JP3224138B2 - Kanji font generation method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for generating a kanji font, and more particularly, to generate a drawing outline functionally by giving a drawing (kaku) geometric parameter, and as a set of drawing outlines. It relates to a method for generating a kanji font.

[0002]

2. Description of the Related Art Kanji has an enormous character type and has a complicated pattern, so the design work is extremely large, and an extremely large amount of memory is required to store the kanji pattern as bitmap data. And In particular, in order to obtain high-quality kanji fonts, the amount of required memory is enormous.

Conventionally, when a high-quality kanji font is required, an outline method is used in which the original kanji character is converted into a bitmap and then converted into an outline vector by general binary image processing.

[0004]

However, the conventional outline method is only a sub-optimal means for reducing the amount of unrealistic memory required for storing bitmap data. There is much room.

A characteristic of a kanji pattern is that it is expressed as a set of unit elements called an image. In the conventional outline method, when a contour vector is formed, a line segment is generated due to interference (intersection, contact) between the images. , The outline data contains many redundant and harmful vectors, and there is a limit to the reduction of the data amount and the memory amount. In the conventional outline method, partial correction of a kanji font was practically impossible, and lacked editability.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a kanji font generation method capable of greatly reducing the amount of data for kanji font generation and having excellent editability.

A kanji is not simply a set of strokes, but is constructed in units of neighboring elements, which are subsets of strokes. By utilizing the hierarchical structure of the side-by-side picture, a great reduction in memory and improvement in operability of font editing can be expected. However, there is also a problem in a method of composing a font in units of sidelines and pictures. For example, when combining kanji, affine transformation of pattern elements is necessary, but if affine transformation is simply applied side by side, distortion due to vertical and horizontality occurs, and it is not possible to combine well-balanced natural fonts There's a problem.

FIG. 19 shows an example of synthesizing "righteousness" using "I" which is adjacent. Reference numeral 400 denotes an original “I” pattern, reference numeral 401 denotes a pattern obtained by simply performing an affine transformation on the entire pattern 400, and reference numeral 402 denotes a target “righteousness” pattern. In the pattern 401, distortion is caused by horizontal and vertical degrees, such as a thick top and a narrow bottom, as seen in the "crossing".

[0009] Such pattern distortion is reduced 402.
To synthesize a font pattern like
It is only necessary that the outline of each image can be independently corrected.
However, such correction is difficult with the conventional outline method. Therefore, in the conventional outline method, it is a fact that the original characters of all kanji must be designed.

[0010] Another object of the present invention is to provide a kanji font generation system capable of effectively utilizing a hierarchical structure of side-by-side images, and to provide a kanji font generation system capable of parametric editing of a kanji font. is there. It is another object of the present invention to provide a kanji font generation method that can easily generate kanji font patterns of various styles, fonts, and sizes using data of a reference font.

[0011]

A feature of the present invention resides in that the outline of an image is defined as one closed loop, and four types of geometric parameters, that is, strokes corresponding to a starting point, a bending point, or an ending point on a stroke. The position of the end point, the accent edge corresponding to the outline part of the emphasis point on the stroke, the stroke edge which is the outline part corresponding to the trajectory on the stroke excluding the accent edge, and the line corresponding to the interval at the beginning and end of the stroke edge It is defined by the diameter, the outline of the image is generated functionally based on these four types of geometric parameters, and a kanji font is obtained as a set of outlines of the image.

A feature of the second aspect of the present invention is that, in the kanji font generation method according to the first aspect of the present invention, a certain type of accent edge pattern is prepared, and the accent edge is designated by an index number assigned to each pattern. is there.

A feature of the third aspect of the present invention is that, in the kanji font generation method according to the first aspect of the present invention, a certain type of stroke edge pattern is prepared and the stroke edge is designated by an index number assigned to each. is there.

According to a fourth aspect of the present invention, in the kanji font generating method according to the first aspect of the present invention, one of the stroke edges is generated functionally from the other edge.

According to a fifth aspect of the present invention, in the kanji font generation system according to the first aspect of the present invention, the coordinates of the stroke end point are defined only for the reference typeface, and the coordinates of the stroke end point of another typeface are defined as a displacement vector with respect to the coordinates. Is to define

According to a sixth aspect of the present invention, in the kanji font generation system according to the first aspect of the present invention, a side element is synthesized as a set of outlines of an image, and at the time of this synthesis, only the stroke end points of the outline of the image are added. To perform the conversion.

