JP3124854B2 - Character string direction detector - 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 string direction detecting device, and more particularly, to an optical character recognition device for reading a printed sheet, particularly a business card, for detecting whether the direction of a character string on the sheet is horizontal or vertical. And a character string direction detecting device.

[0002]

2. Description of the Related Art There are various formats on printing paper, especially on business card paper. The character strings include those in the horizontal direction and those in the vertical direction. Although the information on the character string direction is important when the paper is processed by an optical character recognition device, there is no known technology for automatically and stably determining the information.

[0003]

That is, it has not been possible to automatically determine whether the direction of a character string on the paper surface is horizontal (horizontal) or vertical (vertical).

An object of the present invention is to provide a character string direction detecting device capable of automatically detecting the direction of a character string on paper.

[0005]

[MEANS FOR SOLVING THE PROBLEMS] A plurality of sentences forming an image
Means for extracting character elements, and all elements of extracted character elements
Label with different values for all combinations between character elements
An equivalence relationship based on the proximity
Character element that checks whether or not
Means for relabeling the same value,
Means for generating area rectangle from bell-attached character elements
And the direction of the character string included in the image is changed to the area rectangle.
Means for making a determination based on

[0006]

[Function] Extracting and extracting a plurality of character elements forming an image
Label all elements of the specified character element with different values, and
Proximity relation between character elements for all combinations between character elements
Checks whether an equivalence relation based on the
Re-label the same value for the text elements for which the relationship holds
From the character element labeled with the same value
Generate a rectangle and change the direction of the character string included in the image to the
Since the judgment is made based on the area rectangle, the direction of the character string can be automatically known by inputting the image, and the image processing that matches the direction of the character string by rotating the image 90 degrees as necessary Can be performed, the character element can be recognized, and the user can use the optical character recognition device without being aware of the character string direction on the paper.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS According to the present invention, the direction of a character string in an image is horizontal (x direction) or vertical (y direction) based on the arrangement of characters or character components (hereinafter referred to as character elements) in the image.
It is a device that detects what it is.

First, a method of detecting a character string direction according to an embodiment of the present invention will be described.

FIG. 6 is a diagram showing an example of a business card in which the character string is in the short side direction and has extremely large character elements.

At first, an extremely large character element which may adversely affect the detection, for example, " ○○○○○ " in FIG.
○ NEWS ", or small character elements, for example, a scanner noise, set in advance a reference for detecting the sheet surface of the dust, exclude such character elements from the detection target.

In FIG. 6, unless a character element such as “ XXXXXX NEWS” is ignored, the entire business card becomes one area. If the "○○○○○ NEWS" to process the same as the other character elements of the character elements instead of ignoring such as, because of its size, "○ KawaTaro", "○○○○ Co., Abeno-ku, Osaka XX town XX number XX (〒545) "
And the other character elements are in the same area, and as a result, the entire area is the same, and a desired result cannot be obtained. If you apply the criterion of ignoring text elements of a certain size or more, the processing will be " ○ Kawataro",
"○○○○ Co., Abeno-ku, Osaka ○○-cho, ○○ number ○○
The area of the character string is determined to be horizontal because both areas are long in the horizontal direction. By providing such a criterion, the performance can be further improved from a practical point of view.

FIG. 7 is a diagram for explaining the proximity relationship between the character elements of two characters. With reference to FIG. 7, a criterion for determining whether any two character elements in an image are "close" or "not close" will be described.

In FIG. 7, circumscribed rectangle of character element "c" is rc, circumscribed rectangle of character element "d" is rd, and r
The distance in the x direction between c and rd is lx (= F− (D + E)), y
If the distance in the direction is ly (= C- (A + B)), lx <ma
x {(length of rc in x direction, D), (length of rd in x direction, E), (length of rc in y direction, A), (length of rd in y direction, B)} And ly <max ｛(length of rc in the x direction, D), (length of rd in the x direction, E), (rc
Length of the y-direction, A), (length of rd in the y-direction, B)}
, C and d are close to each other, otherwise it can be determined that they are not close to each other (c and d are not close to each other in FIG. 7). If the proximity of two character elements a and b is described as a to b, the two character elements c and d are described as "a sequence {xi} (i
= 0,1, ... n-1) and c to x0 to x1 to. . . xn-1
~ D ”(this is described as c‖d)
Is a kind of relation mathematically called an equivalence relation.
The equivalence relation has the remarkable property that the entire set in which it is defined is decomposed into several groups consisting only of those that hold the relation.

Therefore, if it is checked whether or not the above-mentioned equivalence relation holds for all combinations of two character elements in the image, all the character elements in the image can be changed to any character belonging to the group. A‖b for elements a and b
(Hereinafter referred to as a region).

