JP2503208B2 - Business card image processing method - Google Patents

Description

Description: TECHNICAL FIELD The present invention is to distinguish between vertical and horizontal business cards and how many times the business cards are rotated (0 °, 90 °, 180 °) in a business card OCR.
Business card image processing method for automatically determining (°, 270 ° rotation, etc.).

[Prior art]

Business cards come in two formats: vertical writing and horizontal writing. Conventionally, in order to recognize such a business card with an optical character reading device (OCR), an operator needs to set the business card one by one in a predetermined direction according to the OCR, which is not necessarily easy to use.

〔Purpose〕

An object of the present invention is to read an OCR for reading a business card,
Vertical and horizontal business cards can be distinguished and the number of times the business card is rotated (0 °, 90 °, 180 °, 270 ° rotation, etc.) is automatically determined to improve the ease of use for this type of OCR. Is to try.

〔Constitution〕

A telephone number is always printed on the business card, but as shown in FIG. 3 (a) and FIG. 4 (a), the telephone number is printed at the end along with the address in any vertically or horizontally written business card. And the size of the letters used in it is the smallest of any business card. Therefore, as shown in FIG. 3 (b) and FIG. 4 (b), the projection is made along the longitudinal direction of the business card, and the side with the few narrowest shadows is the portion where the telephone number exists. It can be inferred that there is.

The present invention pays attention to this point, detects the longitudinal direction of the business card by projecting the entire horizontal and vertical directions of the input pattern,
Cut out characters including the phone number. Then, the cut-out portion is recognized by using a dictionary corresponding to each number rotated by 0 °, 90 °, 180 °, 270 °, and the number with the highest similarity is continuously output to rotate the number. The angle is judged to distinguish between vertical writing and horizontal writing of the business card and whether it is set by rotating 180 degrees.

FIG. 1 shows a schematic structure of an OCR for business cards which is the object of the present invention. In FIG. 1, 10 is an input scanner, 20 is a microprocessor, 30 is a bitmap memory, and 40 is a dictionary memory. The microprocessor 20 is a processing unit (CPU),
It consists of program memory (ROM), data memory (RAM), etc. Projection detection, feature extraction, and
Execute processing such as dictionary search. Since the microprocessor itself is well known, details of its configuration are omitted. The dictionary memory 40 has a dictionary corresponding to numbers rotated by 0 °, 90 °, 180 ° and 270 °.

For the business cards that have been set arbitrarily, input scanner 10
Reads the entire image of the business card and writes it in the bitmap memory 30. The microprocessor 20 performs the following processing on the image of the bitmap memory 30.

First, for the input image of the bitmap memory 30, the entire projection is taken in the horizontal and vertical directions, and the second
As shown in the figure, the length of X and Y is judged to detect the longitudinal direction of the business card.

Next, of the detected projections in the longitudinal direction, the side where there are several projections with the smallest width is examined. Then, for the projection with the smallest width and at the end, a projection for one line is taken out in the direction perpendicular to the longitudinal direction. This is so-called character cutout, and the part of the business card including the telephone number is cut out.

Next, after extracting the features of the cut-out characters (numbers), the similarity between the feature parameter and the number dictionary rotated in the dictionary memory 40 by 0 °, 90 °, 180 °, 270 ° is prepared in advance. Calculate (distance). For each character that is cut out, it is determined how many times the number dictionary is rotated and the best match is made, and the number of times the business card is rotated and placed is detected.

Next, character cutting is performed on all projections of the input image in the bitmap memory 30 in the longitudinal direction, and the image for each character cut out is rotated according to the detected rotation angle, and then recognition processing is performed. Instead of rotating the image for each character and performing the recognition process, the dictionary memory 40
It is also possible to have all the dictionaries corresponding to the characters rotated by °, 90 °, 180 °, 270 ° and search only the dictionary corresponding to the detected rotation angle.

[Effect]

As is apparent from the above description, according to the present invention, when placing a business card on the OCR, it is not necessary to be aware of the direction, and the ease of use of the business card OCR is improved.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a business card reading device targeted by the present invention, FIG. 2 is a diagram showing horizontal and vertical projections of contents of a bitmap memory, and FIGS. 3 and 4 are principles of the present invention. It is a figure explaining. 10 ... Input scanner, 20 ... Microprocessor, 30 ...
… Bitmap memory, 40… Dictionary memory.

Claims (1)

(57) [Claims] 1. A vertical and horizontal business card image processing method, comprising reading a vertical or horizontal business card, and taking the entire projection of the read business card image in the horizontal and vertical directions to determine the longitudinal direction of the business card. Each character pattern is detected by using the narrowest projection in the detected projection in the longitudinal direction as the projection of the telephone number portion, and the cut-out character patterns and 0 °, 90 °, 180 ° prepared in advance, 270
A business card image processing method characterized by comparing a pattern corresponding to a rotated character to distinguish between vertical writing and horizontal writing of a business card and how many times the business card is rotated.