[0017]

According to the first aspect of the present invention, it is possible to generate a parametric outline.
Since the four types of geometric parameters defining an image are independent of each other, a certain parameter can be corrected without causing mutual interference with other parameters. For example, without changing the position and size of a stroke, It is possible to correct only the wire diameter, and the kanji font can be easily edited and processed. On the other hand, in a non-parametric method in which a stroke loop is represented by a geometric point sequence and a curve type, it is difficult to specify a point sequence at which the position of a stroke should be changed.

Further, one image is divided into one data set and one data set.
It can be expressed by a combination of individual drawing routines, which means that the font can be set by the drawing routine. The meanings of the four types of geometric parameters are common regardless of the typeface. Therefore, by interpreting the parameters, that is, changing the function of the image, it is possible to generate a kanji font having a different style as illustrated in FIG. 20 without changing the parameters.

The accent edges of the drawing outline are stylized traces of movement perpendicular to the paper at the end points and nodes on the brush stroke, and have a large operating range (large resolution) like stroke edges. ) Is not required, and the number of bits of a vector forming an edge can be reduced. Furthermore, since the same type of accent (for example, horizontal scale) is quantitatively approximated for all patterns of all characters, prepare patterns with a small number of ranks and select the required pattern from among them. Can be used. Therefore, according to the method of designating the accent edge by the index number of the pattern prepared in advance, the data amount regarding the accent edge can be effectively reduced.

The same applies to stroke edges. A pattern can be selected from predetermined types of patterns by an index number, and a stroke edge of the selected pattern can be functionally generated based on the position of the stroke end point. As a result, the data amount related to the stroke edge can be reduced.

The stroke edge of the image outline represents a real stroke of the stroke, and the resolution of the vector must be able to cover the entire area. However, the stroke edge is a stylization of horizontal movement without vertical movement of the brush,
It can be represented by two edges having a uniform interval, that is, a constant interval or a gradual change. Therefore, it is easy to define one of the stroke edges as a real vector and generate the other edge functionally.

According to the method in which one of the stroke edges is generated functionally from the other edge, the amount of data can be reduced, and when both of the two edges are defined by a real vector, there is a concern. The risk of damaging the uniformity of both edges can also be eliminated.

According to the fifth aspect of the present invention, one character type is defined by one qualitative image structure.
Differences in style are only quantification differences for this image structure. If there is a parameter that expresses the image structure like the stroke end point parameter, one
The positions of the stroke endpoints are distributed within a limited range between characters of different styles. FIG. 21 shows how the stroke end points are dispersed.

Therefore, if the stroke end point of a certain typeface as a reference is defined by a real vector, the stroke end point of another typeface can be defined by a displacement vector with respect to the stroke end point of the reference typeface. The resolution of the displacement vector may be smaller than that of the real vector. Therefore, even when many types of fonts are required, the data amount and the memory amount can be considerably reduced.

Since the kanji has a side-by-side hierarchical structure, the side by side elements can be combined as a set of outlines of the stroke. In this case, it is considered that all common Chinese characters can be combined by combining 28 types of image outlines as shown in FIGS.

More specifically, by expanding these 28 types of image outlines, preparing all of the sidelines (about 120 types), and specifying a size and a position in these fonts, a data amount and a memory amount can be obtained. Decreases dramatically. By performing affine transformation only on stroke end points without performing affine transformation on the entire image outline at the time of this synthesis, pattern distortion that cannot be ignored in font quality as described with reference to FIG. 19 can be removed. .

The degree of freedom of the shape is limited by limiting the affine transformation to only the end points of the stroke (for example, all words have a degree of freedom in position and size, but all have similar shapes). However, if this is considered as a kind of design item, it is not particularly disadvantageous, but rather has a great advantage that the labor and the amount of memory for font creation can be drastically reduced.

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the drawings.

First, the structure and terms of the image outline will be specifically described with reference to FIG. The outline of the image of “left pay” shown in (a) is composed of a pair of stroke edges 100 and 101 and one accent edge 102. 100 is called the main side of the stroke edge, and 101 is called the sub-side. 103 and 104
Are the start and end points of the stroke end points. 105 is the main edge 1
00 and the start point 106 of the accent edge 102 are the end points of the accent edge 102, but at the same time
It is also the starting point of w is a wire diameter.