This can be performed, for example, as follows.
Prepare character element labels that correspond one-to-one with character elements,
All the elements (hereinafter, labels) are initialized to different values. Then, the two character elements in the image are checked to see if they are close to each other in accordance with the above-described criteria. If it is determined that the two are close to each other, a common value (hereinafter referred to as V0) is assigned to the label corresponding to the two character elements and all the labels having the same value as one of the two labels. I do. V0 may be the previous value of the label corresponding to one of these two character elements. When this is repeated for all combinations of character elements, a set of character elements having a common label value is defined as one area.

The region physically corresponds to a group of character elements obtained when character elements that are locally dense on the paper are integrated into one. For each region, a minimum rectangle including all the character elements included therein is created (hereinafter, this is referred to as a region rectangle).

For example, character elements having a common label value are collected, and the minimum value xo of the starting point in the x direction of the circumscribed rectangle, the minimum value yo of the starting point in the y direction, the maximum value x1 of the ending point in the x direction, By taking out the maximum value y1 of the end point in the y direction, it is easy to make a rectangle having a straight line (xo, yo)-(x1, y1) as a diagonal line.

A criterion for determining whether a character string in the image is in the x-direction or in the y-direction, using the position coordinates of the area rectangle as input, is determined in advance. For example, for each region rectangle, R
xy = (length in y-direction) / (length in x-direction) is calculated in advance, and RA = (number of region rectangles in which Rxy is less than 1.0) /
When (the total number of areas in the image) is less than 0.5, the character string is regarded as the y direction on the paper, and otherwise, the character string is regarded as the x direction.

In addition, by adding a criterion for ignoring an extremely small area rectangle considered to be caused by noise or an extremely elongated area rectangle thought to be caused by a design, the reliability of determination can be further increased. By inspecting each area rectangle based on such a criterion, it is possible to obtain information in the character string direction, which is the final purpose.

FIGS. 3 and 4 show examples of business cards in which the character string extends in the long side direction, and FIG. 5 shows examples of business cards in which the character string extends in the short side direction. FIGS. 8, 9 and 10 are diagrams showing a circumscribed rectangle of a character element when a connected portion of black pixels in the image is a character element. FIGS. 11, 12, and 13 are diagrams illustrating an area rectangle in the case where the criterion for determining whether or not two character elements c and d are close to each other is the same as in the above example.

A specific example will be described based on the above criteria.

3 to 5 and 8 to 13, let the long side direction of the business card be the y direction. According to the above-described character string direction determination criterion, in FIG. 11 (corresponding to FIG. 3 and FIG. 8), all three rectangular areas have an aspect ratio of 1.0 or more, so RA becomes 0,
The character string direction is determined to be the y direction. FIG. 12 (FIGS. 4, 9)
The same applies to the above. However, in FIG. 13 (corresponding to FIG. 5 and FIG. 10), all three rectangles in the entire area have an aspect ratio of less than 1.0, so RA is 1.0, and the character string direction is determined to be the x direction. As shown in FIGS. 3, 4, and 5, correct results were obtained in all cases.

FIG. 1 is a block diagram showing a case where the embodiment of the present invention is incorporated in an optical character recognition apparatus. The configuration of the embodiment of the present invention is shown in a dotted line.

In FIG. 1, an image captured from a scanner 1 is stored in an image buffer 2. Image buffer 2
The character elements are extracted by the character element extraction unit 3 constituting extraction means together with a character element buffer 4, a character element exclusion unit 5, and a character element exclusion criterion unit 6 described later, and stored in the character element buffer 4. The character element exclusion unit 5 refers to the character element exclusion criterion unit 6 and removes a character element inconvenient for determination from the character element buffer 4. The character element proximity relation determination unit 7 that forms a determination unit together with a character element label buffer 8 described later first initializes all elements of the character element label buffer 8 to different values, and then sets the two characters for which the above-described equivalence relation holds. All combinations of elements are searched from the character element buffer 4, and the corresponding elements and
Performs an operation to assign the same value to an element having the same value as one of the two. The area rectangle generation unit 9 refers to the character element label buffer 8 to generate an area rectangle circumscribing all character elements having common values, and sends the generated area rectangle to the area rectangle buffer 10 constituting a generation unit together with the area rectangle generation unit 9. Store. The character string direction judging unit 11 refers to the character string direction judging unit 12 constituting the column direction judging unit together with the area rectangular buffer 10 and the character string direction judging unit 11 so that the character string on the sheet is in the x direction and the y direction. Determine which one is facing.

The character recognition unit 13 extracts and recognizes characters from the image buffer 2 while referring to the result of the character string direction determination unit 11 and the recognition dictionary 13, and stores the result in the recognition result buffer 15. The language processing unit 16 refers to the language dictionary 17 and changes the recognition result in the recognition result buffer 15 as needed. The result correction unit 18 recognizes the result buffer 1
5 is output to the display 19, and the user is required to input necessary changes from the keyboard 20. Result output unit 21
Outputs the contents of the recognition result buffer 15 from the printer 22.

