# US 5,708,730 A · Table recognition apparatus

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 5,708,730 A
- Document ID: US5708730A
- Status: Granted
- Kind: A
- Application number: 08/479,974
- Filed: 1995-06-07
- Granted: 1998-01-13
- Earliest claimed: 1992-10-27
- Assignees: Fuji Xerox Co., Ltd.
- Inventors: Katsuhiko Itonori
- Web page: https://patentgrep.com/patent/US5708730A

## CPC

- G06V 30/412: Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A table recognition apparatus for recognizing a table image wherein coexist characters and ruled lines. The apparatus comprises a character/ruled line separating part for separating characters from ruled lines in the table image; a character block extracting part for extracting character blocks from the character image separated by the character/ruled line separating part; and a character block extending part for extending the character blocks extracted by the character block extracting part in such a manner that the edges of the character blocks are aligned according to the positional relations between the blocks.

## Claims

### Claim 1

A table recognition apparatus for recognizing a table image having both a character image and a ruled line image, said table recognition apparatus comprising:
image separating means for separating characters from ruled lines in said table image;
means for acquiring a rectangular area containing a character in the character image;
means for producing a threshold from the dimensions of said acquired rectangular area;
character block extracting means for using said threshold to extract character blocks from the character image separated by said image separating means;
first character block extending means for extending each of said character blocks up to the nearest one of the ruled lines;
means for computing a horizontal and a vertical distribution of character blocks to keep each extended character block from overlapping with other ones of the character blocks;
second character block extending means for further extending the extended character blocks to align the edges of the extended character blocks;
row extracting means for extracting rows of said table image based on the distance between the extended character blocks; and
column extracting means for extracting columns of said table image based on the distance between the extended character blocks.

### Claim 2

A table recognition apparatus for recognizing a table image having both a character image and a ruled line image, said table recognition apparatus comprising:
image separating means for separating characters from ruled lines in said table image;
character block extracting means for extracting character blocks from the character image separated by said image separating means;
rectangular frame extracting means for extracting a rectangle composed of ruled lines, from the ruled line image separated by said image separating means;
means for converting the extracted ruled lines into a plurality of horizontal and vertical vectors;
means for comparing said horizontal and vertical vectors to a threshold angle;
component frame extracting means for extracting as a component frame a rectangular frame including at most one character block;
means for computing a horizontal and vertical distribution of character blocks
first character block extending means for using said computed horizontal and vertical distribution of character blocks to extend a target character block up to the nearest one of the ruled lines;
second character block extending means for further extending the extended character blocks to align the edges of the extended character blocks;
row extracting means for extracting rows of said table image based on the distance between the extended character blocks; and
column extracting means for extracting columns of said table image based on the distance between the extended character blocks.

### Claim 3

A table recognition apparatus for recognizing a table image having both a character image and a ruled line image, said table recognition apparatus comprising:
image separating means for separating characters from ruled lines in said table image;
first character block extending means for extending each of said character blocks up to the nearest one of the ruled lines;
means for computing a horizontal and a vertical distribution of character blocks to keep each extended character block from overlapping with other ones of the character blocks;
second character block extending means for further extending the extended character blocks to align the edges of the extended character blocks;
character block normalizing means for normalizing each character block based on the distance between the extended character blocks;
character block interpolating means for detecting a virtual character block that overlaps the character blocks normalized by said character block normalizing means;
row extracting means for extracting rows based on the distance between the character blocks interpolated by said character block interpolating means; and
column extracting means for extracting columns based on the distance between the character blocks interpolated by said character block interpolating means.