JPH06231306A - Character recognition device - Google Patents

Abstract

PURPOSE:To provide a character recognition device by which a document including a dot character can easily be recognized and which is superior in operability. CONSTITUTION:A primary character area judgement part 1 obtaining the circumscribing rectangle of a connected graphic from a primary binary picture obtained by reading the document by a scanner and judging a character area with the size, a dot area extraction part 2 extracting a dot area from the rate of the change point of black and white as against the whole size from a non-character area, a secondary character area judgement part 3 judging the character area from a secondary binary picture obtained by reading palely the dot area by the scanner in the same way as the primary character area judgement part 1, a character segment part 5 segmenting a character pattern from primary/ secondary character areas, a character characteristic extraction part 6 extracting a characteristic from the character pattern, a recognition certainty calculation part 8 comparing the character characteristic with a character characteristic dictionary 7 storing all the character characteristics and obtaining certainty and a recognition character decision part 9 deciding the recognized character from recognition certainty are provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for reading a document or the like and converting the character into a corresponding character code, in which a dot is applied to the character (hereinafter referred to as a dot character). The present invention relates to a character recognition device capable of character-recognizing a document including a character.

[0002]

2. Description of the Related Art In recent years, with the widespread use of word processors and the like, halftone dot characters have come to be frequently used for emphasizing characters in a document, and a document including the halftone dot characters can be recognized. Character recognition devices are being developed.

A conventional character recognition device will be described below. 3 is a diagram showing a recognition target document including halftone dots, and FIG. 4 is a diagram showing a binary image when the recognition target document shown in FIG. 3 is read from a scanner at a normal density.
FIG. 5 is a diagram showing a binary image when the recognition target document shown in FIG. 3 is read from a scanner with a light density.

When the conventional character recognition apparatus attempts to character-recognize a document to be recognized as shown in FIG. 3, when it is read with a normal density from a scanner, a binary image as shown in FIG. 4 is obtained. Therefore, the normal character portion can be recognized, but the halftone dot character portion cannot be recognized. On the other hand, if the image is read from the scanner at a low density, a binary image as shown in FIG. 5 is obtained, and the halftone dot character portion can be recognized.
The normal character part becomes unrecognizable.

Therefore, in the case of character recognition of a recognition target document including halftone dots as shown in FIG. 3 by the conventional character recognition device, first, the recognition target document is read from a scanner at a normal density and then the normal recognition is performed. The character portion is character-recognized, then the user adjusts the density of the scanner lightly, the document to be recognized is read again from the scanner lightly, and the halftone dot character portion is character-recognized. Character recognition work is performed by integrating the recognition results.

[0006]

However, in the above-mentioned conventional configuration, when the recognition target document including halftone characters is character-recognized, after the recognition target document is read from the scanner at a normal density, the user scans it. Adjust the density of the image to be thin and read the document to be recognized again from the scanner.
Since the two recognition results have to be integrated, there has been a problem that it is complicated, laborious, and lacks in workability.

An object of the present invention is to solve the above-mentioned conventional problems, and an object thereof is to provide a character recognition apparatus having excellent workability, which can easily recognize a recognition target document including halftone characters. .

[0008]

In order to achieve this object, the character recognition apparatus of the present invention is capable of adjusting the density when reading a document to be recognized and outputting a binary image and / or the object to be recognized. A scanner capable of reading a document as multivalued data, and a primary binary image which is read by the scanner with a normal density from the scanner, or read as multivalued data and binarized by a normal threshold value and output. A primary character area determination unit that determines a circumscribed rectangle of the connected figure from the size of the circumscribed rectangle and a primary character area determination unit that determines the primary character area from the non-character area determined by the primary character area determination unit, The halftone dot area extracting unit for extracting the halftone dot area according to the change point ratio, and the recognition target document of the halftone dot area portion extracted by the halftone dot area extracting unit are From the secondary binary image that is read again at a density lower than the normal density than the scanner, or the read multi-valued data of the halftone dot area is binarized by the threshold set to be lighter than the normal threshold and output. A secondary character area determination unit that determines a secondary character area in the same manner as the primary character area determination unit, a primary character area determined by the primary character area determination unit, and a secondary character area determination unit by the secondary character area determination unit A character cutout unit that cuts out a character pattern according to the size and position of a circumscribing rectangle from a secondary character region, a character feature extraction unit that extracts character features from the character pattern cut out by the character cutout unit, and characters of all characters in advance. Recognition accuracy calculation for obtaining recognition accuracy of character candidates, similarity, etc. by comparing the character feature dictionary storing the features with the character features extracted by the character feature extraction unit and the character feature dictionary If has a configuration that includes a recognition character determining unit that determines the recognized character from the recognition accuracy obtained in the recognition accuracy calculation unit.

