JPH0192795A - Character pattern automatic formatting method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] (Summary) Regarding an automatic character pattern formatting processing method for formatting input character patterns, the purpose of this method is to automatically format character patterns to generate beautiful character patterns. Background noise is removed from the binary image, straight edge noise is removed using black runs, vector shaping processing is performed to remove noise at bending points and end points of contour vectors, etc., and character patterns are automatically shaped.

[Industrial application field]

The present invention relates to an automatic character pattern formatting processing method for formatting input character patterns.

[Problems to be solved by conventional technology and invention] Conventionally,
In order to create a full-dot character pattern, the following two steps were followed.

■Read the letters using a scanner or TV camera,
Binarize it, input it into a computer as a character pattern, and store it.

■Display the memorized character pattern on a CRT, etc., and correct it manually.

As mentioned above, since the character pattern of the alphabet displayed on the CRT in step 2 was "manually corrected", it required a great deal of labor, and there was a problem in that the productivity of the character pattern was significantly impaired. Ta. Also, the correction is
There was a problem in that the quality of the character patterns created was not constant, as a large part depended on the aesthetic sense of the people involved.

In order to solve this problem, the present applicant's r character pattern automatic formatting processing method related to the claim of priority (domestic priority) pursuant to the provisions of Article 42-2, Paragraph 1 of the Patent Law (Patent Application No. 1982) No. 035666), Figures 3 to 7
By the means explained using the figures, prominent noises and missing noises shown in Fig. 8(a)
Modified as shown in b).

Furthermore, in order to solve the problem, Article 42-2, Section 1 of the Patent Law
The applicant's claim of priority pursuant to the provisions of ``
"Character pattern automatic formatting method" (Patent application No. 140, 1986)
No. 536), the above patent application No. 035666 of 1988
In addition to the noise removal described in the above issue, by the means explained using figures 9 to 1),
The protrusion like e and the missing part like C1d are the first
It was modified as shown in Figure 2(b).

However, there was a problem in that the quality of the character pattern corrected by the noise removal described above was still insufficient.

An object of the present invention is to automatically format character patterns to generate beautiful character patterns. Therefore, in addition to the noise removal described above, the present invention performs vector shaping processing to remove noise at bending points and end points of the contour vector.

[Means for solving problems]

FIG. 1 shows a block diagram of the principle of the present invention. In FIG.

The graphic input device 2 is for inputting a binary image generated by scanning the character 1 using a scanner or the like into a computer or the like.

The input character pattern 3 is a binary image of a letter read by the graphic input device 2.

The background noise removal means (noise removal means) 4 removes noise from the input character pattern 3.

The linear noise removing means 5 is a means for further removing linear noise from the input character pattern 3 from which the background noise has been removed by the background noise removing means 4.

The vertical/horizontal line shaping means 6 recognizes the vertical lines and horizontal lines of the character pattern from which noise has been removed by the noise removing means 4, and aligns the line widths. At this time, the vector shaping means 6-2 forming the vertical/horizontal line shaping means 6 removes noise present in the contour vector.

The output character pattern 7 is a beautiful character pattern shaped by the vertical/horizontal line shaping means 5.

(Function) As shown in FIG.
The background noise removing means (noise removing means) 4 removes the background m sound of this input character pattern 3, the linear noise removing means 5 removes the noise on the straight line, and the vertical/horizontal line shaping means 6
After noise present in the contour vector is removed, vertical lines and horizontal lines are recognized respectively, line widths are made equal, and a clean output character pattern 7 is output.

Therefore, by removing background noise, removing straight edge noise using black runs, removing noise using contour vectors, and controlling the line width of vertical and horizontal lines from the input character pattern 3, which is a binary character form, It becomes possible to automatically generate beautiful character patterns of good quality.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be explained in detail using FIGS. 2 to 14.