The outline of the "vertical line" image shown in FIG. 3B is composed of the main side 110 and the sub-side 111 of the accent edge,
Consists of accent edges 112. Reference numerals 113 and 114 denote start and end points which are stroke end points, reference numeral 115 denotes a start point of the main side 110 and the accent edge 112, and reference numeral 116 denotes a start point of the sub-side and an end point of the accent edge 112.

The outline of the "key" image shown in (c) is composed of the main side 120 and the sub-side 12 of the first stroke edge.
1. Major side 122 and minor side 12 of second stroke edge
3. It is composed of a first accent edge 124 and a second accent edge 125. In this case, the stroke end point includes a middle point (bending point) 128 in addition to the start point 126 and the end point 127.

Next, the image parameters will be described with reference to the "left-payment" image (black-painted) used in "gi" shown in FIG. As can be understood from FIG. 2, even if the same “leftward payment” is used, the inclination, length, thickness, and the like are slightly different. The image outline is defined using the dynamic parameters. 3 and 4 are explanatory diagrams thereof.

In FIG. 3, P0 is the stroke start point, P
2 is the stroke end point, P1 is the start point of the main side and the accent edge, and P3 is the start point of the sub side and the end point of the accent edge, as described above. C is the distance from the midpoint of the straight line connecting P0 and P2 to the point at which the perpendicular extending in the clockwise direction intersects the main side, which corresponds to the curvature of the main side. w
Is the wire diameter.

Actually, as described later, a certain type of stroke edge pattern (function) is prepared, the pattern is designated, and the curvature C is designated (the designation of the curvature C may be omitted). Possible), the main side can be generated functionally from the stroke end points P0 and P2 and the height of the accent edge (described below) specified separately.
The sub-side can be functionally generated from the main side and a separately specified wire diameter w.

FIG. 5 shows an example of such a previously prepared stroke edge pattern. By using the patterns S1 to S8 shown here, it is considered that almost all stroke edges (main sides) of an image used for a general kanji can be generated.

FIG. 4 shows an accent edge pattern, in which the direction is normalized. P0 and P
3 is the start and end points of the accent edge, which is shown in FIG.
P1 and P3. P1 corresponds to P0 in FIG. P2 is a point farthest from the start point P1 of the accent edge.

When a high degree of freedom is recognized in the accent pattern, the accent edge pattern is defined by three numerical parameters of the height m of the accent edge, the line diameter w, and the vector v (dx, dy) from P3 to P2. I do.

Actually, as will be described later, a certain type of accent edge pattern (function) is prepared in advance, a pattern is selected, and the pattern is scaled and rotated so as to be correctly continued to the stroke edge. Thus, a necessary accent edge can be generated.

FIG. 6 shows an example of such an accent edge pattern prepared in advance. By using the patterns A1 to A8 shown here, it is considered that almost all accent edges of an image used for a general kanji can be generated.

FIGS. 7 to 9 show 28 image outlines that can be generated by combining the eight stroke edge patterns shown in FIG. 5 and the accent edge patterns shown in FIG. Si and Ai in the figure are the same as the pattern numbers (names) in FIG. 5 or FIG. Ki is an image pattern number (name). The broken line is a sub-side of the stroke edge. In the case of a certain font, all the kanji fonts can be generated by the above 28 types of image outline patterns.

FIG. 10 shows an example of an outline generator for generating an image outline according to the present invention. 200 is a stroke generator, 201 is an accent generator, and 203 is an affine transformation unit.

The stroke generator 200 is a set of routines for functionally generating eight types of stroke edges (line segment vectors) shown in FIG. It has a thickness function table 204 (the contents will be described later) which is referred to when a sub-edge is functionally generated from a main edge of a stroke edge.

The accent generator 201 is a set of routines for functionally generating eight types of accent edges (line segment vectors) shown in FIG.

The pattern generated by each of the generators 200 and 201 is a set of points required to draw an outline pattern, for example, five points P0 to P4 in FIG. 11 or points obtained by interpolating these points with spline curves. Group. The important point is that these outline arrays have a coordinate system independent of the character coordinates, the starting point (P0 in FIG. 11) is the origin (0, 0), and the given parameters (stroke end point, accent, line diameter) , Rotation, enlargement / reduction (vertical / horizontal), and position movement, and drawing an image pattern.

