JP2630261B2 - Character recognition device - Google Patents

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 recognizing characters written on a form, and more particularly to a character recognition device capable of accurately recognizing a contacted character.

[0002]

2. Description of the Related Art Conventionally, in a character recognition device for recognizing a character entered on a form, when recognizing a contact character, the height of the character is measured from an input character pattern, and the character height is measured. Character width is estimated and cut out in units of one character for recognition, or touched characters are divided and recognized at positions where the histogram of black pixels constituting a character is equal to or less than a predetermined threshold (for example, Japanese Patent Laid-Open No. -343192).

[0003]

In the above-mentioned conventional method, when recognizing a contact character, the character width is estimated from the character height and the character is forcibly cut out in units of one character. It is not always correct, and there is a drawback that it is not possible to make an erroneous determination or recognize depending on the position where the contact character is cut out in units of one character. When the black pixel histogram is used, the valley V of the histogram is used as the cutout position as shown in FIG. 5, but the valley V may not appear as shown in FIG. In such a case, the cutout position cannot be determined. Therefore, the contact character cannot be read correctly.

The present invention has been made in view of the above points, and has as its object to enable accurate recognition of a contact character.

[0005]

In order to achieve the above object, according to the present invention, a character chunk is extracted from a binarized image sent from a photoelectric conversion means, and whether or not the extracted character chunk is one character is determined. Detects whether a character is a character, and when the character block is a contact character, cuts out the contact character for each character and recognizes it, and also recognizes it in multiple character units. The touch character is recognized based on both results.

[0006]

The contact character is recognized based on both the result of recognition of the extracted character block for each character and the result of recognition for each of a plurality of characters. Thereby, the contact character can be more accurately recognized.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described with reference to the drawings. FIG. 1 is a functional block diagram showing one embodiment of the present invention, which is specifically composed of an image scanner, a memory, a computer, and the like.

In FIG. 1, a photoelectric conversion unit 1 binarizes an image on a form, and converts a binarized form image as shown in FIG. And send to The one-character extracting unit 2 extracts a character block from the binarized image input from the photoelectric conversion unit 1. Specifically, by extracting rectangular features by tracking black pixel connections of characters, FIG.
The circumscribing rectangles 31 and 32 as shown in (B) are extracted, and the height, width and pitch of the character block are measured. Here, rectangle extraction by tracking black pixel connections is a well-known technique in this field. If there is a contact character in the character block, the one-character cutout unit 2 cuts out the contact character for each character as shown in FIG. The one-character recognizing unit 4 performs character recognition for each cut-out character.

The contact character detection unit 3 includes circumscribed rectangles 31 and 32
A contact character is detected from among the contact characters (FIG. 2).
(Rectangle 32 of (B) and (D)) is sent to the contact character recognition unit 5. When the number of touching characters is more than two, the touching character detecting unit 3 cuts out the touching characters for two characters and sends them to the recognizing unit 5. Character information (height, width, pitch, etc. of a character block) is exchanged between the contact character detection unit 3 and the one-character cutout unit 1.

[0010] As shown in FIG.
It comprises a pattern input unit 51, a two-character reading dictionary 52 prepared in advance with two characters as one category, and a comparing unit 53. The two-character reading dictionary 52 stores a pattern in units of two characters.
By comparing the input pattern from No. 1 with the comparing unit 53, the contact character 32 as shown in FIG. 2D is recognized in units of two characters.

The judgment unit 6 receives the recognition results of the one-character recognition unit 4 and the contact character recognition unit 5 and outputs a final judgment result 7 based on the recognition results. When the characters cannot be read by the one-character recognizing unit 4 and the contact character recognizing unit 5, the determining unit 6 instructs the one-character extracting unit and the contact character detecting unit 3 to change the contact character extracting position. Try to cut out again.

Next, the operation of the embodiment will be described with reference to the flowchart of FIG. First, binarized image data is input from the photoelectric conversion unit 1 (S101). Next, a character block is extracted from this data, and the height H, width W, and pitch of the character are measured (S102). Next, based on the character height H and the character width W, HK <W (K: constant) for each character block
It is checked whether it is (S103). If H · K <W, the character width W is small, so it is determined that the character is not a contact character, and normal one-character recognition is performed (S104).

