the feature vector to each group depending on a sum of an average value and standard deviation of the components of the feature vector for each degree, and quantizer means for quantizing the components in each group with an independent quantization level for each group so as to obtain the quantized data having a predetermined number of bits for each group; a dictionary which stores compressed feature vectors of standard characters; and matching means for matching the compressed feature vector obtained from said compression means with each of the compressed feature vectors stored in said dictionary so as to output at least one candidate character corresponding to one of the standard characters described by a compressed feature vector having a minimum difference with the compressed feature vector obtained from said compression means out of the compressed feature vectors stored in said dictionary.
Claim 3
Dependent Claim
47 words
A character recognition apparatus as claimed in claim 1 in which said quantizer means quantizes the components in each group with a quantization level which increases as the sum of the average value and standard deviation of the components of the feature vector for each degree increase.
Claim 4
Dependent Claim
25 words
A character recognition apparatus as claimed in claim 1 in which said quantizer means quantizes the components in each group according to a linear quantization.
Claim 5
Dependent Claim
58 words
A character recognition apparatus as claimed in claim 3 in which said quantizer means determines the quantization level for each degree according to a formula Q=Cl×L, where Q denotes the quantization level, Cl denotes a constant for each degree within each group, l=1, 2, . . . , n, and L denotes a component value of each degree.
Claim 6
Dependent Claim
66 words
A character recognition apparatus as claimed in claim 1 in which said rearranging means groups the components of the feature vector into the predetermined number of groups for every predetermined number of degrees by uniformly distributing the components of the feature vector to each group from a largest sum of the average value and standard deviation of the components of the feature vector for each degree.
Claim 7
Dependent Claim
66 words
A character recognition apparatus as claimed in claim 1 in which said rearranging means groups the components of the feature vector into the predetermined number of groups for every predetermined number of degrees by uniformly distributing the components of the feature vector to each group from a smallest sum of the average value and standard deviation of the components of the feature vector for each degree.
Claim 8
Dependent Claim
121 words
A character recognition apparatus as claimed in claim 1 in which said rearranging means comprises a first table which stores table of average values and standard deviations of the components of the feature vector obtained from said feature extracting means for each degree and a rearranging part for rearranging the degree of the components of the feature vector obtained from said feature extracting means by referring to said first table, and said quantizer means comprises a second table which stores a table of quantization levels and a quantizer part for quantizing the components in each group with the independent quantization level for each group by making access to said second table by use of values of the components in each group.
Claim 9
Dependent Claim
25 words
A character recognition apparatus as claimed in claim 1 in which said quantizer means quantizes the components in each group according to a non-linear quantization.
Claim 10
Dependent Claim
44 words
A character recognition apparatus as claimed in claim 8 in which a threshold value within each quantization level is set in said quantizer means so that a probability of a certain logic value occurring for each bit of the quantized data is approximately constant.
Claim 11
Dependent Claim
91 words
A character recognition apparatus as claimed in claim 8 which further comprises a distance table which stores a table of distances between feature vectors, said matching means matching the compressed feature vector obtained from said compression means with each of the compressed feature vectors stored in said dictionary so as to output at least one candidate character corresponding to one of the standard characters described by a compressed feature vector having a minimum distance from the compressed feature vector obtained from said compression means by making access to said distance table.
Claim 12
Dependent Claim
36 words
A character recognition apparatus as claimed in claim 10 in which said distances stored in said distance table are calculated by using a center value of the quantization levels in each group as a representative point.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1993-09-24 | 1994-03-26 | 1994-09-25 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1997-09-24 | 1998-03-26 | 1998-09-25 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2001-09-24 | 2002-03-26 | 2002-09-25 | $7700.00 | Window Closed / Lapsed |