JPS61255483A - Character recognition device - Google Patents

Abstract

PURPOSE:To use a slip using a conventional print ink by identifying a full area of a picture data for one character's content including a character describing frame from a background area and using data of the difference area and the background area so as to detect the character describing frame. CONSTITUTION:The full area surrounded by black picture elements is identified from a picture data for one character's content including the character describing frame at the 1st identification section 71, while a white picture element area surrounded by black picture elements is identified by the 2nd identification section 72, the results are given to the 3rd identification section 73, where the 3rd area belonging to the 1st area identified by the section 71 and not belonging to the 2nd area identified by the section 72 is identified. The 3rd and 2nd areas are processed by the 1st processing section 74 to detect the picture element area for the character describing frame and a character data separated from the input picture data is extracted by the 2nd processing section 75.

Description

[Detailed Description of the Invention] [Summary] In a character recognition device such as an optical character reader (OCR) that cuts out and recognizes each character written on a form, a 9-character entry frame can be filled with 9 characters without using a dropout color. This is designed to reduce the cost of forms by allowing the use of forms that use normal printing ink.

[Industrial application field]

The present invention relates to a character recognition device, and particularly to a character recognition device using a form on which a character entry frame is printed.

A character recognition device, typically an optical character reader (OCR), uses an observation device such as a photoelectric conversion scanner to observe a character pattern written on a form by hand or by printing, and uses the resulting image data. We took the image data of one character, and extracted the characteristics of that character pattern.

This is a device that determines the character type of a character pattern through a procedure such as extracting each character type from a standard character pattern in advance, storing it in a recognition dictionary, and comparing it with a certain standard feature.

The process of extracting the image data of one character is called character extraction or segmentation, and in order to facilitate this process, multi-character recognition devices use forms with character entry frames printed on them. .

Since the character entry frame should not be included in the recognition target, forms are often used that are printed using dropout (printing using ink that is visible to humans but cannot be observed by the observation unit).

In response to this, in order to reduce the cost of forms.

There is a character recognition device that uses a form printed with ordinary ink and is configured to separate the character entry frame after cutting out, but in this case, it must be configured to ensure that the two character entry frames are separated. It won't happen.

[Conventional technology]

FIG. 4 is a block diagram of a conventional example of a character recognition device using a form in which a character entry frame is printed with ordinary ink.

1 observes a form with a single character pattern written on it by rask scanning and detects 2 image constituent pixels (hereinafter referred to as r black pixels J).
is “1” ・Background constituent pixels (hereinafter referred to as “white pixel j”)
An observation unit that outputs binary image data with "0".

Reference numeral 2 denotes a cutting unit that extracts one character's worth of image data including a seven character entry frame from, for example, one character line's worth of image data obtained by the observation unit 1.

31, 32, 33, 34, and 35 separate nine character entry frames from the image data for one character extracted by the extraction unit 2 (black pixel "1" corresponding to the character entry frame is separated from white pixel "1"). 0”) is a component of the character entry frame separation unit.

Reference numeral 31 denotes an input buffer that stores image data for one character extracted by the extraction unit 2 in the form of a two-dimensional array.

Reference numeral 32 denotes a projection unit (see FIG. 5) that calculates the horizontal and vertical projections of one character's worth of two-dimensional array image data stored in the input buffer 31.

33 is a detection unit that detects pixels corresponding to the 9-character entry frame from the projection obtained by the projection unit 32;

34 is the image data stored in the input buffer 31,
A processing unit that separates the character entry frame detected by the detection unit 33.

Reference numeral 35 denotes an output sofa for storing image data from which character entry frames have been separated by the processing unit 34.

4 is from the image data stored in the output buffer 34,
A feature extraction unit that extracts features of the character pattern.

5 is a feature dictionary that stores features extracted in advance from standard character patterns for each of two character types.

6 is the feature extracted by the feature extraction unit 4.

By comparing the stored contents of the feature dictionary 5.

This is a matching unit that determines the character type of the character pattern.

[Problem that the invention seeks to solve]

The character recognition device configured as described above detects pixels corresponding to a single character entry frame from the horizontal and vertical projection of image data for one character including a single character entry frame, and separates this from the original image data. It is something.

For this reason, for example, if the form is tilted, the image data cut out by the cutout section 2 will also be tilted, and pixels corresponding to the character entry frame will not be correctly detected and removed, as illustrated in FIG. However, there is a problem that the recognition rate decreases.

Therefore, it is an object of the present invention to improve the recognition rate in a character recognition device that inputs a character pattern along with a two-character entry frame.

[Means for solving problems] FIG. 1 is a block diagram of the principle of the present invention.

Reference numeral 71 denotes a first identification unit that identifies the entire area surrounded by black pixels in one character's worth of image data including a two-character entry frame.

A second identification unit 72 identifies a white pixel area surrounded by black pixels in the image data.

