Character recognition apparatus and method for recognizing characters

• US 6,327,384 B1
• Filed: 11/13/1997
• Issued: 12/04/2001
• Est. Priority Date: 11/13/1996
• Status: Expired due to Term

• Current Assignee: Rakuten, Inc. (Rakuten Group, Inc.)
• Original Assignee: NEC Corporation
• Inventors: Mochizuki, Takashi, Hirao, Kouichirou, Yamada, Keiji, Hongu, Takahiro, Arai, Mitsutoshi
• Primary Examiner(s): Tran, Phuoc
• Application Number:
  US08/970,150
• Time in Patent Office: 1,482 Days
• Field of Search: 382/168, 382/185, 382/200, 382/173-178, 382/170-172, 382/209, 358/462, 358/467, 358/470
• US Class Current: 382/173
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1463 Orientation detection or correction, e.g. rotation of multiples of 90 degrees

Abstract

In a projection means black pixel histograms of a binary stationary image are generated in both the vertical and the horizontal direction. In a text type judgment means, in accordance with these histograms, a determination is made of whether the image is vertical text or horizontal text. Based on the result of this determination, a pattern block extraction means extracts either a column or a row from the image. The block is further projected and divided into smaller blocks. Then projection is again applied to these divided blocks and patterns are extracted by a pattern extraction means. A judgment is made as to whether or not joining of the extracted patterns is to be performed and, if joining is required, they are joined by a pattern joining means and finally the offsets of all the extracted patterns are calculated, whereupon data (of extracted patterns) are sent to a pattern matching process.

Claims

1. A character recognition apparatus that is used in performing coding of a pattern in accordance with the results of dividing a binary stationary image into patterns and performing pattern matching with respect to each of said divided patterns, said pattern extraction apparatus comprising:

projection means for projecting black pixels in said binary stationary image and determining a histogram thereof;

document style determining means for determining, in accordance with said histogram, whether said binary stationary image is horizontally written text or vertically written text and for outputting a result of said determination of said document style determining means;

means for extracting a block from said image in accordance with said determination result;

pattern matching means for performing pattern matching operation by comparing said input pattern with a pattern as stored in a library;

coding means for coding said input pattern into coded data depending upon an off-set data and a result of said pattern matching operation;

means for extracting a pattern based on said block;

means for making a joining judgment which determines whether or not said extracted pattern is extracted as a unitary character;

means for joining extracted patterns in accordance with output from said means for making a joining judgment; and

means for calculating a relative distance between a currently extracted pattern and an immediately previously extracted pattern, as an offset.

2. A character recognition apparatus according to claim 1, wherein said binary stationary image has a vertical direction and a horizontal direction, and wherein said projection means takes a projection of black pixels in said vertical direction and said horizontal direction to generate said histogram.

3. A character recognition apparatus according to claim 2, wherein when said judgment result of said document style determining means corresponds to vertically aligned text, said block extraction mean extracts a column as a block from said binary stationary image, and when said judgment result of said document style determining means corresponds to horizontally aligned text, said block extraction means extracts a row as a block from said binary stationary image.

4. A character recognition apparatus according to claim 1, wherein when said judgment result of said document style determining means corresponds to vertically aligned text, said block extraction means extracts a column as a block from said binary stationary image, and when said judgment result of said document style determining means corresponds to horizontally aligned text, said block extraction means extracts a row as a block from said binary stationary image.

5. A character recognition apparatus according to claim 4, wherein with respect to said extracted column or row, said projection means performs further projection in a first direction that is perpendicular with respect to said column or said row and determines a sub-histogram, in accordance to which said block extraction means divides said column or said row and provides said divided block to said pattern extraction means.

6. A character recognition apparatus according to claim 5, wherein, with respect to said block, said projection means further projects in a second direction, which is perpendicular to said first direction and output a result of said projection, and wherein said pattern extraction means, based on said projection result, extracts said extracted pattern from said block.

7. A method of character recognition for recognizing complex characters comprised of simpler patterns, said method comprising the steps of:

projecting black pixels in said binary stationary image and determining a histogram thereof;

determining, in accordance with said histogram, whether said binary stationary image is horizontally written text or vertically written text and outputting a result;

extracting a block from said image in accordance with said result of said step of determining;

extracting a plurality of patterns in accordance with said block;

performing a pattern matching operation on each pattern in said block by comparing each pattern of said plurality of patterns with patterns stored in a library;

determining whether or not each extracted pattern corresponds to a unitary character;

joining extracted patterns based on said step of determining whether or not an extracted pattern corresponds to a unitary character;

coding said input pattern into coded data depending upon an off-set data and a result of said pattern matching operation; and

calculating a relative distance between a currently extracted pattern and an immediately previously extracted pattern, as an offset.

8. A recording medium which stores a controlling program for character recognition from a binary stationary image which can be performed by a computer, wherein said program is for character recognition from a binary stationary image and is used in performing coding of a pattern in accordance with the results of dividing a binary stationary image into patterns and performing a pattern matching with respect to each of said divided patterns, said program comprising the steps of:

projecting black pixels in said binary stationary image and determining a histogram thereof;

determining, in accordance with said histogram, whether said binary stationary image is horizontally written text or vertically written text and outputting a result;

extracting a block from said image in accordance with said result of said step of determining;

extracting a pattern in accordance with said block;

performing pattern matching operation by comparing said input pattern with a pattern as stored in a library;

coding said input pattern into coded data depending upon an off-set data and a result of said pattern matching operation;

determining whether or not an extracted pattern corresponds to a unitary character;

joining extracted patterns based on said step of determining whether or not an extracted pattern corresponds to a unitary character; and

calculating a relative distance between a currently extracted pattern and an immediately previously extracted pattern, as an offset.