JPH03177982A - Image processing method - Google Patents

Abstract

PURPOSE:To express a recessed right angle part sharply by extracting a specific picture element as a boundary picture element when at least one of eight picture elements that exist in the neighborhood of the specific picture element is shown as a background picture element. CONSTITUTION:When at least one of the eight picture elements (b-i) that exist in the neighborhood of the specific picture element (a) is shown as the background picture element, the specific picture element (a) is set as the boundary picture element. As a result, the picture elements a1, a2, a3, and a4 are also recognized as the boundary picture elements, and it is decided that the picture element shown in hutching in figure is the boundary picture element of a graphic, and also, the picture element a5 at the corner of the recessed right angle part is recognized as the boundary picture element. Therefore, when the arrangement of a vector representing relation between adjacent picture elements is generated from a found boundary picture element and the contour of the graphic is reproduced by using the arrangement of the vector, the recessed right angle part can be expressed sharply.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing method, and more particularly to an image processing method in which boundary pixels representing features of a figure are extracted and the figure is vectorized for image processing.

[Conventional technology]

When recording figures in image processing, data formats include data that includes all the pixels that make up the figure, that is, raster data, and a predetermined set of pixels that represent the characteristics of the figure, such as a set of boundary pixels. There is vector data that is expressed as a vector array using two vectors. In general, vector data, particularly vector data of boundary pixels of a figure, requires less data than raster data, so the storage capacity of a memory for storing figures can be reduced. Furthermore, since the amount of data is small, various processing on graphics can be accomplished with fewer calculations. Therefore, for example, processing such as enlarging, reducing, or deforming a figure or forming a convex hull can be performed at high speed, and this vector data is effective data.

A conventional vectorization method for generating vector data of boundary pixels of a figure will be described below.

To generate vector data of boundary pixels of a figure,
As preprocessing, processing to extract boundary pixels of the figure is required. As shown in FIG. 30, the conventional boundary pixel extraction method focuses on the vicinity of a pixel 50 to be determined, and extracts vertically adjacent pixels 54 and 58 and horizontally adjacent pixels 5.
When any of the 4 pixels of 2.56 (so-called 4 neighboring pixels) is a background pixel, pixel 50 is extracted as a boundary pixel of the figure.This extraction process is applied to all pixels in the image or This is done for all pixels that make up the figure.Third
When the extraction process is performed on the figure 60 shown in FIG. 2,
In FIG. 33, the pixels indicated by hatching are the figure 60.
are extracted as boundary pixels.

Next, vector data is generated from the extracted boundary pixels. Tracing the boundary pixel array shown in FIG. 33 in a counterclockwise or clockwise direction, and using vectors that separately represent the eight directions of vertical, horizontal, and diagonal directions as shown in FIG. 31,
Represents the relationship between adjacent pixels. Each of these vectors is assigned a code. FIG. 34 shows, as an example, a code that is attached when the boundary pixel array shown in FIG. 33 is traced counterclockwise. A code string (r
660007660022222444444'' is vector data. Here, vector data generated using the code shown in FIG. 31 is particularly called a chain code.

[Problem to be solved by the invention]

However, in conventional vectorization methods, the second
As shown in FIGS. (A) to (D), if a background pixel exists diagonally to a pixel to be determined, the pixel to be determined is not determined as a boundary pixel. Therefore, as shown in FIG. 33, the pixel 62 at the corner of the concave right angle part of the figure is not extracted as a boundary pixel. Therefore, when the contour of a figure is reproduced from vector data generated by the conventional vectorization method, the corners of the concave right angles in the contour are reproduced obliquely, so the problem is that the concave right angles are not clearly expressed. was there.

The quality of the reproduced figures is important as an evaluation target for vector data, and especially when recording and reproducing vector data generated from characters such as kanji using conventional vectorization methods, the quality of the reproduced figures (characters) is important. Quality has a major impact on product value.

The present invention was developed in consideration of the above-mentioned facts, and an object of the present invention is to provide an image processing method that can clearly express the right angle portion of a concave portion.