[0009]

With this configuration, the primary character area determination unit determines the primary character area, and the secondary character area determination unit rereads the halftone dot area extracted by the halftone dot area extraction unit with a light density from the scanner. Alternatively, the read multi-valued data is binarized by a threshold value that is set lightly, the secondary character area is determined from the output secondary binary image in the same manner as the primary character area determination unit, and the primary character area and the secondary character area are determined. By recognizing the character in the next character area, it is possible to easily and automatically recognize the recognition target document including halftone characters.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A character recognition device according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of a character recognition device in one embodiment of the present invention. Reference numeral 1 is a circumscribed circumscribed rectangle obtained by reading a recognition target document with a normal density from a scanner (not shown) to obtain a circumscribed rectangle of a connected figure from a primary binary image and integrating the circumscribed rectangles according to their distances and sizes. A primary character area determination unit that determines a primary character area and a non-character area based on the size of a rectangle, 2 is a ratio of the overall size and a black-and-white change point from the non-character area determined by the primary character area determination unit 1. A halftone dot area extracting unit 3 for extracting a halftone dot area by means of a scanner (not shown) rereads the recognition target document of the halftone dot area portion extracted by the halftone dot area extracting unit 2 at a density lower than the normal density. The secondary character area determination unit 4 that determines a secondary character area and a non-character area from the secondary binary image obtained in step 1 in the same manner as the primary character area determination unit 1, and the primary character area determination unit 4 determines the primary character area determination unit 1. Character area and secondary character area judgment The character line extraction unit 5 that extracts the character lines by projecting the vertical and horizontal projections of the circumscribed rectangle in the secondary character area determined in step 3 is the character line width and circumscribed line determined by the character line extraction unit 4. A character cutout unit that estimates the size of a character from the size of the rectangle and merges the circumscribing rectangle based on the size of the character to remove noise components according to the size and distance, and cuts out as a character pattern, 6 is a character cutout unit A character feature extraction unit for extracting character features from the character pattern cut out in 5, a character feature dictionary 7 in which character features of all characters are stored in advance, and 8 a character feature extracted by the character feature extraction unit 6. A recognition accuracy calculation unit that compares the character feature dictionary 7 with the recognition accuracy such as character candidates and similarity, and a recognition character determination unit 9 that determines a recognition character from the recognition accuracy calculated by the recognition accuracy calculation unit 8. .

The operation of the character recognition device having the above-described structure according to the embodiment of the present invention will be described below. FIG. 2 is a flowchart of the character recognition device in one embodiment of the present invention. First, the primary character area determination unit 1 obtains a circumscribed rectangle of a connected figure from a primary binary image input by reading a recognition target document with a normal density from a scanner (not shown), and determines the circumscribed rectangle by the distance. , The primary character area and the non-character area are determined based on the size of the circumscribed rectangle after integration (S1). Next, the halftone dot area extraction unit 2 extracts a halftone dot area from the non-character area determined in S1 based on the overall size and the ratio of black and white change points (S2). Next, the secondary character area determination unit 3 reads the recognition target document of the halftone dot area portion extracted in S2 again with a density lower than normal density from a scanner (not shown) and inputs the secondary character. The circumscribing rectangle of the connected figure is obtained from the binary image, the circumscribing rectangle is integrated according to its distance and size, and the secondary character area and the non-character area are determined based on the size of the circumscribing rectangle after integration (S3). . Next, the character line extraction unit 4 takes a histogram of the circumscribing rectangle in the vertical and horizontal directions for the primary character region determined in S1 and the secondary character region determined in S3, and extracts the character line (S4). ). Next, the character cutout unit 5 estimates the character size from the distribution of the character line width and the size of the circumscribing rectangle extracted in S4 (S5). Next, the character slicing unit 5 merges the circumscribed rectangles by removing the noise component from the size and distance based on the character size estimated in S5, and cuts out as a character pattern (S6). Next, the character feature extraction unit 6 extracts character features from the character pattern cut out in S6 (S7). Next, the recognition probability calculation unit 8 compares the character features extracted in S7 with the character feature dictionary 7 to obtain recognition probabilities such as character candidates and similarities (S8). Next, the recognized character determination unit 9 determines the character having the highest degree of similarity among the character candidates obtained in S8 as a recognized character (S9).

