Character recognition device which divides a single character region into subregions to obtain a character code

• US 5,151,951 A
• Filed: 03/11/1991
• Issued: 09/29/1992
• Est. Priority Date: 03/15/1990
• Status: Expired due to Fees

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Aramaki, Takashi, Ueda, Toru, Ishizuka, Yasushi, Togawa, Fumio
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Fox, David
• Application Number:
  US07/667,340
• Time in Patent Office: 568 Days
• Field of Search: 382/14, 382/15, 382/16, 382/19, 382/27, 382/36, 382/37, 382/38, 364/513, 395/21, 395/23
• US Class Current: 382/156
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18086 by performing operations within image blocks or by using histograms

Abstract

A character recognition device has a subdivider, a features calculator and a character code recognition device. Image data for a single character area is extracted from scanned character image data and input to the subdivider. This subdivider divides the image data for the single character area into subregions. The features calculator calculates quantified features in each subregion based on a degree of resemblance between a template and image data in the subregions. When the features of each subregion are calculated for all subregions constituting the single character area, a character code corresponding to the scanned character image data is recognized by the character code recognition device based on the quantified features of each of all subregions.

Claims

1. A character recognition device which isolates the image data for a single character area from scanned character image data, and recognizes a character code corresponding to the character image data based on the image data of this single character area, comprising:

a region divider which divides the image data of the single character area into plural subregions,

a features calculator which calculates quantified features of each subregion based on degrees of resemblance between image data of each subregion divided by said region divider and templates which express in each subregion differences in shape of characters in a recognition character group,

a character code recognition means which recognizes the character code corresponding to the scanned character image data based on the quantified features calculated by the features calculator in all subregions composing the single character area, and

a neural network which during training refreshes contents of a specified weight vector corresponding to an input training vector to approach the contents of the input training vector, and which

trains the neural network using training vectors generated for each subregion based on character image data of plural training characters representative of the recognition character group, and

defines as the templates used during calculation of the quantified features of the subregions by the feature calculator the weight vectors set, as a result of being refreshed to approach the contents of each training vector, to express differences in shapes of characters in each of training characters representative of the recognition character group.

4. A character recognition device according to claim 1 wherein the character code recognition means is a character identification neural network trained to identify a character code associated with an input vector when an input vector based on the quantified features calculated by the feature calculator in all subregions forming the single character area is input.

2. A character recognition device which isolates the image data for a single character area from scanned character image data, and recognizes a character code corresponding to the character image data based on the image data of this single character area, comprising:

a region divider which divides the image data of the single character area into plural subregions, wherein the region divider, when dividing the character image data of the single character area into subregions, generates the subregions so that adjacent subregions overlap,

a features calculator which calculates quantified features of each subregion based on degrees of resemblance between image data of each subregion divided by said region divider and templates which express in each subregion differences in shape of characters in a recognition character group,

a character code recognition means which recognizes the character code corresponding to the scanned character image data based on the quantified features calculated by the features calculator in all subregions composing the single character area, and

a neural network which during training refreshes contents of a specified weight vector corresponding to an input training vector to approach the contents of the input training vector, and which

trains the neural network using training vectors generated for each subregion based on character image data of plural training characters representative of the recognition character group, and

defines as the templates used during calculation of the quantified features of the subregions by the feature calculator the weight vectors set, as a result of being refreshed to approach the contents of each training vector, to express differences in shapes of characters in each of training characters representative of the recognition character group.

3. A character recognition device which isolates the image data for a single character area from scanned character image data, and recognizes a character code corresponding to the character image data based on the image data of this single character area, comprising:

a region divider which divides the image data of the single character area into plural subregions, wherein the region divider, when dividing the character image data of the single character area into subregions, generates the subregions to have various shape so that places which express differences in shapes of characters in a recognition character group are covered by the subregions,

a features calculator which calculates quantified features of each subregion based on degrees of resemblance between image data of each subregion divided by said region divider and templates which express in each subregion differences in shape of characters in a recognition character group,

a character code recognition means which recognizes the character code corresponding to the scanned character image data based on the quantified features calculated by the features calculator in all subregions composing the single character area, and

a neural network which during training refreshes contents of a specified weight vector corresponding to an input training vector to approach the contents of the input training vector, and which

trains the neural network using training vectors generated for each subregion based on character image data of plural training characters representative of the recognition character group, and

defines as the templates using during calculation of the quantified features of the subregions by the feature calculator the weight vectors set, as a result of being refreshed to approach the contents of each training vector, to express differences in shapes of characters in each of training characters representative of the recognition character group.