In FIG. 2, the background noise removing means 4 is composed of an ambient noise removing means 4-1, an isolated noise removing means 4-2, and a contour noise removing means 4-3.

The surrounding noise removal means 4-1 removes noise generated around a required character such as a character frame of a character.

The isolated noise removal means 4-2 removes isolated noise that does not touch the character pattern.

The contour noise removing means 4-3 removes small noises of several dots that touch the character pattern.

The straight line noise removal means 5 is composed of a black run extraction means 5-1, a relationship extraction means 5-2, and a line deletion/insertion means 5-3.

The vertical/horizontal line shaping means 6 is composed of a vectorizing means 6-1, a vector shaping means 6-2, a line width controlling means 6-3, and a full dot forming means 6-4. The vectorizing means 6-1 vectorizes the outline of the character pattern. The vector shaping means 6-2 removes noise existing in the vectorized contour vector. The line width control means 6-3 recognizes vertical lines and horizontal lines from the contour vector and controls the line width. Full dot conversion means ・
6-4 is for restoring the outline of a character pattern from the outline vector and filling the inside thereof to generate a shaped character pattern.

Note that 1 to 3 and 7 have the same effect as those shown in FIG. 1, so their explanations will be omitted.

Next, the operation of the configuration of one embodiment of the present invention shown in FIG. 2 will be explained in detail.

Ambient noise removal means 4-1 = Figure 3 (Since it is necessary to remove the unnecessary surrounding character frame surrounding the character part in el, operate the mask pattern in Figure 3 (bl to Ta). The character pattern part is expressed using "1" and the background is expressed using "0°".

Step (l) The part within distance W from the two-character frame is '2''
Fill it with

Step (2): Using a mask pattern (bl), scan the character pattern with +x in the main scanning direction and +Y in the sub-scanning direction. If there is a "2" at any time, the "x" pixel is set to "2", thereby removing noise mainly present in the upper left part of the character area.

Step (3): Using the mask pattern (C1, scan the character pattern with -X in the main scanning direction and +Y in the sub-scanning direction, with "1" in the "x" part and one in the "O" part) But if there is “2”, set the pixel of ′×” to “2”,
As a result, noise mainly present in the upper right portion of the character area is removed.

Step (4): Using mask pattern Td), scan the character pattern with +X in the main scanning direction and -Y in the sub-scanning direction. However, if there is a “2”, the 1×” pixel is set to “2”.

As a result, noise mainly present in the lower left portion of the character area is removed.

Step (0: Using mask pattern (el), scan the character pattern with -x in the main scanning direction and -Y in the sub-scanning direction. If there is '2', the pixel of 'x' is set to '2'. This removes noise mainly present in the lower right part of the character area.

Step (6): Change the pixel value of “2” to “0°”
(set to zero) (set as background).

Through the above operations, noise existing around the character pattern is removed.

Isolated noise removal means 4-2: FIG. 4 shows an example of an isolated noise removal mask pattern, which removes isolated noise that does not touch a character pattern.

Step (1): Scan the character pattern using the mask pattern in Figure 4, and all “O” (circle) parts are “O”
(zero), set the "x0" part to 0" (zero).

Contour noise removal means 4-3: FIG. 5 shows an example of a contour noise removal mask pattern, which removes small noises of several dots that touch a character pattern.

Step (Step 1: Scan the character pattern using the mask pattern in Figure 5, and the "x" part is 1" and 10"
If there are three or more consecutive “1”s in the (circle) part, “
Set the “x” part to “2”.

Step (2) Set the pixel of "62" to "0" (zero).

Next, the removal of noise that has entered the character pattern using steps (3) to 6) will be explained.

Step (3) Scan the character pattern with the z main scanning direction set to +x and the g+J scanning direction set to +Y to create the pattern shown on the left side of Figure 6 Ta) (1 dot correction) and (al' (2 dot correction)). If they match, they are modified to the patterns shown on the right.This removes the small one-dot and two-dot noises on the left side that touched the character pattern.