[Means to solve the problem]

In order to achieve the above object distantly, the invention as claimed in claim (1),
As shown in FIG. 1, a specific pixel is a pixel of a figure, and 8 pixels bSc, d, eS f existing in the vicinity of the specific pixel a are
, when at least one pixel of gShS i is a background pixel, the process of making the specific pixel a a boundary pixel of the figure is performed from the specific pixel and 8 pixels existing in the vicinity of the specific pixel. The outer boundary pixel column and the inner boundary pixel column are determined by 9 pixels each in a fixed order. , vertical direction vectors V2, V6, horizontal direction vectors V1, which express the direction of eight pixels bSc, dSe, f, gSh, i existing in the vicinity of the specific pixel a, with the specific pixel a as a reference point, starting from the specific pixel a.
The relationship between adjacent pixels is represented by an array of eight vectors, Vo, and diagonal direction vectors V1, 4, 3, and 2, and image processing is performed.

Further, the invention described in claim (2) traces the outer boundary pixel column and the inner boundary pixel column of the figure so that the outer boundary pixel column and the inner boundary pixel column can be distinguished, and as shown in FIG. 8 pixels near the specific pixel a, c, dSeS f
, g, h, i, with the specific pixel a as the reference point, vertical vectors v2, v6, horizontal vectors V4, Vo, and diagonal vectors Vl,
In an image processing method that expresses the relationship between adjacent pixels by arraying eight vectors V3, v5, and V7, and performs image processing, for the outer boundary pixel column, one of the left and right vectors is placed before and after the diagonal vector. and the diagonal direction vector of the part where one of the up and down direction vectors exists and the direction changes by a predetermined acute angle in the opposite direction to the direction tracing the outer boundary pixel when the starting point of these three vectors is common, It is expressed by decomposing it into vectors that exist before and after the diagonal direction vector, and for the inner boundary pixel column, one of the left and right vectors and one of the up and down direction vectors exist before and after the diagonal direction vector, and these three vectors The diagonal vector in the part where the direction changes by a predetermined acute angle in the same direction as the direction tracing the outer boundary pixel when the starting point of It is characterized by the fact that

Furthermore, in the inventions of claims (1) and (2), eight vectors can be represented by codes, and the relationship between adjacent pixels can be represented by a chain code made up of code strings.

[Effect]

In the invention described in claim (1), a specific pixel of a figure is defined as a background pixel, and as shown in FIG. In this case, the specific pixel a is set as a boundary pixel of the figure.

As a result, for example, the pixels at, a2, a3, and a4 shown in FIGS. 2(A) to 2(D) also become boundary pixels, and the figure 60 shown in FIG.
When the boundary pixels of the figure are determined, the pixels indicated by hatching in FIG. 3 are determined to be the boundary pixels of the figure. In other words, the pixel a at the corner of the right angle portion of the concave portion is also set as a boundary pixel.

For this reason, an array of vectors indicating the relationship between adjacent pixels is generated from the boundary pixels determined based on the invention of claim (1), and the outline of the figure is reproduced using the array of vectors. In this case, the right angle part of the concave can be clearly expressed.

In the invention described in claim (2), in order to distinguish between the outer boundary pixel column and the inner boundary pixel column, for example, in the figure shown in FIG. 4, the outer boundary pixel column is rotated counterclockwise (arrow C in FIG.
The inner boundary pixel row is traced clockwise (in the direction of the arrow in FIG. 4), and the relationship between adjacent pixels is represented by an array of eight vectors.

For the outer boundary pixel column, one of the horizontal direction vector and one of the vertical direction vector exists before and after the diagonal direction vector, and the direction when the starting point of these three vectors is common is a predetermined acute angle outward. A diagonal vector of a portion changing in the opposite direction to the direction of tracing the boundary pixel is decomposed into vectors existing before and after the diagonal vector and represented.

