JPS61117670A - Character cutting-out processing system - Google Patents

Abstract

PURPOSE:To cut out accurately a circumscribed rectangle corresponding to one character by calculating a circumscribed rectangle circumscribing a picture element group and uniting this circumscribed rectangle and another circumscribed rectangle adjacent to it to generate a new circumscribed rectangle if said calculated circumscribed rectangle is smaller than one character. CONSTITUTION:A picture input part 1 scans characters hand-written on a circuit diagram or the like by a scanner and converts them to picture signals and inputs them to a character cutting-out part 2. The character cutting-out part 2 extracts successively picture elements connected to picture elements constituting characters from picture signals inputted from the picture input part 1 and calculates a circumscribed rectangle circumscribing this picture element group to cut out a rectangular area when a character exists. In this case, if one character described on the circuit diagram is cut out as, for example, two circumscribed rectangles because of partial blur or the like, they are united to generate another circumscribed rectangle. Thus, the circumscribed rectangle corresponding to one character is cut out accurately even if one character is cut out as plural circumscribed rectangles.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a character extraction processing method, especially when a character string handwritten on a circuit diagram or the like is extracted character by character into a circumscribed rectangle. The present invention relates to a character extraction processing method configured to extract a circumscribing rectangle that integrates the separated characters even if they are separated due to various reasons.

(Problems to be solved by conventional techniques and inventions) Conventionally,
Characters handwritten on circuit diagrams, etc. are scanned by a scanner and converted into image signals, and then character recognition is performed using a computer system. It is necessary to accurately cut out the eccentric weight position.

However, in order to accurately cut out a handwritten character with an emphasis on 1, we calculate a circumscribed rectangle that circumscribes all the pixel groups that are 0 connected to the pixels of a certain character, and make the circumscribed rectangle correspond to one character. It was divided into areas. For this reason, for example, because the characters are blurred on the circuit diagram, there is a problem in that one character is separated into two circumscribed rectangles and cut out.

(Means for Solving the Problems) In order to solve the above problems, the present invention provides a form that circumscribes all other pixel groups connected to a certain pixel that constitutes a character handwritten on a circuit diagram, etc. , and if the calculated circumscribed rectangle is smaller than the size equivalent to one character, a new circumscribed rectangle that is a composite of the other adjacent circumscribed rectangles is generated, so that exactly 1 A circumscribed rectangle corresponding to a character is cut out.Therefore, in the character cutout processing method of the present invention, a character string is cut out character by character. and a character cutting unit that sequentially extracts pixels connected to pixels constituting a character from the image signal converted by the image input unit and generates a circumscribing rectangle circumscribing a group of pixels constituting the character. , a character width determination unit that determines whether the circumscribed rectangle generated by the character cutting unit has the width of one character separated or a width, and the character width determination unit determines whether the circumscribed rectangle is 1
11II characters do not have separate widths, a left/right/top/bottom character integration unit that detects other circumscribed rectangles adjacent to the circumscribed rectangle and determines whether they should be integrated; - A character extraction unit that generates a new circumscribed rectangle that integrates the circumscribed rectangles found to be integrated by the upper and lower character integration unit.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a configuration diagram of one embodiment of the present invention, FIGS. 2 to 6 are operation explanatory diagrams explaining the operation of the configuration of one embodiment of the present invention shown in FIG. 1 Figure Illustrating the present invention
A flowchart illustrating the operation of the embodiment configuration is shown.

In the figure, l is an image input section, 2 is a character cutting section, 3 is a dust processing section, 4 is a character width discrimination section, 5 is a left, right, top and bottom character integration section,
6 represents a character string division section, and 7 represents a character extraction section.

In FIG. 1, an image input section 1 is used to scan characters handwritten on a circuit diagram or the like using a scanner, convert them into an image signal, and input the converted image signals to a character cutting section 2.

As will be described later, the character extraction section 2 in the figure sequentially extracts other pixels connected to the pixels constituting the character from the image signal input from the image input section 1, and extracts a shape circumscribing the extracted pixel group. This is for cutting out a rectangular area in which characters exist by extracting a so-called circumscribed rectangle. At this time, if one character written on the circuit diagram is partially shifted or cut out in the form of a circumscribed rectangle of 211, for example, two circumscribed rectangles are cut out as described later. The parts cut out in the form are integrated to generate a new circumscribed rectangle. The configuration and operation of the present invention will be explained in detail below.

In the figure, the dust processing unit 3 includes an image input unit l as described below.
When there are minute pixels that do not form a character in the characters converted into an image signal by Xji, the purpose is to delete the minute pixels and remove so-called dust.

