# US 2001/0024520 A1 · Method and apparatus for table recognition, apparatus for character recognition, and computer product

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

- Number: US 2001/0024520 A1
- Document ID: US20010024520A1
- Status: Published application
- Kind: A1
- Application number: 09/804,241
- Filed: 2001-03-13
- Published: 2001-09-27
- Earliest claimed: 2000-03-15
- Assignees: Ricoh Company, Ltd.
- Inventors: Toshifumi Yamaai
- Web page: https://patentgrep.com/patent/US20010024520A1

## 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

The character recognition apparatus incorporates a table recognition section. This table recognition section extracts circumscribing rectangles from connected components of black pixels on a whole surface of the image as an area where read image data are desired to be processed, sorts rectangles of character candidates and rectangles of frame candidates from the extracted circumscribing rectangles, fills the rectangles sorted as the character candidates with white pixels, recognizes and extracts frames as ruled lines from the rectangles which remain as the frame candidates.

## Claims

### Claim 1

A table recognition method of recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition method comprising the steps of:
extracting circumscribing rectangles of connected components of black pixels of the document image;
sorting the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
filling the image inside the area of the rectangles which have been sorted as the character candidates with white; and
recognizing frames from the rectangles which have been sorted as the frame candidates.

### Claim 2

A table recognition method of recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition method comprising the steps of:
extracting circumscribing rectangles of connected components of black pixels of the document image;
sorting the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
filling the image inside the area of the rectangles which have been sorted as the character candidates with white;
rearranging the rectangles sorted as the frame candidates in order of increasing area;
extracting frames successively from the rectangles sorted as the frame candidates in the order of increasing area; and
repeatedly filling the images of the rectangles which have been sorted as the frame candidates from which the frames were extracted with white.

### Claim 3

A table recognition method of recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition method comprising the steps of:
extracting circumscribing rectangles of connected components of black pixels of the document image;
sorting the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
filling the image inside the area of the rectangles which have been sorted as the character candidates with white;
rearranging the rectangles sorted as the frame candidates in order of increasing area;
extracting frames successively from the rectangles sorted as the frame candidates in the order of increasing area;
when a number of the rectangles sorted as frame candidates is not more than two and they do not establish a positional relationship that they are partially overlapped with each other, filling the image inside the area of the rectangles which have been sorted as the character candidates with white; and
when a number of the rectangles sorted as frame candidates is not more than two and they establish a positional relationship that they are partially overlapped with each other, suppressing filling the image inside the area of the rectangles which have been sorted as the character candidates with white and extracting a frame from the rectangle of next candidate.

### Claim 4

A table recognition apparatus for recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition apparatus comprising:
a circumscribing rectangle extracting section which extracts circumscribing rectangles of connected components of black pixels of the document image;
a rectangle sorting section which sorts the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
a frame candidate extracting section which fills the image inside the area of the rectangles which have been sorted as the character candidates with white; and
a ruled line extracting section recognizes frames from the rectangles which have been sorted as the frame candidates.

### Claim 5

A table recognition apparatus for recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition apparatus comprising:
a circumscribing rectangle extracting section which extracts circumscribing rectangles of connected components of black pixels of the document image;
a rectangle sorting section which sorts the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
a frame candidate extracting section which fills the image inside the area of the rectangles which have been sorted as the character candidates with white; and
a ruled line extracting section which extracts frames from the rectangles sorted as the frame candidates successively in order of increasing area and repeatedly filling the images of the rectangles which have been sorted as the frame candidates from which the frames were extracted with white.

### Claim 6

A table recognition apparatus for recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition apparatus comprising:
a circumscribing rectangle extracting section which extracts circumscribing rectangles of connected components of black pixels of the document image;
a rectangle sorting section which sorts the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
a frame candidate extracting section which fills the image inside the area of the rectangles which have been sorted as the character candidates with white; and
a ruled line extracting section which rearranges the rectangles sorted as frame candidates in order of increasing area, extracts frames from the rectangles sorted as frame candidates successively in order of increasing area, when a number of the rectangles sorted as frame candidates is not more than two and they do not establish a positional relationship that they are partially overlapped with each other, fills the image inside the area of the rectangles which have been sorted as the character candidates with white, and when a number of the rectangles sorted as frame candidates is not more than two and establish a positional relationship that they are partially overlapped with each other, suppresses filling of the image inside the area of the rectangles which have been sorted as the character candidates with white and extracts a frame of the rectangle of next frame candidate.