When tracing an outer boundary pixel counterclockwise, the diagonal vectors that meet the above conditions are a left diagonal downward vector where a downward vector exists before and a leftward vector after, and a leftward diagonal vector where a leftward vector exists before. diagonally upward left vector with an upward vector after, diagonal downward right vector with a right vector before and a downward vector after, and diagonal downward right vector with an upward vector before and a right vector after. There are four types of vectors: a diagonally upward right direction vector; Furthermore, when tracing an outer boundary pixel clockwise, the diagonal vectors that meet the above conditions are diagonally upward right vectors where there is a right vector before and an upward vector after, and diagonal upward right vectors where there is a downward vector before. A diagonally downward right direction vector that exists and a rightward vector exists after, a diagonally upward left direction vector that exists before an upward vector and a leftward vector, and a downward direction vector that exists before a leftward vector. There are four types: a diagonally downward left direction vector in which Each of these four types of diagonal direction vectors is decomposed into vectors that exist before and after them and is expressed.

As an example, if we pay attention to the part pointed by arrow A in Figure 4, there is an upward vector v2 before the diagonal vector V+ shown in Figure 5 (A>), and a rightward vector V exists after the diagonal vector V1. As shown in Fig. 5(C), when these three vectors V2, V+, and Vo have a common starting point, their directions move toward the outer boundary pixel by a predetermined acute angle θ (here, 45°). The diagonal direction vector V1 of this part is changed in the opposite direction (clockwise in this case) to the direction of tracing the diagonal direction vector V, as shown in FIG. 5(B). It is divided into a vector V2 and a rightward vector v0.

In addition, for the inner boundary pixel row, one of the horizontal direction vector and one of the vertical direction vector exists before and after the diagonal direction vector, and when the starting point of these three vectors is made common, the direction is a predetermined acute angle. The diagonal vector of the portion changing in the same direction as the direction of tracing the outer boundary pixel is divided into vectors existing before and after the diagonal vector and represented.

The diagonal vectors that meet the above conditions when the inner boundary pixels are traced clockwise are the same as the four types of diagonal vectors when the outer boundary pixels are traced counterclockwise.

Furthermore, the diagonal vectors that meet the above conditions when tracing the inner boundary pixel counterclockwise are the same as the four types of diagonal vectors described above when tracing the outer boundary pixel clockwise.

Each of these four types of diagonal direction vectors is decomposed into vectors that exist before and after them and is expressed.

For example, if we pay attention to the part indicated by arrow B in Figure 4,
A downward vector V6 exists before the diagonal vector V5 shown in FIG. 6(A), and a leftward vector V4 exists after the diagonal vector v5. As shown in FIG. 6(C), when these three vectors V6, Vs, and Vs have a common starting point, their directions are at a predetermined acute angle θ (here, 4
5°) in the same direction as the direction of tracing the inner boundary pixel (clockwise here). The diagonal vector v5 in this portion is divided into a downward vector V6 and a leftward vector V4, which exist before and after the diagonal vector V5, as shown in FIG. 6(B).

This antagonist, the corner pixel of the concave right angle part of the figure, is also incorporated into the vector array, and when the contour of the figure is reproduced using the vector array, the concave right angle part can be clearly expressed.

In the invention described in claim (3), the eight vectors are represented by codes, and the relationship between adjacent pixels is represented by a chain code made up of code strings. The codes representing the eight vectors in the chain code are numbers, and the chain code is a string of numbers. Therefore, it is easy to handle in a device that processes digital data, such as a computer.

〔Example〕

First Embodiment Hereinafter, the process of attaching a chain code to the figure 10 shown in FIG. 7 using the image processing method of claim (1) of the present invention will be explained with reference to the flowchart of FIG. 10. . In addition, in Fig. 7, the pixel value of the hatched part (figure pixel) = 1,
It is assumed that the pixel value of the white part (background pixel) is −0, which is a binary image.

In step 102, it is determined whether the pixel to be determined is a boundary pixel by referring to the pixel to be determined as a boundary pixel and eight pixels located around the pixel to be determined (hereinafter referred to as eight neighboring pixels). The determination as to whether or not the pixel is a boundary pixel is made when the pixel value of the pixel to be determined is l and the pixel value of at least one pixel among the eight neighboring pixels is 0. It is determined that That is, the pixel to be determined is a, and the three neighboring pixels are c, dSe, and f.