In the figure, the character width determination section 4 is connected to the character cutting section 2 as described later.
Width (width in the X-axis direction) of the circumscribed rectangle cut out by
This is for determining whether the vertical width (width in the X-axis direction) is smaller, approximately the same, or larger than the horizontal and vertical widths corresponding to a predetermined character. The circumscribed rectangle cut out by the character cutting unit 2 by performing this discrimination is: (1) A circumscribed rectangle cut out in a form in which a character is separated into a plurality of pieces, a circumscribed rectangle cut out in a form corresponding to a single character,
Alternatively, it is possible to determine whether the cutout corresponds to a plurality of characters (for example, two characters).

The left/right/top/bottom character integration section 5 in the figure is for detecting a pair of circumscribed rectangles cut out by separating one character into a plurality of shapes by the character width discriminating section 4, as will be described later. Horizontal direction of the cut out circumscribed rectangle ('#i direction)
For example, if the width of the specified width is less than half of the specified width, a process is performed to find a pair of circumscribed rectangles to be combined. When it is determined that the width of a new circumscribed rectangle obtained by integrating the circumscribed rectangles is approximately equal to the predetermined width, it is determined that a pair of circumscribed rectangles to be integrated has been detected. As a result, pairs of circumscribed rectangles cut out in separate shapes are detected.

In the figure, the character string dividing unit 6 determines the optimal position for dividing the circumscribed rectangle into each character when the circumscribed rectangle cut out by the character width determination unit 4 is cut out as including a character of ``Zoba'' as described later. This is for calculating l.

The character extraction unit 7 in the figure integrates the pairs of circumscribed rectangles to be integrated detected by the left/right/top/bottom character integration unit 5 into one circumscribed rectangle, or divides them by the character string division unit 6. This is for dividing the circumscribed rectangle into a plurality of circumscribed rectangles at the position calculated as the exponent position. As a result, each circumscribed rectangle corresponding to one character written on the circuit diagram is accurately cut out.

The operation of the configuration shown in FIG. 1 will be explained in detail below using FIGS. 2 to 8.

FIG. 2 is an explanatory diagram illustrating an operation for cutting out a circumscribed rectangle that is integrated when one character "j' is separated in the horizontal direction (x direction).

Figure 2 (a) shows a single character “J” handwritten on the circuit diagram.
As shown in the figure, the character "J' indicating " is partially faded in the left-right direction (x direction),
The image (converted to the form of No. 8 is shown).

FIG. 2(b) shows two circumscribed rectangles (l) and (2) that circumscribe the character "J" shown in FIG. 2(a), cut out. This is done by the middle character cutout section 2 in the figure M1.This is done by the circumscribed rectangle f in the figure.
Since there was another circumscribed rectangle (2) within a distance ゛α” in the right direction (character string direction) of l+, it was cut out to be the object of integration.The distance “α” °
is a predetermined value, and other circumscribed rectangles within the distance “α” are the targets of integration, which will be described later. In addition, the circumscribed rectangles u+, i, and I2) have sizes m, Xn, and m y x mouth □ in the horizontal and vertical directions, respectively.
It is cut out.

FIG. 2(C) shows a result of integrating the circumscribed rectangles il+ and (2) in FIG. 2(B) to generate a circumscribed rectangle (3) corresponding to one character "J". As a result, for example, if a part of one character disappears on the circuit diagram, two
The circumscribed rectangles tl+ and those cut out in the form of (2) are automatically integrated into the form of one circumscribed rectangle (3).

FIG. 3 is an explanatory diagram illustrating the operation when cutting out a number in which a horizontal bar is written above a single number "0".

Figure 3 (a) shows a single number °0 handwritten on the circuit diagram.
An example is shown in which a horizontal bar is written at the top of ゛.

Figure 3 (b) shows two circumscribed rectangles +41 and (5) cut out from the number "O" with a horizontal bar written at the top of the figure in figure 3 (a). .

This is because a circumscribed rectangle (5), which is a circumscribed rectangle of a pond, existed at a distance “β” in the vertical direction of the circumscribed rectangle (4) in the figure, so it was cut out to be an object of integration. The distance "β" is given in advance (direct), and represents the longest distance between the circumscribed rectangles to be integrated.
) and (5) are respectively m in the horizontal and vertical directions,
It is cut out as a size of Xn and m%Xns.

Figure 3 (c) shows the circumscribed rectangles (4) and (5) in Figure 3.
This shows that a circumscribed rectangle (6) corresponding to one number is generated by integrating the 2 numbers.For example, if there are characters such as horizontal bars above and below one character on the circuit diagram, 2 The pieces cut out in the form of the circumscribed rectangles (4) and (5) are automatically integrated into the form of the single circumscribed rectangle (6).

FIGS. 4 and 5 are operation explanatory diagrams illustrating an operation for separating each character when two characters are cut out in the form of one circumscribed rectangle in the horizontal or vertical direction.

