JPH08255237A - Data storage device - Google Patents

Abstract

PURPOSE: To improve a character recognition rate and to store massive image data by discriminating a model which is newly inputted and having characters described with handwriting or typing, out of plural models in a model image storage device after inclination correction is performed. CONSTITUTION: First of all, various kinds of slips, on which nothing is described, are previously inputted by an image input device 1, erected by an image distortion corrector 2 and stored in a model storage device 3. Next, the new model, in which characters are described, is inputted as object data while using the image input device 1 and the image distortion corrector 2. An image discriminator 4 discriminates which model among plural model images stored in the model storage device 3 corresponds to the object data. The object image is turned to the image composed of only described characters by finding difference from the model image. This image composed of only the characters is compressed by an image compressing device 6 and stored in a digital image storage device 7 together with model discrimination information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data storage device for encoding a character portion in a character image input by using a facsimile, a scanner or the like and storing it in a storage medium.

[0002]

2. Description of the Related Art Conventionally, character recognition devices by various character recognition methods have been invented and put into practical use. Regarding character recognition, there are various things such as recognizing characters from a medium written in free format and recognizing the character part of a form that is predetermined as a slip, but in the latter case, a fill-in frame is often set in advance. Has been done.

[0003]

Usually, a facsimile or a scanner has a dropout color set,
Do not read characters or ruled lines printed in that color. Therefore OC
A user of R (Optical Character Reader) printed the entry frame using the color. However, the dropout color differs depending on the model, and it was necessary to purchase the same input device in order to operate the OCR device. In addition, when a fax is sent from an unspecified number of people, the entry frame is not sometimes dropped out, and the entered characters overlap the entry frame or partially pop out, which hinders character recognition. It was.

The drop-out color has a problem in that it is difficult for the writer to fill in because the printing fee for the OCR paper is expensive because the special color ink is used at the time of printing and the color is relatively light. Also, in consideration of future inquiries, it is necessary to store the slip itself or digitally store the read image. However, the storage of the slip itself is not appropriate because it requires a large amount of work and a storage location when searching. Digital storage is easy to search, but if the slip image is complicated, high compression cannot be expected even if image compression is performed, and there is a problem in storing a large amount of data.

[0005]

To solve this problem, a data storage device of the first invention is an image input device, an image distortion correction device, an image discrimination device, a template image storage device, a character recognition device, and an image compression device. Device and a digital data storage device. The image input device inputs a plurality of template images in advance, the image distortion correction device performs tilt correction, and the image data is then registered in the template image storage device. A plurality of templates in the template image accumulating device using the image discriminating device after tilt correction is performed using the image correcting device on a template newly input by using From the newly input template to obtain image data consisting of only characters, character parts are character coded by a character recognition device, and then the image compression device is used. It is characterized by the accumulation in the digital data storage device the image data and the character code consisting of only the characters compressed Te.

[0006]

The data storage device of the present invention comprises an image input device. This image input device is a facsimile device or a scanner device. The input image data is binarized, and in the case of a facsimile device, it is compressed by the compression method recommended by CCITT (International Telegraph and Telephone Consultative Committee). Data is output. In addition, the image distortion correction device erects the input image with a tilt at the time of input,
In general, many vertical and horizontal ruled lines are printed on a slip, and by detecting the ruled lines and determining the inclination, or in the case of print text, detecting the gap between lines and determining the inclination. A method of erecting an image is known.

The image input by the image input device is erected by the image distortion correction device. A plurality of slip images in which nothing is entered in advance are input using the image input device and the image distortion correction device,
It is stored in the template storage device as a template. In general, figures such as characters, ruled lines, and logo marks are printed on the slips. A new voucher in which characters such as the amount, the number, and the address are entered is input as target data using the image input device and the image distortion correction device. The image discriminating apparatus discriminates which template of the plurality of template images stored in the template storing apparatus corresponds to the target data.

Generally, a template
A technique called matching is used, which is a method of evaluating the overlapping ratio by overlapping. The target image is only the character image entered by taking the difference from the template image. The image of only this character image has the entry frame erased, the character to be recognized can be easily cut out, and is normally coded by the character recognition device and stored in the digital data storage device. Further, the image containing only the character image is stored in the digital image storage device together with the model identification information after being compressed by the image compression device, but the compression rate is remarkably improved as compared with the image to which the model is added. It is possible to store a large amount of image data.

The accumulated image data can be restored by superimposing it on a template image as needed, and can be used for confirmation of an inquiry.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The details of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of the apparatus of the present invention. In FIG. 1, reference numeral 1 is an image input device. This image input device is a facsimile device or a scanner device. The image input by facsimile is 200 dp in the main scanning direction according to the GIII standard.
i (dots per inch), 100dpi (dpi) in the sub-scanning direction, an A4 size document has an image of about 250 Kbytes, and the MR compression method recommended by CCITT (T4 recommendation)
Compressed with. 200 to 300d for both vertical and horizontal for scanner
After inputting with pi and binarizing, it becomes an image of about 500K-1MB.