### Claim 7

A character recognition apparatus comprising a table recognition apparatus for recognizing a frame as a ruled line in a table area of a document image including the table area, said table recognition apparatus including,
a circumscribing rectangle extracting section which extracts circumscribing rectangles of connected components of black pixels of the document image;
a rectangle sorting section which sorts the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
a frame candidate extracting section which fills the image inside the area of the rectangles which have been sorted as the character candidates with white; and
a ruled line extracting section recognizes frames from the rectangles which have been sorted as the frame candidates.

### Claim 8

A character recognition apparatus comprising a table recognition apparatus for recognizing a frame as a ruled line in a table area of a document image including the table area, said table recognition apparatus including,
a circumscribing rectangle extracting section which extracts circumscribing rectangles of connected components of black pixels of the document image;
a rectangle sorting section which sorts the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
a frame candidate extracting section which fills the image inside the area of the rectangles which have been sorted as the character candidates with white; and
a ruled line extracting section which extracts frames from the rectangles sorted as the frame candidates successively in order of increasing area and repeatedly filling the images of the rectangles which have been sorted as the frame candidates from which the frames were extracted with white.

### Claim 9

A character recognition apparatus comprising a table recognition apparatus for recognizing a frame as a ruled line in a table area of a document image including the table area, said table recognition apparatus including,
a circumscribing rectangle extracting section which extracts circumscribing rectangles of connected components of black pixels of the document image;
a rectangle sorting section which sorts the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
a frame candidate extracting section which fills the image inside the area of the rectangles which have been sorted as the character candidates with white; and
a ruled line extracting section which rearranges the rectangles sorted as frame candidates in order of increasing area, extracts frames from the rectangles sorted as frame candidates successively in order of increasing area, when a number of the rectangles sorted as frame candidates is not more than two and they do not establish a positional relationship that they are partially overlapped with each other, fills the image inside the area of the rectangles which have been sorted as the character candidates with white, and when a number of the rectangles sorted as frame candidates is not more than two and establish a positional relationship that they are partially overlapped with each other, suppresses filling of the image inside the area of the rectangles which have been sorted as the character candidates with white and extracts a frame of the rectangle of next frame candidate.

### Claim 10

A computer readable medium for storing instructions, which when executed on a computer, causes the computer to realize a table recognition method of recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition method comprising the steps of:
extracting circumscribing rectangles of connected components of black pixels of the document image;
sorting the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
filling the image inside the area of the rectangles which have been sorted as the character candidates with white; and
recognizing frames from the rectangles which have been sorted as the frame candidates.

### Claim 11

A computer readable medium for storing instructions, which when executed on a computer, causes the computer to realize a table recognition method of recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition method comprising the steps of:
extracting circumscribing rectangles of connected components of black pixels of the document image;
sorting the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
filling the image inside the area of the rectangles which have been sorted as the character candidates with white;
rearranging the rectangles sorted as the frame candidates in order of increasing area;
extracting frames successively from the rectangles sorted as the frame candidates in the order of increasing area; and
repeatedly filling the images of the rectangles which have been sorted as the frame candidates from which the frames were extracted with white.

### Claim 12

A computer readable medium for storing instructions, which when executed on a computer, causes the computer to realize a table recognition method of recognizing a frame as a ruled line in a table area of a document image including the table area, the table recognition method comprising the steps of:
extracting circumscribing rectangles of connected components of black pixels of the document image;
sorting the extracted circumscribing rectangles as character candidates and frame candidates according to information such as an aspect ratio, a number of black pixels and a number of black runs of the extracted circumscribing rectangles;
filling the image inside the area of the rectangles which have been sorted as the character candidates with white;
rearranging the rectangles sorted as the frame candidates in order of increasing area;
extracting frames successively from the rectangles sorted as the frame candidates in the order of increasing area;
when a number of the rectangles sorted as frame candidates is not more than two and they do not establish a positional relationship that they are partially overlapped with each other, filling the image inside the area of the rectangles which have been sorted as the character candidates with white; and
when a number of the rectangles sorted as frame candidates is not more than two and they establish a positional relationship that they are partially overlapped with each other, suppressing filling the image inside the area of the rectangles which have been sorted as the character candidates with white and extracting a frame from the rectangle of next candidate.