Step (4): -X for main scanning direction, +Y for sub-scanning direction
The character pattern is scanned as follows, and if it matches the pattern shown on the left in FIGS. 6(b) (1-dot correction) and (b)' (2-dot correction), it is corrected to the pattern shown on the right. This removes the small one-dot and two-dot noises on the right side that touched the character pattern.

Step (51: Main scanning direction +Y, sub-scanning direction +X
The character pattern is scanned as shown in FIG. 6, and if it matches the pattern shown on the left side of FIG. The small one-dot and two-dot noises in the upper part that touched the character pattern are removed.

Step (6): -Y in the main scanning direction, 10X in the sub-scanning direction
The character pattern is scanned as 61st ffl(d+(1
(dot correction) and (d)' (two-dot correction)
If the pattern matches the pattern shown on the left side, the pattern shown on the right side is corrected. This eliminates small one-dot and two-dot noises in the lower part that touched the character pattern.

Next, using steps (7) to α, noise that has protruded from the character pattern into the background is similarly scanned and removed.

Step (7): Change the main scanning direction to +x! ! The character pattern is scanned with the +1 scanning direction set to +Y, and the
dot correction) and (al' (two dot correction)
If the pattern matches the pattern shown on the left side, the pattern shown on the right side is corrected. This eliminates the small one-dot and two-dot noises on the left side that touched the character pattern.

Step (8): -X for main scanning direction, +Y for sub-scanning direction
The character pattern is scanned as follows, and if it matches the pattern shown on the left side of FIG. Small one-dot and two-dot noises on the right side touching the character pattern are removed.

Step (9): +Y in main scanning direction, +X in sub-scanning direction
The character pattern is scanned as follows, and if it matches the pattern shown on the left side of FIG. 7(e) (1 dot correction) and (C1' (2 dot correction)), it is corrected to the pattern shown on the right side. , the small one-dot and two-dot noises in the upper part that touched the character pattern are removed.

Step αI: Scan the character pattern with -Y in the main scanning direction and +X in the sub-scanning direction, and match the pattern shown on the left side of Fig. 7 Ta) (1 dot correction) and (d3'' (2 dot correction) In this case, the patterns shown on the right are corrected.As a result, the small one-dot and two-dot noises in the lower part that touched the character pattern are removed.

FIG. 8 shows an example of noise removal processing obtained by the invention described in Japanese Patent Application No. 035666 of 1988 by the present applicant. This is an example of processing obtained by the vectorization means 6-1, line width control means 6-3, and full dot formation means 6-4.

FIG. 8(a) shows the input character pattern of the character capital 1 read by the graphic input device 2, and FIG. For convenience of production, only the outlines of input character pattern 3 and output character pattern 7 are shown. In reality, the inside of this contour line is filled in.As can be seen by comparing the two patterns, the inside of this outline is filled in. Steps in straight lines (■ in the figure), small noises in character patterns (@ in the figure), etc. are automatically removed.

Next, in addition to the noise removal described above, linear noise removal obtained by the invention described in Japanese Patent Application No. 140536 of 1988 by the present applicant will be explained in detail using FIGS. 9 to 12. do.

Black run extraction means 5-1: Here, black runs in the horizontal and vertical directions are extracted from the character pattern.

Step (2) Scan the character pattern with +X in the main scanning direction and +Y in the sub-scanning direction, and perform the processing in the next steps (2) to (2).

Step 2: Capture the point where the pixel value changes from 0 to 1, and set that point as x.

Step ■ Capture the point where the pixel value changes from 1 to 0, and mark that point with X! shall be.

Step N If (Xi -X+) is greater than or equal to the threshold, pixels (x, ~ run).

Step v Scan the character pattern with +Y in the main scanning direction and →X in the sub-scanning direction, and perform the processing in steps ① to ②.