If gShS is l and AND in logical operation j is represented by X, then aX XCX XeX XgX XI =
It can be expressed by one logical formula.

If the determination target pixel is determined to be a boundary pixel in step 102, the determination target pixel is registered as a boundary pixel in a storage means such as a frame memory in step 104. If it is determined that the determination target pixel is not a boundary pixel, step 104 is not executed. In step 106, it is determined whether the extraction of boundary pixels has been completed. If the extraction of boundary pixels has not been completed, the process returns to step 100 and steps 100 to 106 are repeated. Step 106
If it is determined that the extraction of boundary pixels has been completed, the process moves to step 108.

At this time, pixels indicated by hatching in FIG. 8 are pixels registered as boundary pixels. Furthermore, the pixel 12 indicated by dark hunting is a pixel that cannot be extracted using the conventional boundary pixel extraction method. in this way,
Conventionally, at the outer boundary and inner boundary of the figure 10,
The pixel 12 at the corner of the right angle part of the concave that was not extracted is newly extracted. Therefore, when vector data is generated from the registered boundary pixels using a vectorization method similar to the conventional method, and the generated vector data is used to reproduce the figure 10, the right angle part of the concave in the reproduced figure 10 will not be clearly expressed. Ru.

The process of generating vector data from registered boundary pixels will be briefly described below.

In step 108, the registered boundary pixels are assigned different labeling numbers for each boundary pixel column. As a result, in the figure 10, different labeling numbers are assigned to outer boundary pixels and inner boundary pixels to distinguish them.

In step 110, a starting point for each boundary pixel column to which a labeling number has been assigned is determined and registered. In the graphic 10, the starting point of the outer boundary pixel and the starting point of the inner boundary pixel are registered. Note that the end point of each boundary pixel column may be registered in step 110.

In step 112, vector data (here, chain code) is generated. In other words, the boundary pixels with the same labeling number are traced counterclockwise or clockwise from the registered starting point, and a vector indicating the direction in which the next boundary pixel exists is shown in FIG. The code is converted into a code corresponding to the eight vectors shown, and the code is attached to the boundary pixel. A vector array of these codes is called a chain code. The generation of the code can be determined based on the pixel to be processed, eight pixels in the vicinity of the pixel to be processed, and the code attached immediately before. As a result of attaching the chain code to the figure 10, the chain code is also attached to the pixel 12 at the corner of the right angle part of the concave as shown in FIG.

As explained above, in the first embodiment, the pixel value of the determination target pixel is the pixel value of the figure 10, and the pixel value of at least one pixel among the eight neighboring pixels of the determination target pixel is the background pixel. Since the pixel to be determined is registered as a boundary pixel of the figure when the value is a value of When vector data is generated and the contour of a figure is reproduced using the generated vector data, the concave right angle part of the figure can be clearly expressed.

Note that when the image processing method of the first embodiment is used to extract boundary pixels of a step-like boundary in units of one pixel as shown in FIG. 11, pixels indicated by dark hatching are extracted as boundary pixels. . Here, the pixel 14 is a boundary pixel that is not extracted by the conventional boundary pixel extraction method but is newly extracted by the image processing method of the first embodiment. The step-like boundary of each pixel is a straight line l in an oblique direction, and there is no need to clearly express the right angle part of the step portion, and there is no need to use the pixel 14 as a boundary pixel. However, since the pixel 14 is extracted as a boundary pixel, the number of boundary pixels extracted in a step-like boundary of one pixel is approximately twice that of the conventional method.

This has an adverse effect such as an increase in processing time during processing such as chain code generation. For this reason, when the step-like boundary of one pixel unit continues for a predetermined length or more, an exception process or post-processing may be performed in which the pixel 14 is not set as a boundary pixel.

In addition, in the first embodiment, eight vectors in the vertical, horizontal, and diagonal directions are used to express the relationship between adjacent pixels for the boundary pixels of extracted figures. However, when boundary pixels are extracted using the boundary pixel extraction method of this embodiment, the relationship between adjacent pixels of a boundary pixel is limited to four directions, vertical and horizontal directions, so only four vectors are extracted. can be used to express the relationship between adjacent pixels. For this reason, it is only necessary to prepare four types of codes to be attached to vectors, and when the above code is used as a chain code, the amount of data in one code is only 2 bits, which reduces the amount of data to half compared to the conventional chain code. can do.

