JPS63251874A - Segmenting method for contact character - Google Patents

Abstract

PURPOSE:To read even the characters touching with a nondrop-out color entry frame with high reliability by separating the characters touching with said entry frame from this frame without missing parts of characters included in a entry frame picture. CONSTITUTION:It is supposed that character picture is not included in an entry frame picture and the character picture is cut by the 1st boundaries B1-B4 set inside of the frame picture. These separated character pictures are taken out. In case the cut points P10 and P11 are produced with cuts at the boundaries B1-B4, it is supposed that a character picture is included between both points P10 and P11. Then the character picture is cut by the 2nd boundaries B1'-B4' including the outer part of the contour of the entry frame picture and extracted through the points P10 and P11. Then the types of characters are read from the character pictures segmented at the boundaries B1-B4 and those segmented at the boundaries B1', B4'. Thus it is possible to read even the characters touching with a nondrop-out color entry frame with high reliability.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a method for cutting out touching characters, and in particular, a method for cutting out touching characters suitable for reading touching characters written in a non-dropout color writing frame. Regarding.

[Prior Art] Most conventional character reading methods read characters within a writing frame printed in a dropout color.

However, in recent years, copying machines have become easier to create forms.

There is a need for technology that can read forms with black print boxes output by laser beam printers.

In this type of character reading technology, it is necessary not only to automatically distinguish between two characters and a writing frame, but also to cut out characters in order to separate the characters that have touched the writing frame from the writing frame and decipher them. This type of conventional technology includes JP-A-60
There is a technique described in Japanese Patent No.-126777.

[Problems to be Solved by the Invention] In the prior art described above, characters that are in contact with a writing frame are separated by cutting the writing frame and the characters at the contact portion.

Therefore, there is a problem in that there are many cases where a part of the characters existing in the entry frame image are missing, separated as normal characters, and cannot be read.

SUMMARY OF THE INVENTION An object of the present invention is to solve the problems of the prior art, to separate characters that are in contact with the frame from the frame without missing any part of the characters in the frame image, and to use non-dropout coloring. To provide a method for cutting out touching characters that can be read with high reliability even if the characters are touching a writing frame.

[Means for solving interlayer points] The purpose is to assume that the character image is not included in the frame image, and to cut the character image at the first boundary set inside the outline of the frame image. When a plurality of cutting points are generated by taking out and cutting at the first boundary,
Assuming that a character image is included between the plurality of cutting points, the character image is extracted by cutting from the cutting point at a second boundary including a portion outside the outline of the fill-in frame image, and the character image is extracted from the first. This is achieved by selecting which of the second boundaries the characters can be read from the separated character images.

[Operation] In the present invention, when separating the entry frame and characters displayed on a form, it is assumed that the character image is not included in the entry frame image, and the first boundary set inside the entry frame image is used to separate the entry frame and characters displayed on the form. Cut it and extract the character image.

Next, when a plurality of cutting points are generated by cutting at the first boundary, it is assumed that a character image is included between the plurality of cutting points. Based on this assumption, the character image is cut and extracted from the cutting point at a second boundary that includes a portion outside the outline of the fill-in frame image.

Then, the character image cut out at the first boundary and the second
What kind of character is read based on the character image cut out at the boundary.

This allows characters that touch the frame to be separated from the frame, and characters that touch the non-dropout color printing frame to be separated from the frame without missing any part of the characters within the frame image. However, it can be read with high reliability.

[Example] Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a flowchart showing an embodiment of the method of the present invention;
FIGS. 2(a) to 2(j) are diagrams showing specific examples of character extraction, and FIG. 3 is a block diagram showing an example of an apparatus for carrying out the method of the present invention.