Step (2) Capture the point where the pixel value changes from 0 to 1, and set that point as Y.

Step (2) Capture the point where the pixel value changes from 1 to 0, and set that point as Yt.

Step ■ If CYt Y+) is greater than or equal to the threshold, then (Y
+ to yz) are extracted as valid black runs (see FIG. 9 (bl)).

Relationship extraction means 5-2: Here, black run extraction means 5-1
Examine the interrelationships of the black runs extracted with
Extract parts that appear to be vertical lines (see Figure 10).

Step I +Y force direction scanning and black run extraction means 5-
Detect the horizontal black run extracted in step 1.

Step■ Step! The length of the black run extracted in step 1 is compared with the length of the upper and lower black runs adjacent to the black run, and if the difference is within the threshold value, the black run is extracted as a horizontal line (see Figure 1.theta.ial).

Step m +x force direction scanning and black run extraction means 13
to detect vertical black runs.

Step ■ Compare the length of the black run extracted in step ■ with the lengths of the left and right black runs adjacent to the black run, and if the difference is within the darkness value, extract the black run as a vertical line (No. (See Figure 10).

Line deletion/insertion means 5-3: Here, relationship extraction means 5-3
Remove the noise adjacent to the horizontal and vertical lines extracted in step 2 (see Figure 1).

Step I Scan in the +Y force direction to detect a black run that is the boundary above the horizontal line, and set the Y coordinate of that row as Y, and set the X coordinates of the left and right ends of the black run as X, , and X8, respectively. do.

Step n (yt-1) row and (y,-2) row X
The number of white pixels between 1 and X1 is measured and set as Wr and Wz, respectively.

Step m (1) WI/W! If it is above a certain threshold, among the black pixels located between X and X1 in the (yt-1) row, those that have no black pixel directly above them are set as white pixels "0".

Step m (2) If Wr /Wz is smaller than a certain threshold, X in the (yt-1) row.

-X2 is set to black pixel "1".

Step IV Scan in the +Y direction to detect a black run that forms the lower boundary of the horizontal line, and set the X coordinate of that line to Y2. Also, let the X coordinates of the left and right ends of the black run be X3 and X4, respectively.

Step V X in (Yx+1) row and (Yt+2) row,
The number of white pixels between ~X4 is measured, and Wr, W! respectively. shall be.

If the step Vl tt+ W'+ /Wz is equal to or greater than a certain threshold value, among the black pixels located between X3 and X4 in the (Yi+1) row, those that have no black pixel directly below are made white pixels.

Step Vl (21Wr) If /Wt is smaller than a certain darkness value, set 1 between X3 and X4 in the (Yt+1) row.

Step (2) Scan in the +X direction to detect a black run that is the left boundary of the vertical line, and set the X coordinate of that column to Xl. Also, let the X coordinates of the upper and lower ends of the black run be Yl and yx, respectively.

Step ■ 71 in columns (XI-1) and (XI-2)
The number of white pixels between 72 and 72 is counted and set as WISWt.

Step IX-tt) If W l/ W z is greater than or equal to a certain darkness value, Yl to Y2 in the (X r 1 ) column
Among the black pixels in between, those that have no black pixel immediately to the left are made white pixels.

If step IK (21Wt /W! is smaller than a certain threshold, Y in column (XI-1).

-Y2 is set to 1.

Step x Scan in the +X direction to detect a black run that is the right boundary of the vertical line, and set the X coordinate of that column to X, and set the X coordinates of the top and bottom ends of the black run to Y, respectively.

%Y4.

Step XI Y in columns (XI+1) and (XI+2)
The number of white pixels between , and Y4 is measured and set as W, , W, respectively.

Step Xll (1) If Wr /Wt is greater than or equal to a certain darkness value, Y in the (X1+1) column.

Among the black pixels between ~Y4 , there is no black pixel immediately to the right of it. Make things white pixels.

