Character recognition system
First Claim
Patent Images
1. A character recognition system for reading character image data from a document comprising:
- optical scanning means for scanning the document to obtain a video signal containing image data from said document;
extracting means for extracting character blocks from the video signal, wherein each character block contains image data of individual characters from said document, said extracting means further producing predetermined history data of each character block indicative of the process of extraction of said character block;
combining means for comparing ratios of predefined dimensional attributes of each character block with predetermined standard ratios and selectively recombining two or more of the extracted character blocks based on such comparison in order to form a combined character block;
character recognition means for receiving the extracted character blocks and the recombined character blocks for performing character recognition according to predetermining algorithms and outputting a signal indicative of a recognized character; and
encoding means connected to said extracting means to obtain the extraction history data, for controlling the operation of said combining and character recognition means by means of the extraction history data in such a manner that when said character recognition means fails to recognize a character in said character block, said encoding means selectively causes said combining means to successively combine character blocks until said character recognition means recognizes the character in said character block formed by said combining means.
1 Assignment
0 Petitions
Accused Products
Abstract
A character recognition system extracts character blocks to contain a minimum character unit therein. The character recognition system has the capability of re-combining and/or re-extracting character blocks during the character recognition process. Re-combination and re-extraction of the character blocks is controlled based on the history of the extraction process for each of the character blocks so that re-combination and re-extraction can be performed effectively.
65 Citations
63 Claims
-
1. A character recognition system for reading character image data from a document comprising:
-
optical scanning means for scanning the document to obtain a video signal containing image data from said document; extracting means for extracting character blocks from the video signal, wherein each character block contains image data of individual characters from said document, said extracting means further producing predetermined history data of each character block indicative of the process of extraction of said character block; combining means for comparing ratios of predefined dimensional attributes of each character block with predetermined standard ratios and selectively recombining two or more of the extracted character blocks based on such comparison in order to form a combined character block; character recognition means for receiving the extracted character blocks and the recombined character blocks for performing character recognition according to predetermining algorithms and outputting a signal indicative of a recognized character; and encoding means connected to said extracting means to obtain the extraction history data, for controlling the operation of said combining and character recognition means by means of the extraction history data in such a manner that when said character recognition means fails to recognize a character in said character block, said encoding means selectively causes said combining means to successively combine character blocks until said character recognition means recognizes the character in said character block formed by said combining means. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 62)
-
-
19. A character recognition system comprising:
-
scanning means for obtaining a video signal containing image data of said document; extracting means, which is supplied with the video signal, for segmenting said image data into character blocks, each of which contains image data of an individual character unit, and outputting a corresponding character block signal, said extracting means including a plurality of segmentation stages including a first segmentation stage for segmenting character-string blocks and a second segmentation stage for segmenting each character-string block into separate character blocks each containing character image data of a character unit and wherein said first and second stages further output segmentation history data indicative of the manner in which each character-string block and character block have been segmented; combining means, which is supplied with the character block signal, for selectively recombining two or more character blocks to form a combined character block and outputting a combined character block signal; character recognition means, which is selectively supplied with either the character block signal or the combined character block signal, for performing character recognition according to predetermined algorithms and outputting a "recognized character" signal; re-segmenting means, which is supplied with the character block signal and the combined character block signal, for re-segmenting a selected character block for forming a re-segmented character block; and encoding means connected to said extracting means and supplied with the segmentation history data with respect to each of said character blocks segmented by said extracting means, for controlling the operation of said combining, re-segmentation and character recognition means based on said segmentation history data in such a manner that said encoding means selectively operates one of said combining and re-segmenting means when said character recognition means fails to recognize a character in said character block, so as to recombine or re-segment, respectively, one or more selected character blocks to assure recognition in said character recognition means of the character block formed by a selected one of said combining and re-segmenting means. - View Dependent Claims (20, 21, 22, 23, 24, 25, 26, 27, 28, 29)
-
-
30. A character recognition system comprising:
- scanning means for obtaining a video signal containing image data corresponding to said document and wherein said scanning means scans the document along horizontal and vertical lines;
extracting means for deriving from the video signal a plurality of character blocks, each of which contains image data of individual characters in said document, and outputting a corresponding character block signal, said extracting means including a plurality of segmentation stages including a first segmentation stage for segmenting character-string blocks, a second segmentation stage for segmenting each character-string block into separate character blocks each containing character image data of a character unit and a third segmentation stage for checking the continuity of image data containing bits within each character block for detecting discontinuities in order to separate the character blocks into two or more character blocks, each containing a minimum character unit; combining means, which is selectively supplied with the character block signal, for recombining two or more character blocks to form a combined character block and outputting a combined character block signal; re-segmenting means, which is supplied with the character block signal and the combined character block signal, for selectively re-segmenting one or more selected character blocks to form a re-segmented character block and outputting a corresponding re-segmented character block signal; character recognition means, which is supplied with the character block signal, the combined character block signal and the re-segmented character block signal, for performing character recognition according to predetermined algorithms and outputting a "recognized character" indicative signal; and encoding means for controlling the operation of said combining, re-segmentation and character recognition means in such a manner that said encoding means selectively operates one of said combining and re-segmenting means when said character recognition means fails to recognize a character in said character block, so as to assure recognition of character in said character recognition means with the character block formed by the selected one of said combining and re-segmenting means. - View Dependent Claims (31, 32, 33, 34, 35, 36, 37, 38)
- scanning means for obtaining a video signal containing image data corresponding to said document and wherein said scanning means scans the document along horizontal and vertical lines;
-
39. A character recognition system comprising
first means for receiving a video signal containing character image data; -
extracting means for deriving from the video signal a plurality of character blocks, each of which contains image data of individual characters, said extracting means including a plurality of segmentation stages inclucing a first segmentation stage for segmenting character-string blocks into character blocks, a second segmentation stage for segmenting each character block containing a character unit of image data and a third segmentation stage for checking the continuity of image data containing bits within each character block for detecting discontinuity in order to separate character blocks into more than one character block so as to obtain character blocks which each contain a minimum character unit and wherein the extracting means generates segmentation history data with respect to at least said second and third segmentation stages performed by said extracting means, which segmentation history data is indicative of the order of segmentation of each character block in each of said second and third stages; combining means for recombining two or more character blocks for forming a combined character block; re-segmenting means for re-segmenting character block for forming a re-segmented character block; character recognition means for performing character recognition according to predetermined algorithms and outputting recognized character indicative signal; and encoding means connected to said extracting means for obtaining the segmentation history data, and controlling the operation of said combining and character recognition means based on said segmentation history data in such a manner that said encoding means seletively operates one of said combining means and re-segmenting means when said character recognition means fails to recognize a character in said character block, so as to assure recognition by said character recognition means of the character in the character block formed by the selected one of said combining and re-segmenting means. - View Dependent Claims (40, 41, 42, 43, 44, 45, 46)
-
-
47. A method for recognizing characters contained in a document comprising the steps of:
-
generating a video signal containing image data of said document; from the video data, extracting character blocks, each of which contains image data of individual characters in said document; combining two or more character blocks for forming a combined character block; re-segmenting a character block for forming a re-segmented character block; from the character blocks, the combined character blocks and the re-segmented character blocks, performing character recognition according to predetermined algorithms and outputting a "recognized character" indicative signal; and controlling the combining step, re-segmenting step and character recognition step in such a manner that the combining step, re-segmenting step and character recognition step are selectively performed to ensure recognition of characters by repeating said combining step, re-segmenting step and character recognition step with respect to a selected one or a selected succession of character blocks, combined character blocks or re-segmented character blocks until character recognition is achieved. - View Dependent Claims (48, 49, 50, 51, 52, 53, 54, 55, 56, 59, 60, 61, 63)
-
- 57. A character recognition method as set forth in cliam 56, wherein said character block segmentation step includes the steps of checking the level of an x-projection profile of the document image data with respect to a predetermined threshold level and using that information to select the vertical scanning lines of the scanning step to be included in said character block.
Specification