The selection of the routine (pattern) of each of the generators 200 and 201 is designated externally according to the target image. For example, when generating the outline of "vertical stop" (k1), the routine of A1 and A4 is selected for the accent edge, and the routine of S3 is selected as the stroke edge.

The affine transformation unit 203 is provided for performing affine transformation on the position of the stroke end point as required when combining a character or a character from an image outline.

Next, "left payment" (K20) is taken as an example.
The process of generating an image outline will be specifically described. FIG. 12 is an explanatory diagram thereof.

As can be seen from FIG. 9, this picture outline is composed of the stroke edge pattern of S5 and the accent edge pattern of A1, and the following parameters are designated externally.

First, a pattern number (name), that is, a routine number (name) is used as a parameter related to stroke generation.
There are S5 and P0 and P1 positions (stroke end points) shown in FIG. This position is a character coordinate (for example, 256 × 25
6) Specified as absolute coordinates above. Also, the curvature C is specified. However, if it is acceptable to sacrifice the degree of freedom and naturalness of the shape of the stroke edge, it is not impossible to set C as a default and not specify it from the outside.

The parameters related to accent edge generation include a pattern number (name), that is, a routine number (name) A1, a pattern height m, and a vector v from the end point to the farthest point. Note that, similarly to the curvature, it is not impossible to set m and v as defaults. Since the wire diameter w is also used for generating an accent edge, it is also input to the accent generator 201. The generation of the outline is performed as follows.

The stroke pattern of S5 is selected. FIG.
As shown in (a), the position of the midpoint P2 of the straight line connecting P0 and P1 is calculated. Next, a perpendicular line is drawn in a clockwise direction from P2, and a point P2 ′ separated by the curvature C is calculated. Then, spline interpolation is performed for three points P0, P2 ', and P1 to generate a point group P0-P1. This point group constitutes a part of the accent edge and the main side.

Select the accent edge pattern of A1.
As shown in FIG. 12B, the size of the pattern A1 is adjusted so that the height is m and the interval between the start point P3 and the end point P4 is equal to the wire diameter w. The degree of the bulge of the pattern is determined by the vector v.

Next, as shown in FIG.
The point P0 ′ on the point sequence obtained by the step P, which is separated from P0 by m, is set as the starting point of the main side, and the difference d in the horizontal and vertical directions between P0 and P0 ′
The inclination of the main side is calculated from x and dy. (D), the accent edge pattern, tilts by the tilt and the same amount of primary edges. Next, as shown in (e), the start point P3 of the accent edge pattern is moved to the start point P0 ′ of the main side,
Connect the accent edge pattern to the main edge.

Next, as shown in FIG.
The interval from 0'-P1 is divided into four equal parts (or three equal parts), and a vertical line is drawn clockwise from each division point. Length w of each perpendicular
1, w2 and w3 are calculated according to the thickness function table corresponding to the pattern S5 (see FIG. 13). Then, the tip position of each perpendicular, the end point P4 of the accent edge (the start point of the sub-side),
The sub-side is generated by obtaining a point group connecting the end point P1 of the main side (the end point of the sub-side) by spline interpolation.

Here, the thickness function table is a numerical sequence representing the thickness at each point on the stroke edge as a ratio to the wire diameter W. The thickness such as “pay-left”, “pay-right”, “clamp-right”, etc., as described above, changes non-linearly along the stroke progression. Therefore, by creating a “thickness function table” suitable for the design item, the thickness can be easily controlled. Also, there is a greater degree of freedom in thickness control as compared to the method of defining the thickness with a function.

As described above, the outline of the image can be generated. By combining the outline on one character coordinate, the outline of the sideways or Chinese character can be obtained. In this case, there are two types of data formats for generating a kanji font using the image outline generator.

The first format is a format in which each character is represented as a set of a series of strokes constituting the character. In this case, for example, the data for “I” has the content as shown in FIG. In this format, the parameters of each image can be specified in units of individual characters, so that subtle differences in outlines can be expressed. However, from the viewpoint of reducing the amount of memory for storing data, it is disadvantageous compared to the second format described below.