In this embodiment, the primary character area determination unit 1 and the secondary character area determination unit 3 change the densities at the time of reading and read the document to be recognized twice by a scanner (not shown). However, it is assumed that the scanner (not shown) can read the recognition target document as multivalued data, and the primary character area determination unit 1 reads the recognition target document as multivalued data from this scanner (not shown). , A primary binary image is obtained by binarizing with a normal threshold value, and the secondary character area determination unit 3 binarizes the multi-valued data read by the primary character area determination unit 1 with a threshold value that is thinner than usual to obtain a secondary image. Two
It is preferable to obtain the value image because the number of times the document to be recognized is read by the scanner (not shown) is only once, and the time required for character recognition work is shortened.

If information indicating that the character was a halftone dot character is added to the recognized character recognized from the secondary character area, the recognition result makes the halftone character in the document to be recognized. It can easily detect the part that was
This is preferable because it is possible to display the recognition result as a halftone dot character when displaying the recognition result.

[0014]

As described above, according to the present invention, the primary character area determination unit determines the primary character area, and the secondary character area determination unit detects the halftone dot area extracted by the halftone dot area extraction unit from the scanner. The secondary character area is determined in the same way as the primary character area determination unit from the secondary binary image that is output by reading again with the density or by binarizing the read multi-valued data with a thinly set threshold value. By recognizing characters in the primary character area and the secondary character area, it is possible to realize a character recognizing device with excellent workability that can easily and automatically recognize a recognition target document including halftone characters. .

[Brief description of drawings]

FIG. 1 is a configuration diagram of a character recognition device according to an embodiment of the present invention.

FIG. 2 is a flowchart of a character recognition device according to an embodiment of the present invention.

FIG. 3 is a diagram showing a recognition target document including halftone characters.

FIG. 4 is a diagram showing a binary image when the recognition target document shown in FIG. 3 is read from a scanner at a normal density.

5 is a diagram showing a binary image when the recognition target document shown in FIG. 3 is read from a scanner with a light density.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Primary character region determination unit 2 Halftone dot region extraction unit 3 Secondary character region determination unit 4 Character line extraction unit 5 Character cutout unit 6 Character feature extraction unit 7 Character feature dictionary 8 Recognition accuracy calculation unit 9 Recognized character determination unit

Claims (1)

[Claims] 1. A scanner capable of adjusting density when reading a recognition target document and outputting a binary image and / or reading the recognition target document as multivalued data, and the recognition target document The circumscribed rectangle of the connected figure is obtained from the primary binary image that is read by the scanner with normal density, or is read as multi-valued data and binarized by the normal threshold value, and the primary character area is determined by the size of the circumscribed rectangle. A primary character area determination unit that determines the primary character area determination unit, a halftone dot area extraction unit that extracts a halftone dot area from the non-character area determined by the primary character area determination unit based on the ratio of black and white change points to the overall size, The document to be recognized in the halftone dot area portion extracted by the halftone dot area extraction unit is read again at a density lower than the normal density by the scanner, or the halftone dot area is read. The secondary character area is determined in the same manner as the primary character area determination unit from the secondary binary image output by binarizing the read multi-valued data of the part with a threshold value set to be thinner than the normal threshold value. A character that cuts out a character pattern according to the size and position of the circumscribed rectangle from the next character area determination unit, the primary character area determined by the primary character area determination unit, and the secondary character area determined by the secondary character area determination unit A cutout unit, a character feature extraction unit that extracts character features from the character pattern cut out by the character cutout unit, a character feature dictionary that stores character features of all characters in advance, and a character feature extraction unit that extracts the character features. A recognition accuracy calculation unit that obtains recognition accuracy such as a character candidate and similarity by comparing a character feature with the character feature dictionary, and a recognized character is determined from the recognition accuracy obtained by the recognition accuracy calculation unit. Character recognition apparatus characterized by comprising a that recognition character determining unit.