FIG. 4(a) shows the letters "AB" written on the circuit diagram that are in contact with each other converted into a self-portrait (number 8).

FIG. 4(b) shows one circumscribed rectangle (7) cut out by the character cutting part 2 in FIG. In the figure, the size of the circumscribed rectangle (7) is expressed in the form 2mtXny.

Figure 4 (c) shows whether the character string division part 6 in Figure 1 or Figure 4 (
b) Figure 0 shows a state in which the position for horizontally moving Ip1 is detected near the center of the mutually touching letters "AB" in the figure and at the position where there are the least number of black pixels in the vertical direction. °α″1 indicates the range to be scanned to detect the area where the minimum pixel exists on the left and right sides from the vicinity of the center.

FIG. 4(d) shows that the character extraction unit 7 in the first causal diagram is the character "AB" that is in contact with each other at the horizontal position where the minimum black pixel exists.
゛ is divided into circumscribed rectangles (8) and (9), respectively. As a result, circumscribed rectangles (8) and (9) are generated for each character “A” and “B”, respectively. Become.

FIG. 5(a) shows characters "FH" touching each other in the vertical direction converted into an image signal.

FIG. 5(b) shows one circumscribed rectangle (II is a box 41) cut out by the character cutout part 2 in FIG. In the figure, the size of the circumscribed rectangle Ql is cut out in the form of m, ox20.

Figure 5 (c) shows whether the character string division part 6 in Figure 1 or Figure 5 (
b) It is located approximately in the center of the letters “Fll” that touch each other in the figure, and is located horizontally as a man’s 1! ! In Figure 9, which shows the state of detecting the position for dividing in the vertical (vertical) direction with the smallest number of j pixels, "β" indicates the state in which scanning is performed upward and downward from near the center to detect the area where the smallest number of pixels exists. Indicates the range.

FIG. 5(D) shows that the character extraction unit 7 in FIG. “F” and “
This shows a state in which each character is divided into circumscribed rectangles (11) and (12) corresponding to ``H''.
11) and (12) will be generated.

FIG. 6 shows an operation explanatory diagram for explaining the operation of dust removal processing.

FIG. 6(a) shows an image signal when there are fine dust-like black pixels in the vicinity of the letter "A".

FIG. 6(b) shows that the character cutout portion 2 in FIG. 1 is a circumscribed rectangle (1
3) and a minute circumscribed rectangle (14) are shown to show the state of cutting out, in this case, the minute circumscribed rectangle (14) is smaller than the pixels constituting the smallest character included in the character to be cut out, for example. In this case, the size of the circumscribed rectangles (13) and (14) in the 0 diagram that is cut out is m12
x n , and mla×n+a.

FIG. 6(C) shows that the character string division part 6 in FIG.
b) Indicates a circumscribing rectangle (15) obtained by removing the dust-like black pixels from the character "A" in which the dust-like black pixels exist in the figure. Removal processing will be executed.

Next, the operation of the configuration shown in FIG. 1 will be explained in more detail using FIG. 712I and FIG.

■ in the diagram is black! 1! This shows the state of discovering the i-element, which is
This means performing a scan or the like to discover the black pixels that make up the characters as shown in FIG. 2(a).

■ in the figure indicates a state in which fill tracking is enabled.

This means that other black pixels connected to any black pixel of the interrupted letter "J" in FIG. 2(a) are sequentially traced.

In the figure, ■ indicates a state in which the width (m, mouth) of a circumscribed rectangle is calculated. This means calculating the width (m)8 in the x direction and the vertical width (width n in the Y direction).

■ in the figure indicates "0" in the register number for the first time! This shows the state set as the lII value.

In the figure, ■ indicates that the horizontal width (m) of the circumscribed rectangle obtained by tracking in state ■ is less than half the radius (M) corresponding to a predetermined character (m<M/2 in the figure), and is approximately equal. ([iJ pH:
m-1M) or 1.5 times or more (1 small m>
M” M / 2), and depending on each state, state ■, state @
Or execute state 0 or lower.

■ in the figure indicates the horizontal width (m) of the circumscribed rectangle obtained by tracking.
) is less than half, it is executed, and indicates a state to determine whether the content stored in the register No. is "1". If YES, execute the state @. If NO Execute state ■.

In the figure, ■ indicates a state in which another pixel existing within a position a distance “α” away from the circumscribed rectangle in the X direction is searched. This means that the width in the X direction of the circumscribed rectangle determined in state ■ is specified. Since the width (M) is less than half of the width (M) of
This means searching for other pixels existing within a position α″ away.

In the figure, ■ indicates a state in which it is determined whether there are other circumscribed rectangles within a position separated by a distance of α° in state ■.If YES, all other circumscribed rectangles connected to the black pixel in state ■ are determined. After tracing the pixels of , and finding a circumscribed rectangle that includes all the pixels, the state Li@ is executed.If the answer is No, the state 0 is executed.