If HK <W, it is regarded as a contact character, and a character block is cut out so that the character width = HK (S105). The character pattern thus separated is as shown in FIG. The separated characters are recognized one by one (S106). It is determined whether the one character can be read (S107). If the one character can be read, the process proceeds to the next step S108. If reading is not possible, the parameter is changed (S109), one character is cut out at a position different from the previous position (S105), and recognition is performed again using the newly cut out character (S106). An example of the parameter change at this time is to change the constant K. Alternatively, a cutout position obtained from the black pixel histogram may be used. FIG. 5 shows an example of a black pixel histogram of a contact character. Since the valley V of the histogram can be determined as a separation position between characters, the position thus obtained is determined in step S109. Can be used.

Returning to step S103, if HK <W (that is, a contact character), it is further checked whether or not this contact character is more than two characters. That is,
In step S110, it is determined whether or not 2H · K <W, and if 2H · K <W, it is considered that there are more than two characters.
The character block is cut out every two characters (S111). For example, cutting is performed so that the character width becomes 2H · K, and the contact character recognition unit 5 performs character recognition for each two characters (S
112). Of course, if 2H · K <W is not satisfied, that is, if the contact character is two characters, step S111 is skipped and character recognition is performed without cutting out (S112).
If the two characters can be read (S113), the read result is compared with the result read one character at a time in step S108. The results of the one-character recognition (S106) and the two-character recognition (S112) are stored in an appropriate storage unit, and when the recognition results of a plurality of characters (two characters in the embodiment) appear, the two are compared. (S108).

If the reading is impossible in step S113, the parameter is changed (S114), the cutout position is changed (S111), and the reading is performed again (S112).
As a parameter change, for example, there is a method of changing the above K, but it is not limited thereto.

In step S115, if the result of the one-character recognition (S106) and the result of the two-character recognition (S112) match, the value is determined as a determination result (S116). If not, one of them is adopted as the determination result. In the case of the embodiment, the result of one-character recognition is adopted as the final determination result (S117).

In the above embodiment, 2 is used as the contact character.
Although an example of characters has been described, the present invention is not limited to two characters, and may include more characters than two characters.

[0018]

As described above, according to the present invention,
Since the determination is made based on the result of cutting out and recognizing the contact character in units of one character and the result of cutting out and recognizing the contact character in units of a plurality of characters, it is possible to recognize the contact character with high accuracy.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention.

FIG. 2 is a diagram illustrating a recognition unit of a character block.

FIG. 3 is a diagram illustrating a detailed configuration of a contact character recognition unit.

FIG. 4 is a flowchart illustrating the operation of the embodiment.

FIG. 5 is a diagram illustrating an example of obtaining a cutout position using a black pixel histogram.

FIG. 6 is a diagram illustrating an example in which a cutout position cannot be obtained even when a black pixel histogram is used.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Photoelectric conversion part 2 1 Character cut-out part 3 Contact character detection part 4 1 character recognition part 5 Contact character recognition part 6 Judgment part

Claims (3)

(57) [Claims] An extracting means for extracting a character chunk from the binarized image sent from the photoelectric conversion means; a means for detecting whether the extracted character chunk is one character or a contact character; Cut out one character at a time when it is a contact character 1
Includes a character extraction unit, 1 and a character recognition means for recognizing the extracted character in a character cutout section, and a recognizing touching character recognition means in multiple units of characters when said character lumps is touching characters, the When the character block is a contact character, the one character
Recognition means and the contact character recognition means.
Character recognition means and recognizes the touching characters. 2. The character according to claim 1, further comprising a determination unit configured to determine a result of the one-character recognition unit and a result of the contact character recognition unit, and changing a cutout position of the contact character according to a determination result of the determination unit. Recognition device. 3. A character chunk is extracted from a binarized image sent from the photoelectric conversion means, and whether the extracted character chunk is a single character or a contact character is detected. In addition, the touched character is cut out and recognized for each character, and the touched character is also recognized in units of a plurality of characters, and the touched character is recognized based on the results of both the recognition in units of one character and the recognition in units of a plurality of characters. Character recognition method.