Second Embodiment Next, an embodiment of the image processing method according to claim (2) of the present invention will be described.

FIG. 13 shows the result of extracting boundary pixels from the figure 20 shown in FIG. 12 by a conventional method, tracing the boundary pixels counterclockwise and attaching a chain code. The chain code for the outer boundary of shape 20 is r6666666002210
0076600222222244665444322
It becomes 44J. As is clear from FIG. 13, the pixel 16 at the corner of the right angle part of the concave is not extracted as a boundary pixel by the conventional boundary pixel extraction method, and therefore no chain code is attached to it.

Thereafter, based on the chain code, the detection of the right angle part of the concave, that is, the detection of pixel 16, and the process of including pixel 6 in the vector data are performed.

Here, if we assume that the scanning line direction is the horizontal direction, the direction perpendicular to the scanning line direction is the vertical direction, and the direction intersecting the scanning line direction at an angle of 45° is the diagonal direction, the pattern showing the right angle part of the concave in question is , Fig. 14 (A), Fig. 15 (A), Fig. 16 (A), Fig. 1
There are only four types shown in Figure 7 (A). Note that in a concave right-angled part formed by a straight line in an oblique direction, any of the four neighboring pixels of the corner pixel 18 is a surrounding pixel, as shown in FIG. 18, and therefore is extracted as a boundary pixel.

Figure 14 (A) shows a right-angled part of a concave that opens to the lower left, and the chain code progresses to "...007" in the horizontally arranged part of the boundary pixel row and reaches the corner pixel 20. After that, it moves to the vertically arranged part and progresses as "66...". The vectors that define right angles in this figure are the rightward vector attached to pixel 22, the diagonally downward rightward vector attached to pixel 24, and the downward direction vector attached to pixel 26, and these chain codes are r076". Also, when these three vectors have a common starting point, their directions change clockwise (
(See Figure 31). This ro 76 in chain code row
When the combination J occurs, it is immediately clear that it is a concave right-angled part that opens to the lower left. Now modify the chaincode to include the corner pixel 20. Figure 14 (
B) shows the result of changing the chain code, changing the chain code attached to 124 from "7" to "
0” and then add a chain code to pixel 20 in the corner.
6" is given. This is chain code set r076
This corresponds to the operation to change J to r0066.

By changing this chain code, the vector attached to the last pixel 24 in the horizontally arranged portion of the boundary pixels has been changed to a vector pointing to the corner pixel 20, and the corner pixel 20 is A vector pointing to the first pixel 26 of the portion arranged in the direction is attached.

Figure 15 (A) shows a right-angled part of a concave that opens to the lower right, and the chain code progresses as "...221-" in the vertically arranged part of the boundary pixel row, and moves to the corner pixel. 20, and then moves to the portions arranged in the left and right direction and progresses as "00...". The vectors that define the right angles in this figure are the upward direction vector attached to pixel 22, the diagonally upward right direction vector attached to pixel 24, and the right direction vector attached to pixel 26, and these chain codes are It becomes "210". Furthermore, when these three vectors have a common starting point, their directions change clockwise (see FIG. 31). This r210 in the chain code row
'', it is immediately clear that it is a concave right-angled part that opens to the lower right. Now modify the chaincode to include the corner pixel 20. Figure 15 (
B) shows the result of changing the chain code, changing the chain code attached to pixel 24 from "l" to "2".
” and then add the chain code “0” to the corner pixel 20.
” is given. This is a set of chain cords R210.”
Corresponds to the operation of changing '2200J'.

By changing this chain code, the vector attached to the last pixel 24 of the vertically arranged part of the boundary pixels is changed to a vector pointing to the corner pixel 20, and the corner pixel 20 is A vector pointing to the first pixel 26 of the portion arranged in the direction is attached.