The second format is to define a set of strokes in units of preselected collaterals or characters, and to obtain the desired outline by combining the collaterals or characters constituting the target character by affine transformation. It is. In this data format, it is sufficient to read out the predefined array or the character arrangement and the affine transformation parameters (position and size) and provide them to the image outline generator, so that the amount of memory required for data storage can be extremely reduced.

For example, the image shown in FIGS.
By preparing about 20 kinds of “nearby”, practically all kanji fonts can be generated.

FIG. 15A shows data in the second format for generating "right". As the affine parameters, as shown in FIG.
A height V and a width H are specified.

In the case of the second format, an affine parameter is input to a picture outline generator together with a picture parameter of a side or a character which is a part of a target kanji. In the image outline generator, the affine transformation unit 203 performs affine transformation on stroke end points in the image parameters according to the affine parameters,
The stroke end point after the conversion is used by the stroke generator 200. Other than this is as described above.

As described above, since the affine transformation is applied only to the stroke end point and other parameters are not affected,
19, it is possible to avoid the occurrence of distortion due to the horizontality and verticality of the stroke as described above. That is, the affine transformation is applied only to the end point of the image stroke of the “I” pattern 400 in FIG.
That is, a distortion-free pattern very close to 02 can be obtained.

Next, an example of a specific hierarchical data structure based on the hierarchical structure of Chinese characters will be described. FIG. 16 is an explanatory diagram of this data structure. In this data structure, data is hierarchized by a character table 300, a side table 301, and an image (kaku) table 302.

The drawing table 302 is composed of individual subordinates (or characters).
Are set at the bottom of the hierarchy. The sidelines table 301 is a set of sidelines constituting a character, and is located in the middle layer of the hierarchy. Since a character can be an element of another character, the character table 300 can be referred to from the side table 301. The character table 300 is located at the top of the hierarchy, and is a table indicating a by-side set or a stroke set corresponding to a character selected from outside.

The contents of each table are described as "word", "right", "gi",
FIG. 17 shows the four characters "I" as an example. (A) Character table 3
00B, (b) shows the contents of the side table 301, and (c) shows the contents of the image table 302.

One record of the character table 300 and the side table 301 has a length of 2 bytes and has a structure shown in FIG. 18, and the lower 13 bits are an index for a table of the type indicated by the upper 3 bits. One record of the picture table 302 is 1 byte long and represents the number of one picture.

When the font of “I” is generated, the record of the address 89E4 corresponding to the character table 300 is the image table 3
02 indicates an address r. Therefore, the parameters (including affine parameters) of each image constituting "I" are sequentially read from the parameter memory according to the contents of the records from the address r of the image table 301 to the position before the NULL record, and input to the image outline generator. By generating an outline of the above and synthesizing the outline on the same character coordinates, a desired font can be obtained.

The same applies to the case of “word”, and the image table 30 indicated by the record of the corresponding address 8CBE of the character table 300 is displayed.
The image parameters are read in accordance with the contents of each record from the address p to the position before the NULL record, and an outline is generated.

In the case of “right”, the records from the address m of the side table 301 indicated by the corresponding record of the address 8B60 in the character table 300 to the position before the NULL record are referred to. First, based on the contents of the record below address q of the image table 302 pointed to by the record at address m of the side table 301, the parameters of the image forming the upper side are read from the parameter memory and given to the image outline generator, and the outline is generated. . Next, the side table 301
Of the character table 300 indicated by the record of the address m + 1 of the address 89E4. Since this record points to the address r of the picture table 302, the parameters of the picture are read out according to the contents of the record below this address and given to the picture outline generator, whereby the picture of Generate an outline. The same applies to "gi".

As described above, by changing the drawing routine of each of the generators 200 and 201, a kanji font having a different typeface as shown in FIG. 12 can be generated using the same image parameters.

Also, the stroke end point is prepared as a real vector for the reference font, and for the other fonts,
A displacement vector with respect to the stroke end point of the reference typeface is prepared, and the actual vector of the stroke end point of another typeface is obtained from the displacement vector and the stroke end point of the reference typeface.
It will be understood from the above description that the image outline can be generated by giving this to the image outline generator. As described above with reference to FIG. 21, this is effective in reducing the amount of memory.

[0072]

According to the first aspect of the present invention, it is possible to parametrically generate an image outline, and to correct only the line diameter without changing the position and size of a stroke. Editing and processing of kanji fonts are facilitated, and kanji fonts having different styles can be easily generated without changing parameters.