Step X1) If 121 Wt /Wt is smaller than a certain darkness value, Y in the (XI+1) column.

-Y4 is set to 1.

Figure 1) takes a horizontal line as an example, where the - column is a detected horizontal black run, and the - columns above and below it are black dots adjacent to the horizontal line. Some of these black dots can be included in the horizontal line, while others are noise and should be removed. In this process, these black dots are inserted/deleted. Above the horizontal line, W+/W
If z<threshold, a black dot is inserted, and on the lower side, Wr/Wt≧dark value, a black dot is deleted.In the case of insertion, it is determined that a part is missing and is replenished, and in the case of deletion, it is added at the intersection with the vertical line. It is excluded as noise that tends to appear in

Vectorization means 6-1: Here, the outline of the character pattern is vectorized. This vectorization is based on the “pattern information compression method” (patent application No. 6O-048895) by the present applicant.
), extracting bending points of the contour, generating vectors connecting the bending points, etc.

Figure 12 is Patent Application No. 140536 of 1988 filed by the present applicant.
This shows an example of processing for straight line noise removal obtained by the invention described in the above-mentioned No.
3. This is an example of the processing obtained by the processing of the full dot forming means 6-4. Figure 12 shows the (al input character pattern) and the (bl output character pattern shaped by linear noise removal).
Prominent noise of input character pattern (8) a % b s
e, etc. are deleted, missing noises c, d, etc. are interpolated, and surrounding noises/sounds f, g, h, etc. are removed.

Next, in addition to the linear noise removal described above, noise removal at bending points and end points of the contour vector by the vector shaping means 6-2 will be explained in FIGS. 13 and 14.
This will be explained in detail using figures.

Vector shaping means 6-2 Ni-step ■ For a series of bending points (Pz: 1=O1), =・n+1) sandwiched between horizontal lines or vertical lines, calculate the distance between each bending point and the horizontal line/vertical line by D5. shall be.

At this time, if Dl≦darkness value, but x=1%2...n holds, the inflection point sequence (PI: i・0.1,...n+1) is deleted, and the noise generated between the horizontal and vertical lines is removed. (see FIG. 13(a)).

Step ■ For a series of bending points connected to a horizontal line or a vertical line (Pl: i=0.1.-n+1), set the distance MDt between each bending point and the horizontal line/vertical line, where 1≦i≦ If Dm≦threshold holds true in the range of
k) and extend it to the position of straight line Pm (see Fig. 13(b)).

Step ■ For the contour line V between horizontal line v1 and vertical nvx, (angle between V+ and v8)≦threshold value or (angle between VZ and V)≦threshold value or 1v81
If ≦threshold holds true, V is deleted and an intersection point is generated on the extension of vl and V. As a result, noise in the right angle portion can be removed (see FIG. 1) (see C1).

FIG. 14 shows an explanatory diagram of the vector shaping processing result.
Figure 4+8) shows the input character pattern, and Figure 14+8l
indicates output character pattern 7 without vector formatting, and the first
As can be seen by comparing Figure 4 (C1 shows the output character pattern 7 with vector shaping) and (C1), the part a in the figure is the result of linear noise removal processing (Figure 13). The noise has been removed by (a)), the part marked with square in the figure is the noise removed by the end point noise removal process (Fig. 13(b)), and the part C in the figure is the result of right angle correction. Noise has been removed by the processing (FIG. 13(C)).

Line width control means 6-3: This is for shaping the contour vector of a character pattern by recognizing vertical lines and horizontal lines from the contour vector vectorized by the vectorization means 6-1 and controlling the line width. It is. Here, for example, the present applicant's "phase conversion method" (patent application No. 62-26053)
No. 9) is used.

Full dot forming means 6-4: Generates a shaped character pattern by restoring the outline of the character pattern from the outline vector and filling the inside thereof. Here, for example, the applicant's "polygon internal area filling method"
(Japanese Patent Application No. 62-048896) is used.

