JPH02220185A - Document image recognizing device - Google Patents

Abstract

PURPOSE:To attain a document syntactic analysis with the use of position information on a host device side, and to improve recognizing ratio by means of the addition of a meaning processing, etc., by outputting the coordinate data of a converted character or a character string. CONSTITUTION:A picture processing part 4 extracts a character string area from inputted character image data, a character string area 8 is expressed by coordinates (X1, Y1) and (X2, Y2) of two points A and B at mutually opposite angles of a rectangle, and the extracted character string area 8 is identified as the sets of coordinate values. The image processing part 4 extracts a character image 6 from each character image area by means of a projecting method, etc., and each extracted character image 7 is impressed to a character recognizing part 5, and converted into a character code. Further the image processing part 4 stores the character code to be the recognition result of the character recognizing part 5 corresponding to the coordinate data of the character image 7, and based on the recognition result, text data are generated. Thus the recognition ratio can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a document image recognition device that converts a document image into text data.

[Prior Art] Document image recognition devices that convert document image data into text data have been put into practical use by reading standard slips and manuscript paper, and in recent years there have been some changes in the size and arrangement of characters. A device that can also be read has been developed.

[Problems to be Solved by the Invention] In the conventional document image recognition device described above, the main function is to output recognition results as text data, so recognition processing using information regarding character arrangement is difficult. It was inconvenient. Furthermore, for characters rejected as a result of single character recognition, only the candidates are output, and the original image of the character needs to be confirmed by looking at the document, which poses a problem in operability.

[Means for Solving the Problems] The present invention has been made to solve the above problems, and is directed to a document image recognition device that converts a document image consisting of pixel data into a text consisting of character code data. The present invention proposes a document image L&I device that outputs coordinate data of characters or character strings to be converted, or character image data.

[Function] The document image recognition device of the present invention receives a signal from an image processing unit that extracts a character string region and a character image from a document image that is input data, and a signal from a character recognition unit that converts the character image into a character code. This outputs character strings, character coordinate data, and character image data.

[Example] Embodiments of the present invention will be described with reference to the accompanying drawings.

FIG. 1 is a schematic block diagram of the entire document image recognition apparatus of the present invention, and FIG. 2 is an explanatory diagram of characters.

Reference numeral 1 is a document image recognition device, which includes an image processing section 4 and a character recognition section 5, and a scanner 3 is a document image recognition device I! 1, and the host computer 6 processes the text data, character string coordinate data, character coordinate data, and character image data output by the document image recognition device 1. It is a device for A document image of document 2 is captured by scanner 3 and applied to image processing section 4 of document image recognition device 1 . This image processing unit 4 extracts a character string region from the input character image data by a method such as a contour tracing method or a projection method. These are rectangular areas 8 lined up at a distance. Also, this character string area 8 consists of two diagonal points A and B of a rectangle.
It is represented by the locus 4m (Xi, YL) (Xis Y3). The extracted character string area 8 is identified as this set of coordinate values.

The image processing unit 4 extracts a character image 7 from each character string area using a method such as a projection method.The 0 character image 7 is an image of a rectangular area surrounding an image of one character. 8, coordinate data is obtained from the coordinates of two diagonal points, and each character image 7 is identified by the combination of these coordinate values. Further, each extracted character image 7 is applied to a character recognition section 5 and converted into a character code.0 The character recognition section 5 is cold-represented by a known character recognition method.1 Image processing section 4 is a character recognition section. The character code that is the recognition result of No. 5 is stored in correspondence with the coordinate data of character image 7.

The image processing unit 4 generates text data based on the above recognition results according to a first-order rule.

(1) Character codes are arranged according to the order of character images in the character string area 8 to generate character string data.

(2) Character string regions that have overlapping x (y) coordinates with character string regions are treated as character strings belonging to the same line.

Combine string data with spaces etc. to create string data representing a line.

(3) Add a line feed code to the end of the line.

If text data is to be output without position information, the text data generated in the above procedure may be output as is.

Next, when outputting text data with position information added, the output data is configured and output as in the example of the linear equation.

String 1...xi, yt, xi, yi String n...J, B* The position information for Xp, Yp characters is the same if you replace the string in the above formula with the character code. It can be expressed as Furthermore, when outputting a character image, the image data may be encoded and added according to the above formula.

[Effects of the Invention] Since the present invention has the above configuration, it is possible to output not only text data but also character strings and character position information at the same time. Document structure analysis is now possible.

The recognition rate can be improved by adding semantic processing, etc. Furthermore, since character images can also be output, it is extremely effective when correcting characters, etc. on the host machine side.

[Brief explanation of the drawing]

The drawings show one embodiment of the present invention, and FIG. 1 is a schematic block diagram, and FIG. 2 is an explanatory diagram of characters. 1...
・Document image recognition device, 2... document, 3... scanner, 4... image processing section, 5... character recognition section,
7... Character image. 8...Character string area

Claims (1)

[Claims] 1) A document image recognition device that converts a document image consisting of pixel data into a text consisting of character code data, which outputs coordinate data of a character or character string to be converted. recognition device. 2) A document image recognition device that converts a document image made of pixel data into text made of character code data, and outputs the converted character image data.