In the figure, [phase] indicates a state in which the width (m, n) of the circumscribed rectangle is corrected. This is corrected to the shape of a new circumscribed rectangle that circumscribes the circumscribed rectangle obtained in state ■ and the original circumscribed rectangle. It means to do.

In the figure, O indicates a state in which °l" is set in register NO. For example, in state ■, if no other pixel is found within a distance "α" from the circumscribed rectangle, or if another pixel is If a new circumscribed rectangle is found, but the new circumscribed rectangle that is a combination of the found circumscribed rectangle and the original circumscribed rectangle is still less than half of the specified width (M), for example, in the case of a circumscribed rectangle such as garbage, repeat the process. This is to prevent the state ■ or [stock] from being executed.

◎ in the figure indicates that the width (m) of the circumscribed rectangle currently being processed is almost equal to the specified width (M) in state ■, so the width (m) of the circumscribed rectangle is determined and registered. After completing the liquid state @, which indicates the state in which the liquid is present, "[Stock]" in FIG. 8 is executed as a process for the vertical direction.

O in the figure indicates that the width (m) of the circumscribed rectangle currently being processed in state ■ is found to be 1°5 times or more compared to the specified width (M), so the width of the circumscribed rectangle is (m>) in the horizontal direction. The division process is performed in the fourth
As already described using the figure (Fig. 4 (c)), the frequency of black pixels in the vertical direction becomes minimum within the range of distance "α" in the left and right directions at approximately the center (center) position of the illustrated circumscribed rectangle. This is done by calculating the position and dividing into circumscribed rectangles at the calculated positions.

In FIG. 8, 0 indicates a condition for determining whether the vertical width (n) is 1.5 times or more the specified width (N), 15
If it is less than double, execute phase B [phase]. If it is 1.5 times or more, execute Bo.

0) in the figure means that the vertical width (n) of the circumscribed rectangle is the specified width (
Since it is 1.5 times or more of N), a state in which vertical division υ1 processing is executed as explained using FIG. 5 is shown.

In the figure, [phase] means that the vertical width (n) of the circumscribed rectangle is less than 1.5 (ia) of the specified width (N), so the distance ° β " If there are other black pixels within a distant position, it indicates a state in which the width of the circumscribed rectangle in the Y direction is calculated by tracking other black pixels connected to the black pixel, and the width of the circumscribed rectangle in the Y direction is calculated. A new circumscribed rectangle that circumscribes the original circumscribed rectangle and the original circumscribed rectangle is calculated, and if the width of the calculated circumscribed rectangle in the Y direction is within the specified width, the Y of the new circumscribed rectangle is calculated in state O. The width (n) in the direction is determined and registered.

As mentioned above, one character may be blurred, etc., and the
Even if the character is cut out in the form of a plurality of circumscribed rectangles separated in the direction and the Y direction, the circumscribed rectangle will be modified to correspond to one original character.

〔Effect of the invention〕

As explained above, according to the present invention, RF! Calculate a circumscribed rectangle that circumscribes all other pixel groups connected to the pixels constituting a character handwritten on a J road map, etc., and make sure that the calculated circumscribed rectangle has a size of 4 degrees corresponding to one character. If Di is small, a new circumscribed rectangle is generated by combining other neighboring circumscribed rectangles, so even if a part of the character is blurred or the like, it is cut out in the form of several circumscribed rectangles. It is also possible to cut out a circumscribed rectangle that exactly corresponds to one character.

[Brief explanation of the drawing]

FIG. 1 shows the configuration of an embodiment of the present invention, FIGS. 1 is a flowchart illustrating the operation of one embodiment of the present invention; FIG. In the figure, l is an image input section, 2 is a character cutting section, 3 is a garbage processing section, 4 is a character width discrimination section, 5 is a left/right/top/bottom character integration section, 6 is a character string division section, and 7 is a character extraction section. represent.

Claims (1)

[Claims] A character extraction processing method for cutting out a character string into character units includes: an image input unit that converts the character string into an image signal; a character extraction section that sequentially extracts pixels contained in the character and generates a circumscribing rectangle that circumscribes a group of pixels constituting the character; and a circumscribing rectangle generated by the character extraction section has a width equivalent to separating one character. a character width determination unit that determines whether the circumscribed rectangle is adjacent to the circumscribed rectangle; A new circumscription that integrates a left/right/top/bottom character integration unit that detects other circumscribed rectangles and determines whether they should be integrated, and a pair of circumscribed rectangles that are found to be integrated by the left/right/top/bottom character integration unit. A character extraction processing method comprising: a character extraction unit that generates a rectangle.