The apparatus of the embodiment shown in FIG. 3 includes a form conveyance mechanism 1 for conveying a form 2, a light emitting source 3, a light receiving element 4, a binarization circuit 5, a storage section 6, and a cutting/recognition section 7. It is equipped with

In the apparatus shown in FIG. 3, the form 2 is transferred to the light emitting source 3 by the form transport mechanism 1. The light receiving element 4 is transported to a reading section.

The light emitting source 3 irradiates the form 2 with light, the light is reflected on the form 2, and enters the light receiving element 4.

This light receiving element 4 converts the sequentially incident patterns into electrical signals and outputs them. In this light receiving element 4.

A -dimensional semiconductor CCD sensor or the like is used.

The binarization circuit 5 outputs the output signal from the light receiving element 4 to the storage unit 6 as a binary signal of "'O" 1 (l 11) according to a predetermined threshold value.

The storage unit 6 stores the pattern on the form 2 as a binary image of 1 (Q 11゜゛1”).
A semiconductor memory or the like is used for this purpose.

The extraction/recognition unit 7 is composed of a microprocessor. The cutout/recognition unit 7 cuts out a character pattern character by character from the binary image in the storage unit 6, compares the cut out character image with a standard pattern, and outputs the result as a recognition result.

Next, a method for cutting out touching characters by the cutting/recognition section 7 will be specifically explained with reference to FIG. 1 and FIGS. 2(a) to (j).

When a character pattern including a writing frame as shown in FIG. 2(a) is input, first, in order to separate the writing frame and the character pattern, in step 100 shown in FIG. Extract as C2 and C3.

Next, in step 200 shown in FIG.
Classify I, C2, and C3 into entry frame contours and character axis contours,
Select the outline of the entry frame. Various methods can be considered for classification, but in this embodiment, classification is performed based on whether the size of the outline circumscribed rectangle is larger or smaller than a specified value. Figure 2 (
In the example b), C2 and C3 are selected as the entry frames.

At this time, the character outline touching the entry frame outline is included within the entry frame outline, as shown in FIG. 2(c).

Note that the character contour C1 that is not in contact with the entry frame contours C2 and C3 is extracted and used in combination with other character contours during character recognition.

Next, in step 300 shown in FIG. 1, in order to separate the character outline within the entry frame outline, a boundary between the character outline and the entry frame outline is set. This boundary setting is shown in FIG. 2(d). As shown, first, assuming that a part of the character outline is not included within the entry frame line segment, first boundaries B1 to B1 are placed inside the entry frame outline.
Set B4. Here, inside the rectangle formed by the first boundaries 81 to B4, the entry frame contours C2 and C3,
The point where the first boundaries 81 to B4 intersect is the cutting point pt
o, pH. The cutting direction at the cutting point PIO, pH is the same direction as the boundary Bl.

When the number of cutting points corresponding to each of the first boundaries 81 to B4 is plural, it is assumed that a part of the character axis is included between the cutting points. Then, based on this assumption, a second boundary including a portion outside the outline of the filling frame is set from the plurality of cutting points. In the case of this example, as shown in FIG. 2(e), it is assumed that a part of the character outline is included between the cut points P1o and P, and the cut points pto, perpendicular to the pH. A second boundary B't that includes a part of the entry frame contour C2
~ Set B4. Note that if the character outline is not in contact with the writing frame outline, there is no problem because the cutting point will not be detected even if this process is executed.

Next, in step 400 shown in FIG.
~B4 and second boundaries B'1 to B4 separate the character contour from the writing frame contour. The first boundary B1-84 separates the character axes C4 and Cs in FIG. 2(f), and the second
The character contour C6 in Fig. 2(g) is determined by the boundaries B'1 to B4.
are separated.

Next, in step 500 shown in FIG. 1, character recognition is performed. Character recognition is performed using a combination pattern of character outline C1 extracted in advance and character axis outlines C4, C5 and C6 separated from entry frame outlines C2 and C3.

That is, in the case of character contours separated by the first boundaries 81 to B4, as shown in FIG. 2(h), the character axis contour C1°C4
, C5, and this character axis Ct.