Various kinds of slips in which nothing is entered in advance are input by the image input device 1 and uprighted by the image correction device 2 and then stored in the template storage device 3. The erect image of the slip is detected by detecting many vertical and horizontal ruled line components printed on the slip, obtaining the inclination of the ruled line component, and rotating the slip backward. The Hough transform is well known in the field of image recognition and widely used for line segment detection. Using Figure 2 Ho
Explain the ugh transformation. In FIG. 2, 21 is a black image,
When the entire image is scanned and a black dot is detected, a straight line 22 passing through the point and having various angles θ is drawn, and a perpendicular line 23 is drawn from the origin to the straight line to obtain a foot length r of the perpendicular line.

For example, assuming that the angle resolution is 1 degree, 360 pairs of (r, θ) are detected each time one black dot is detected.
Is obtained. The (r _k , θ _k ) of the peak position of the (r, θ) frequency distribution represents the ruled line component. In FIG. 1, a new model in which characters such as amount, number, and address are entered is input as target data using the image input device 1 and the image distortion correction device 2. In FIG. 1, reference numeral 4 denotes an image discriminating apparatus, and the image discriminating apparatus discriminates which template among the plurality of template images stored in the template storing apparatus 3 corresponds to the target data.

Template matching is used for image discrimination. This is to calculate the overlapping ratio for each model when the target data and various model images are superimposed, and specify the model showing the largest overlapping ratio as the corresponding model. When the images are vertically and horizontally displaced from each other, the overlapping ratio becomes small. However, the overlapping ratio is calculated by shifting the images vertically and horizontally within a predetermined range, and the maximum overlapping ratio is calculated and used as the overlapping ratio of the template. However, if a threshold value is set and the overlapping ratio of the specified template is equal to or less than the threshold value, it is considered that there is no corresponding template.

The target image is an image of only the characters entered by taking the difference from the template image. FIG. 3 illustrates the difference calculation. In FIG. 3, 30 is the target image, and 31
Is a corresponding template, and 32 is an image with only character images. The image including only the character image is coded by the character recognition device 5 of FIG. 1 and stored in the digital data storage device 6. In FIG. 4, (A) shows the numeral 41 of 9 entered in the character frame 40. In this case, since the numbers are in contact with the character frame, it has been difficult to cut out the character to be recognized, and the recognition rate is lowered.
(B) shows the numeral 42 of the image of only the character image, the character to be recognized can be easily cut out, and the character can be normally recognized.

Further, the image which is only the character image is shown in FIG.
After being compressed by the image compression device 6, the image data is stored in the digital image storage device 7 together with the template identification information. As the compression method, the MR compression method (T4 recommendation) or MMR compression method (T6 recommendation) recommended by CCITT in the field of facsimile is used, and compared with the uncompressed image data with template.
It can be compressed to about 1/100. The stored image data is restored by superimposing it on the template image retrieved from the template storage device 3 based on the template information as necessary, and the image display device, the image printing device, or the communication device is added to the device of the present invention. By adding it, it can be used for confirmation of the inquiry.

[0016]

As described above, according to the present invention, the character recognition rate is improved as compared with the conventional apparatus without using the dropout color for the OCR paper, and a large amount of image data can be stored. Is possible.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a configuration of the present invention.

FIG. 2 is an explanatory diagram illustrating Hough transform.

FIG. 3 is an explanatory diagram of creating a difference image between a target image and a template image.

FIG. 4 is an explanatory diagram illustrating a difference in a difference image between a target image and a template image.

[Brief description of reference numerals]

 1 image input device 2 image distortion correction device 3 template storage device 4 image discrimination device 5 character recognition device 6 image compression device 7 digital data storage device 21 black image 22 straight line passing through black dot 23 perpendicular line passing through origin 24 black dot 30 on voucher Target image with text 31 Template image 32 Image with only text image 40 Entry frame 41 Text 42 Text entered

Claims (1)

[Claims] 1. A data storage device for digitally inputting image data, converting character portions in an image into character codes, and storing the image data, an image input device, an image distortion correction device, an image discrimination device, a template image storage device, and a character recognition device. And an image compression device and a digital data storage device, the image input device is used to input a plurality of template images in advance, the image distortion correction device performs tilt correction, and then is registered in the template image storage device. After the inclination correction is performed using the image correction device for the template in which characters are newly written by handwriting or in print using the image input device, the template is stored in the template image storage device using the image discrimination device. Image data consisting only of characters is obtained by distinguishing from a plurality of templates and removing from the newly input template, and character parts are character coded by a character recognition device. Data storage apparatus characterized by storing the digital data storage device the image data and the character code consisting of only the characters that are compressed by the image compression apparatus after.