According to the second and third aspects of the present invention, it is possible to effectively reduce the amount of data relating to accent edges and stroke edges in an image outline.

According to the fourth aspect of the present invention, it is possible to obtain the effect of reducing the amount of data related to the stroke edge and to eliminate the risk of impairing the uniformity of the two stroke edges forming a pair.

According to the fifth aspect of the present invention, even when many types of fonts are required, the amount of data and the amount of memory can be considerably reduced as compared with the method of facilitating data independently for each typeface.

According to the invention of claim 6, it is possible to reduce the distortion of the outline due to the affine transformation required when the data amount is reduced by utilizing the hierarchical structure of the kanji sidelines and strokes.

[Brief description of the drawings]

FIG. 1A is an explanatory diagram of an outline of “left payment”. (B) It is explanatory drawing of the outline of a "vertical line."
(C) It is explanatory drawing of the outline of "key."

FIG. 2 is a diagram showing “left-payment” used in “giken”;

FIG. 3 is an explanatory diagram of a stroke parameter of “pay left”;

FIG. 4 is an explanatory diagram of an accent parameter of “left payment”.

FIG. 5 is a diagram showing a stroke pattern.

FIG. 6 is a diagram showing an accent pattern.

FIG. 7 shows an image outline obtained by combining the patterns of FIGS. 5 and 6;

FIG. 8 shows an image outline obtained by combining the patterns of FIGS. 5 and 6;

FIG. 9 shows an image outline obtained by combining the patterns of FIGS. 5 and 6;

FIG. 10 is a block diagram illustrating an example of an image outline generator.

FIG. 11 shows an explanatory pattern of a generator.

FIG. 12A is an explanatory diagram of a main edge and generation. (B) It is explanatory drawing of size adjustment of an accent edge pattern.
(C) It is explanatory drawing of the inclination in the starting point of a main side. (D) It is explanatory drawing of rotation of an accent edge pattern. (E) is a diagram showing the connection of the accent edge to the main side. (F)
It is explanatory drawing of generation of a sub-edge.

FIG. 13 is an explanatory diagram of a thickness function table.

FIG. 14 shows one format of data for generating “I”.

FIG. 15A shows one format of data for generating “right”. (B) It is explanatory drawing of an affine parameter.

FIG. 16 is an explanatory diagram of a hierarchical data structure.

FIG. 17A shows the contents of a character table. (B) Shows the contents of the side table. (C) Shows the contents of the image table.

FIG. 18 is a diagram showing a data structure of a character table and a side table.

FIG. 19 is a diagram for explaining pattern distortion due to affine transformation.

FIG. 20 is a diagram illustrating an example of typeface conversion using an image function.

FIG. 21 is a diagram illustrating dispersion of stroke end point positions due to differences in style.

[Explanation of symbols]

 200 Stroke generator 201 Accent generator 203 Affine conversion unit 204 Thickness function table 300 Character table 301 Sideline table 302 Picture table

Claims (6)

(57) [Claims] An outline of an image is defined as one closed loop, and stroke ends corresponding to a start point, a bending point or an end point on the stroke, an accent edge corresponding to an outline portion of a power point on the stroke, and a stroke on the stroke excluding the accent edge are removed. Are defined by four geometric parameters: a stroke edge, which is an outline part corresponding to the trajectory, and a line diameter corresponding to the interval at the start or end of the stroke edge. Based on these four geometric parameters, a functional A kanji font generation method characterized by generating an outline of a picture functionally and obtaining a kanji font as a set of outlines of the picture. 2. The kanji font generation method according to claim 1, wherein a certain type of accent edge pattern is prepared, and an accent edge is specified by an index number assigned to each pattern. 3. The kanji font generation method according to claim 1, wherein a predetermined type of stroke edge pattern is prepared, and the stroke edge is designated by an index number assigned to each. 4. The kanji font generation method according to claim 1, wherein one of the stroke edges is generated functionally from the other edge. 5. The kanji font generation method according to claim 1, wherein coordinates of a stroke end point are defined for a reference typeface, and coordinates of a stroke end point of another typeface are defined as a displacement vector with respect to the coordinates. 6. The kanji font generation method according to claim 1, wherein bye components are synthesized as a set of outlines of the image, and in this synthesis, affine transformation is performed only on stroke end points of the outline of the image.