Character recognition is performed using the combination character pattern of C4 and C5, and the recognition result (unreadable) is output. On the other hand, in the case of character contours separated by the second boundaries B'r to B4, one character contour consists of a combination of C+ and Ca, as shown in FIG. 2(i), and this character contour C1°Cs Character recognition is performed using the combination of character patterns, and the recognition result "5" is output.

After character recognition is completed, in step 600, the above-mentioned FIG. 2 (h
), character contours separated by the most appropriate boundaries are selected from the recognition results shown in (i). In other words, in this embodiment, the character axes separated by the second boundaries B'l to B4, which are the correct reading, are selected, and as shown in FIG. 2(j), the correct reading character [5
”. If there is no boundary worthy of correct reading, or if there are multiple boundaries worthy of correct reading, the character is treated as an illegible character.

In this embodiment, the boundary is selected based on whether the recognition result is correct reading or incorrect reading, but it goes without saying that the determination may be made based on other recognition result information such as similarity.

In addition, the processing time can be shortened by allowing the 1-separation and recognition process to be interrupted as soon as it becomes clear that there is no need to perform separation recognition on the boundary remaining from the previous recognition result. .

In this way, according to this embodiment, even if a part of the character to be written comes into contact with the writing frame and a part of the character outline is buried within the outline of the writing frame, the writing frame can be printed in non-dropout color. Even if there is, the first. The character outline is separated and recognized from the entry frame outline by the second boundary, and based on the recognition result information, the first border is used.
It is possible to accurately determine at which of the second boundaries it is possible to read the character with the separated character contour.

Note that the present invention is not limited to the method of separating a character image from a fill-in frame image based on the fill-in frame contour and character contour; Any method may be used as long as the character image is separated from the entry frame image at the second boundary and it is possible to select at which boundary the character can be read from the separated character image.

[Effects of the Invention] According to the present invention described above, it is assumed that the character image is not included in the entry frame image, and the character image is cut and extracted at the first boundary set inside the outline of the entry frame image, and the When a plurality of cutting points are generated by cutting at the first boundary, assuming that a character image is included between the plurality of cutting points, the part outside the outline of the frame image is cut from the cutting point. Since the character image is cut and taken out at the second border containing the text, and the character image can be read from the separated character image by either the first or second border, the text can be read from the separated character image. You can separate characters that are touching the entry frame from the entry frame without missing any part of the character,
Therefore, even characters touching the non-dropout color entry frame can be read with high reliability.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing an embodiment of the method of the present invention;
FIGS. 2(a) to 2(j) are diagrams showing specific examples of character extraction, and FIG. 3 is a block diagram showing an example of an apparatus for carrying out the method of the present invention. DESCRIPTION OF SYMBOLS 1... Form transport mechanism, 2... Form, 3... Light emitting source, 4... Light receiving element, 5... Binarization circuit, 6... Storage unit, 7... Cutting out/ Recognition unit, C1, C2, C3・
...Entry frame and character outline, B+-84...first boundary, B'l~B4...second boundary, PIO, pH...
- Cutting point, Ct. C4, C5, CG...Separated character outlines. Agent Patent Attorney Tadashi Akimoto Figure 1 Figure 2

Claims (1)

[Claims] 1. In a contact character extraction method that separates and reads the entry frame image and character image from a binary image that includes the entry frame and characters displayed on a form, it is possible that the character image is not included in the entry frame image. Assuming that a character image is extracted by cutting at a first boundary set inside the contour of the entry frame image, and a plurality of cutting points are generated by cutting at the first boundary, the plurality of cutting points Assuming that a character image is included between the points, the character image is extracted by cutting from the cutting point at a second boundary that includes the outer part of the outline of the fill-in frame image, and the character image is extracted from the first and second boundaries. A contact character extraction method characterized by selecting at which boundary a character image separated by which a character can be read.