〔Effect of the invention〕

As explained above, according to the present invention, a configuration is adopted in which background noise is removed from a binary image of a letter, straight edge noise is removed using black runs, and noise is removed from bending points and end points of contour vectors. Therefore, background noise can be removed, partial protrusions and missing parts of character contours can be automatically corrected, and curved points and end points of contour vectors can be automatically shaped. Thereby, it is possible to automate the shaping of character patterns, which was conventionally done manually, and it is possible to improve the productivity of character patterns and to make the quality of character patterns more uniform.

[Brief explanation of the drawing]

Figure 2 is a block diagram of the principle of the present invention, Figure 2 is a block diagram of the principle of the present invention.
Embodiment configuration diagram, Figure 3 is an explanatory diagram of ambient noise removal, Figure 4 is an example of an isolated noise removal mask pattern, Figure 5 is an example of a contour noise removal mask pattern, and Figures 6 and 7 are corrections in contour noise removal. Mask pattern example, Fig. 8 is a processing example, Fig. 9
The figure shows an example of extracting black runs, Fig. 10 shows an example of horizontal and vertical line extraction, Fig. 1) is an explanatory diagram of straight edge noise removal, Fig. 12 is an explanatory diagram of processing results using black runs, and Fig. 13 The figure shows an explanatory diagram of the vector shaping process, and FIG. 14 shows an explanatory diagram of the vector shaping process result. In the figure, 1 is a letter, 2 is a graphic input device, 3 is an input character pattern, 4 is a background noise removal means, 5 is a linear noise removal means,
6 is a vertical/horizontal line shaping means, 6-2 is a vector shaping means, 7
represents the output character pattern. Block diagram of the principle of the present invention Figure 1 R - J Configuration diagram of one embodiment of the present invention Figure 2 Example of a mask pattern for removing ambient noise Diagram for explaining ambient noise removal Figure 3 Example of a mask pattern for removing isolated noise Contour noise removal mask/ Example of f-turn Figure 4 Figure 5 (a) (a)'(c)
(c) 'Example of correction mask pattern for contour noise removal (Part 1) Example of correction mask pattern for contour noise removal (Part 2) m8 example of black foot Extraction example of horizontal and vertical lines > 2 lines adjacent to (Y+) line If 1≧threshold・・・・・・・・・−・Delete 1・・・・・・↑
De...Explanatory diagram for removing straight edge noise 1) Figure 12 Di P. (a) Straight-line noise removal process (1)) End point noise removal process (C) One-angle correction process Figure 14: Explanatory diagram of the result of formatting process (Part 2)

Claims (3)

[Claims] (1) In a character pattern automatic formatting processing method that formats an input character pattern, a background noise removal means (4 ), and vertical/horizontal line shaping means (6) for shaping the vertical lines and horizontal lines of the character pattern from which noise has been removed by the background noise removing means (4), and the vertical/horizontal line shaping means (6) An automatic character pattern formatting method characterized by being configured to output a formatted character pattern. (2) A method for shaping a character pattern input from an input device such as a scanner, which includes a background noise removing means (4) for removing noise around characters, isolated noise, and noise that contacts the outline of characters, and a character pattern. Linear noise removal means that extracts black runs from , detects horizontal and vertical lines from the mutual relationship of each black run, and corrects black dot groups adjacent to the contours of the horizontal and vertical lines by inserting/removing black dots ( 5), and a vertical/horizontal line shaping means (which vectorizes the outline of a character pattern, recognizes vertical and horizontal lines from the outline vector, aligns the line widths, and restores the character pattern from the outline vector whose line width has been controlled). 6) An automatic character pattern formatting method comprising: (3) In the vertical/horizontal line shaping means (6), vector shaping means (6-2) is provided for determining a contour vector and removing noise at bending points and end points of this contour vector. The character pattern automatic formatting method described in item 1) or item (2).