Figure 16 (A) shows the right angle part of the concave that opens to the upper left, and the chain code progresses as "...665" in the vertically arranged part of the boundary pixel row and reaches the corner pixel 20. After that, it moves to the part arranged in the left and right direction and progresses as "44...". The vectors that define the right angles in this figure are the downward direction vector attached to pixel 22, the diagonally downward left direction vector attached to pixel 24, and the left direction vector attached to pixel 26, and these chain codes are r654". Also, when these three vectors have a common starting point, their directions change clockwise (
See Figure 31>. This r654 in the chain code string
When the combination occurs, it is immediately clear that it is a right-angled part of a concave that opens to the upper left. Now modify the chaincode to include the corner pixel 20. Figure 16 (B
) indicates the result of changing the chain code, changing the chain code attached to pixel 24 from "5" to "6".
and add chain code "4" to pixel 20 in the corner.
is giving. This corresponds to the operation of changing chain code set r654J to r6644J.

By changing this chain code, the vector attached to the last pixel 24 of the vertically arranged part of the boundary pixels is changed to a vector pointing to the corner pixel 20, and the corner pixel 20 is A vector pointing to the first pixel 26 of the portion arranged in the direction is attached.

FIG. 17(A) shows a right-angled part of a concave opening to the upper right, and the chain code progresses to "...443" in the part arranged in the left-right direction among the boundary pixel rows, and then returns to the corner pixel s.
20, and then moves to the vertically arranged portions and progresses as "22...". The vectors that define the right angles in this figure are the left direction vector attached to pixel 22, the diagonally upward left direction vector attached to pixel 24, and the upward direction vector attached to pixel 26, and these chain codes are It becomes r432J. Also, when these three vectors have a common starting point, their directions change clockwise (
(See Figure 31). This r432J in the chain code row
When the combination occurs, it is immediately clear that it is a right-angled part of the concave that opens to the upper right. Now modify the chaincode to include the corner pixel 20. Figure 17 (B
) indicates the result of changing the chain code, changing the chain code attached to pixel 24 from "3" to "4".
and add chain code "2" to corner pixel 20.
is giving. This corresponds to the operation of changing chain code set r432J to "4422J."

By changing this chain code, the vector attached to the last pixel 24 of the portion arranged in the horizontal direction among the boundary pixels has been changed to a vector pointing to the corner pixel 20. A vector pointing to the first pixel 26 of the portion arranged in the direction is attached.

FIG. 19 shows the result of performing the above processing on the graphic 20 shown in FIG. 13. The chain code for the outer boundary of shape 20 is r66666660022200000
6660022222224466644444222
It becomes 44J. As is clear from Figure 19, figure 20
Pixels 16 at the corners of the right-angled portion of the concave are included in the chain code, and the right-angled portion of the concave is clearly expressed.

The above process was for the chain code attached by tracing the outer boundary pixels of the figure counterclockwise, but when processing the inner boundary pixels, that is, the hole, the chain code is also attached by tracing the outer boundary pixels of the figure in a clockwise direction. You can process it in the same way by adding .

Figure 20 (A) shows a state in which a chain code is attached to the inner boundary pixel, and the chain code is r000766.
6544432221”. Here, if we consider that the end of the chain code string follows the beginning of the chain code string, the chain code string at the right angle of the concave will be ro 76J, r654J,
"321", r210". If these are processed in the same way as the outer boundary pixels, the chain code will be "0066".
”, r6644J, r4422J, and “2200” (see FIG. 20(B)). That is, the diagonal direction vectors "7", "5", "3", and "1" are decomposed into vertical and horizontal vectors that exist before and after, and the pixel 28 at the corner of the right angle section is Codes "0", "6", "4", and "2" are assigned to the existing pixels, and codes "6" and "4" are assigned to the pixels 30 that are behind the pixel 28. ”, “2”, and “0” are added. Thereby, it is possible to clearly express the right angle part of the concave even with respect to the inner boundary pixel, that is, the hole.