73 is a third identification for identifying a pixel belonging to a third area that belongs to the first area identified by the first identification unit 71 and does not belong to the second area identified by the second identification unit 72; Department.

74 is a first processing unit that detects pixels corresponding to the character entry frame from the third area identified by the third identification unit 73 and the second area identified by the second identification unit 72;

Reference numeral 75 denotes a second processing unit that separates pixels corresponding to the character entry frame detected by the first processing unit 74 from the image data for one character including the character entry frame.

[Effect]

That is, the first identification unit 71 scans the two-dimensional array image data for one character including the character entry frame as illustrated in FIG. 5, with the vertical direction as the main scanning direction. By identifying the range from the detected black pixel to the last detected black pixel, the area shown in Fig. 2(a) is obtained, and similarly, by rask scanning with the horizontal direction as the main scanning direction, the mountain ) is obtained, and a first region (C) such as (C) is obtained as a common region between (a) and middle).

The second identification unit 72 scans the same image data with the vertical direction as the main scanning direction, and identifies a white pixel area sandwiched between black pixels.

Find the area as shown in Fig. 2(d), and similarly perform rask scanning with the horizontal direction as the main scanning direction to find the area as shown in (8), and as the area common to (d) and (e), Find the second region ■ such as f).

The third identification unit 73 determines a third area (hatched area) as shown in FIG.

Next, in the first process i74, as shown in FIG. 2(g), the image data identified as the second area (2) and the third area (2) is rask-scanned from all four directions, vertically and horizontally.

Among the pixels belonging to the third area (3), the pixels corresponding to the 9-character entry frame are detected by detecting the pixels that are detected first in each scan and are consecutive at a predetermined number or less, excluding the first detected pixels.

As a result, pixels in the area shown by hatching in FIG. Image data as shown in 1) is obtained.

In the following, the input character pattern is recognized by extracting features in the same manner as in the conventional example and comparing them with the features of the standard character pattern.

〔Example〕

FIG. 3 is a configuration diagram of the main parts of the embodiment.

The first identification unit 71 includes a scanning unit 71a that scans the two-dimensional array image data stored in the input buffer 31.
, pants 771b that stores the pixels identified in the vertical scan, a buffer 71c that stores the pixels identified in the horizontal scan, and pixels that are common to the storage contents of the buffer 71b and the buffer 71c are extracted. and an AND circuit 71d.

The second identification unit 72 includes a scanning unit 72a that scans the two-dimensional array image data stored in the input buffer 31.
, a buffer 72b that stores the pixels identified in the vertical scan, a buffer 72c that stores the pixels identified in the horizontal scan, and pixels that are common to the storage contents of the buffer 72b and the buffer 72c are extracted. and an AND circuit 72d.

The first processing unit 74 processes the image data to which the identification of the second area ■ obtained by the second identification unit 71 and the identification of the third area ■ obtained by the third identification unit 73 are attached. An input cover sofa 74a stores image data in a two-dimensional array, a scanning unit 74b scans the image data stored in the input cover sofa 74a in four directions vertically and horizontally, and scans pixels corresponding to character entry frames extracted by each scan. External cover sofa 7 to store
4c.

〔Effect of the invention〕

As explained above, according to the present invention, even if the character entry frame cut out with a single character pattern is tilted, the
Since the character entry frame can be accurately separated and only the character pattern can be extracted accurately, the recognition rate can be improved.

[Brief explanation of drawings]

FIG. 1 is a block diagram of the principle of the present invention. FIGS. 2(a) to 2(1) are explanatory diagrams of the action. FIG. 3 is a configuration diagram of the main parts of the embodiment. FIG. 4 is a configuration diagram of a conventional example. FIG. 5 is an explanatory diagram of a conventional example. FIG. 6 is an explanatory diagram of the problem. In the figure. ■ is the observation part, 2 is the cutting part. 4 is a feature extraction unit, and 5 is a feature dictionary. 6 is a collation unit, and 31 is an input buffer. 35 is an output buffer, and 71 is a first identification section. 72 is a second identification part, and 73 is a third identification part. 74 indicates a first processing section, and 75 indicates a second processing section. /Otsu A\≠Ku n month e genri vis 2 shi〕91 phantom Rod 2 mouths Female sewage example e bamboo silk 2 Dou 52 Antsae 1 Hi Seikyu District Rod 6 days

Claims (1)

[Claims] A first identification unit (
71), a second identification unit (72) for identifying a background pixel area surrounded by image constituent pixels in the image data, and a second identification unit (72) for identifying a background pixel area surrounded by image constituent pixels in the image data; A third identification unit (73) that identifies pixels belonging to a third area that does not belong to the second area identified by the identification unit (72) of
), and a first processing unit (7) that detects pixels corresponding to the character entry frame in the image data from the second area and the third area.
4); and a second processing unit (75) that separates pixels corresponding to the detected character entry frame from the image data.