FIG. 2 is a flowchart for explaining the operation of the embodiment of the present invention.

In FIG. 2, inconvenient characters are excluded from character elements according to a predetermined standard (2-1).
. Next, two character elements (Ci, Cj (corresponding to c and d described above), but i and j are not equal) are extracted once (2-2). It is determined whether or not the extracted Ci and Cj are close to each other (2-3), and if they are close to each other, the label corresponding to Ci and Cj has a value equal to one of the labels. Substitute a new value common to all labels (2-4). Ci
When Cj and Cj are not close to each other and when a common new value is substituted, all combinations (all combinations of two different character elements in the arbitrarily extracted image, that is, the character element If there are _{n, the} number of combinations is _n C ₂ ) If not processed, return to step 2-2 to generate an area rectangle if all combinations are processed (2-6), The character string direction is determined according to a predetermined standard (2-7).

[0028]

According to the present invention, a plurality of character elements forming an image are extracted.
And assign different value labels to all the extracted character elements.
For all combinations between character elements, between character elements
Whether equivalence relations based on proximity relations hold
And the label of the same value is assigned to the character element for which the equivalence relation holds.
From the character element labeled with the same value
Generates a region rectangle and the direction of the character string included in the image
Is determined based on the region rectangle, so that the direction of the character string can be automatically obtained by inputting an image,
By rotating the image 90 degrees as needed, image processing can be performed according to the direction of the character string, character elements can be recognized, and the user can use the optical character recognition device without being aware of the character string direction on the paper. Can be used.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a case where an embodiment of the present invention is incorporated in an optical character recognition device.

FIG. 2 is a flowchart for explaining the operation of the embodiment of the present invention.
It is.

FIG. 3 is a diagram illustrating an example of a business card in which a character string is in a long side direction.

FIG. 4 is a diagram showing another example of a business card in which a character string is in a long side direction.

FIG. 5 is a diagram illustrating an example of a business card in which a character string is in a short side direction.

FIG. 6 is a diagram illustrating an example of a business card in which a character string is in a short side direction and has extremely large character elements.

FIG. 7 is a diagram illustrating a proximity relationship between character elements of two characters.

FIG. 8 is a diagram illustrating a circumscribed rectangle of a character element corresponding to FIG. 3 when a connected portion of black pixels in the image is a character element;

FIG. 9 is a diagram illustrating a circumscribed rectangle of a character element corresponding to FIG. 4 when a connected portion of black pixels in the image is a character element;

FIG. 10 is a diagram showing a circumscribed rectangle of a character element corresponding to FIG. 5 in a case where a connected portion of black pixels in an image is a character element;

FIG. 11 is a diagram showing an area rectangle corresponding to FIGS. 3 and 6;

FIG. 12 is a diagram showing an area rectangle corresponding to FIGS. 4 and 9;

FIG. 13 is a diagram showing an area rectangle corresponding to FIGS. 5 and 10;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Scanner 2 Image buffer 3 Character element extraction part 4 Character element buffer 5 Character element exclusion part 6 Character element exclusion criterion part 7 Character element proximity relation judgment part 8 Character element label buffer 9 Area rectangle generation part 10 Area rectangle buffer 11 Character string direction Judgment unit 12 Character string direction judgment reference unit 13 Character recognition unit 14 Recognition dictionary 15 Recognition result buffer 16 Language processing unit 17 Language dictionary 18 Result correction unit 19 Display 20 Keyboard 21 Result output unit 22 Printer

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Minako Kuwata, 22-22 Nagaikecho, Abeno-ku, Osaka-shi, Osaka Prefecture Inside Sharp Corporation (72) Inventor Kazuhiro Takehara 22-22 Nagaikecho, Abeno-ku, Osaka-shi, Osaka Sharp Corporation (72) Inventor Yoshihiro Kitamura 22-22 Nagaike-cho, Abeno-ku, Osaka-shi, Osaka Inside Sharp Corporation (56) References JP-A-2-311975 (JP, A) JP-A-63-26784 (JP, A) JP-A-3-139783 (JP, A) JP-A-3-198177 (JP, A) JP-A-62-166479 (JP, A) JP-A-5-81474 (JP, A)

Claims (2)

(57) [Claims] 1. Extraction of a plurality of character elements forming an image
Means that all elements of the extracted character element have different values.
Add a bell and use characters for all combinations between
Whether equivalence relation based on proximity relation between elements holds
Are checked, and character elements for which an equivalence
Means to relabel and label with the same value as above
Means for generating an area rectangle from character elements;
Determine the direction of the included character string based on the area rectangle
A character string direction detecting device. 2. The character string direction detecting device according to claim 1, wherein the image is an image of a business card paper .