Above, we have shown an example in which the outer boundary pixel row is traced counterclockwise and the inner boundary pixel row is traced clockwise to attach a chain code. Conversely, the outer boundary pixel row is traced clockwise and a chain code is attached. It is also possible to attach In this case, if a chain code is attached by tracing the inner boundary pixel column counterclockwise, the chain code attached to the concave right angle part of the outer boundary pixel column and the inner boundary pixel column will be the same, and the same process will be performed. can.

FIG. 22(A) is a diagram similar to that shown in FIG. 14 with clockwise chain cords attached, and the set of chain cords defining the right angle is r234J. This is the 22nd
As shown in Figure (B), by changing it to "2244", a sharp right-angled portion can be obtained.

FIG. 23(A) is a diagram similar to that of FIG. 15 with clockwise chain cords attached, and the set of chain cords defining the right angle is r456J. This is the 23rd
As shown in Figure (B), by changing it to "4466", a clear right angle part can be obtained.

FIG. 24(A) is a diagram similar to that shown in FIG. 16 with a clockwise chain code added thereto, and the chain code set defining the right angle is "012". This is the 24th
As shown in Figure (B), by changing it to "0022", a clear right angle part can be obtained.

FIG. 25(A) is a diagram similar to that of FIG. 17 with clockwise chain cords attached, and the set of chain cords defining the right angle is r670J. This is the 24th
As shown in Figure (B), by changing the value to "6600", a clear right-angled portion can be obtained.

FIG. 26 shows processing when a chain code is attached counterclockwise to an inner boundary pixel, that is, a hole. As shown in FIG. 26(A), this chain code is r5666700012223444J. Considering that the end of the chain cord string follows the beginning of the chain cord string, the chain cord at the concave right corner is r234J, "4", similar to that shown in FIGS.
56'', ``012'', and ``670'', and this becomes the 26th
As shown in figure (B), r2244J, "4466", r
o By changing to 022J and r6600J, the right angle part of the concave can be clearly expressed.

In this way, in the second embodiment, when three consecutive code strings in the vector array consisting of chain codes are code strings indicating a concave right angle part, codes indicating diagonal vectors are present before and after the chain code. Since the image processing method shown in the second embodiment is used to decompose the contour of a figure into a vector array consisting of chain codes processed by the image processing method shown in the second embodiment, the right angle part of the concave part can be clearly seen. can be expressed.

Although the second embodiment has been described using a chain code as an array of vectors to be processed,
It can also be applied to arrays of vectors other than chaincodes. As an example, if a code as shown in FIG. 27 is attached to a vector in eight directions, and a code is attached to the figure 20 shown in FIG. 12 using this as a reference, a code representing the vector arrangement as shown in FIG. Column rCCCCCCCDDA
AHDDDGCCDDAAAAAAAAAABBCCFBBB
EAABBJ is generated. In this code string, before and after the codes H, G, F, and E indicating the diagonal direction vector, there is one of the code A1C indicating the vertical direction vector and the codes B and D indicating the left and right direction. In addition, the three vectors indicated by the three consecutive codes sandwiching the code indicating the diagonal direction vector change their directions clockwise when the starting point of the three vectors is common (see Figure 27). . This reveals that it is a right-angled part of a concave, and the codes H, G, and PSE are broken down and expressed into codes that exist before and after them. As a result, the concave right angle portion of the figure 20 is clearly expressed as shown in FIG.

In addition, in the second embodiment, the right angle part of the concave was determined using the vector code attached to consecutive "3@element", but the right angle part of the concave is determined by using the vector code attached to the consecutive "3@element". In some cases, it is preferable to express the stepped boundary of two pixels as shown in the figure not as a series of right angles, but rather as a straight line l.

In this case, the above-mentioned judgment may be made only when a straight line of a certain length or more is continuous in at least one of the front and back of the three pixels to be judged. -For example, instead of determining "076", r
o07666", when one side of the right-angled part is 3 pixels or more long, it is determined to be a right-angled part.
You may also convert it to .

〔Effect of the invention〕

In the image processing method according to claim (1), the specific pixel is extracted as a boundary pixel when at least one of the eight pixels existing in the vicinity of the specific pixel is a background pixel. Right angles can be clearly expressed,
In the image processing method according to claim (2), which has the excellent effect of Since it is decomposed into vectors existing before and after , it has the excellent effect of being able to clearly express the right angle part of the concave.

[Brief explanation of drawings]

FIG. 1 is a conceptual diagram showing the concept of eight neighborhoods and eight vectors according to the image processing method of the present invention, and FIGS. 2 and 3 show boundary pixels newly extracted according to the invention of claim (1). Conceptual diagrams, FIGS. 4 to 6 are conceptual diagrams for explaining the invention of claim (2) in detail, and FIGS. 7 to 9 are conceptual diagrams showing figures for explaining the first embodiment of the present invention. 10 is a flowchart explaining the operation of the first embodiment, FIG. 11 is a conceptual diagram explaining exception handling in the first embodiment, and FIGS. 12 and 13 are diagrams explaining the second embodiment. 14 to 18 are conceptual diagrams showing the concave right angle part when tracing the outer boundary pixel row counterclockwise. FIG. 19 is a conceptual diagram showing the processing results of the second embodiment. , FIG. 20 is a conceptual diagram showing the right angle part of the concave when the inner boundary pixel row is traced clockwise,
FIG. 21 is a conceptual diagram explaining exception handling in the second embodiment, FIGS. 22 to 25 are conceptual diagrams showing the concave right angle part when the outer boundary pixel row is traced clockwise, and FIG. A conceptual diagram showing the right angle part of the concave when tracing the inner boundary pixel column counterclockwise, FIG. 27 is a conceptual diagram showing examples of other codes attached to eight vectors, and FIGS. 28 and 29 are Conceptual diagram explaining the processing using the code shown in Figure 27, No. 30
The figure is a conceptual diagram explaining the so-called 4-neighborhood concept, Figure 31 is a conceptual diagram showing the codes assigned to eight vectors when generating a chain code, and Figures 32 to 34 are the conventional boundary pixel extraction method and FIG. 2 is a conceptual diagram showing figures for explaining a vectorization method. 10... Figure, 12... Pixel at the right angle corner of the concave, 16... Pixel at the right angle corner of the concave, 20... Figure.

Claims (3)

[Claims] (1) A specific pixel is defined as a pixel of a figure, and when at least one of the eight pixels existing in the vicinity of the specific pixel is a background pixel, the process of making the specific pixel a boundary pixel of the figure is performed as described above. An outer boundary pixel column and an inner boundary pixel column are obtained by performing nine pixels each consisting of a specific pixel and eight pixels existing in the vicinity of the specific pixel in a fixed order, and determining the outer boundary pixel column and the inner boundary pixel column. The outer boundary pixel row and the inner boundary pixel row of the figure are traced so that they can be distinguished from each other, and the vertical and horizontal vectors, which represent the direction based on the 8 specific pixels existing in the vicinity of the specific pixel, with the specific pixel as the starting point, and the horizontal direction By arraying eight vectors, vectors and diagonal vectors,
An image processing method that represents the relationship between adjacent pixels and processes the image. (2) Tracing the outer boundary pixel row and inner boundary pixel row of the figure so that they can be distinguished from each other, and in a direction based on eight specific pixels existing in the vicinity of the specific pixel. In an image processing method in which the relationship between adjacent pixels is expressed and image processed by an array of eight vectors: a vertical vector, a horizontal vector, and a diagonal vector, which represent a specific pixel as a starting point, is the direction in which one of the horizontal direction vector and one of the vertical direction vector exists before and after the diagonal direction vector, and when these three vectors have a common starting point, the direction is opposite to the direction of tracing the outer boundary pixels by a predetermined acute angle. The diagonal vector of the part where the direction changes is broken down into vectors that exist before and after the diagonal vector. The diagonal direction vector of the part where one of the direction vectors exists and the direction changes in the same direction as the direction tracing the outer boundary pixel by a predetermined acute angle when the starting point of these three vectors is made common is defined as the diagonal direction. An image processing method characterized by decomposing and representing a vector into vectors that exist before and after a vector. (3) Claim (1) or (2) characterized in that the eight vectors are represented by codes, and the relationship between the adjacent pixels is represented by a chain code consisting of a code string.
Image processing method described.