JP2656203B2 - Method for analyzing shape characteristics of figures, etc. - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to graphic processing, and more particularly to a method for analyzing the shape characteristics of various characters, symbols or graphics.

[0002]

2. Description of the Related Art When printing various symbols, designs, logos, etc. on printed matter for distribution and display, or attaching them to a structure such as a building or a signboard, a mechanical process in manufacturing is necessary. , You must know the geometric properties of each. For example, in the case of printing, various considerations and processes are required for space distribution on paper, harmony and distribution with other characters, symbols, and designs used together or in combination. In addition, if it is to be installed on exhibitions and structures, it is necessary to perform basic design concerning the use amount, occupied space, weight, etc. of various materials, so numerical values indicating various geometric properties are required for that purpose .

[0003] Although symbols, characters, and other figures may be used alone, most are used side by side or in combination with other symbols and characters. The characters and symbols of the established series (typeface) have been tried and put to practical use through countless combinations (mainly within the series) over the years.
It is known that they are in harmony with each other. Therefore, when drawing or creating a new symbol or a new typeface, it is necessary to spend a great deal of effort and time to determine harmony in consideration of all combinations and arrangements. When combining with characters and symbols of other series, it is necessary to know the characteristics of the shape of each character or symbol.

[0004]

As described above, the arrangement and adjustment of characters, symbols, figures, and the like, whether they are print manuscripts, exhibits such as posters, or structures, must first be sufficiently performed on a desk. Need to be done. Therefore, it is preferable that the geometrical properties of the innumerable existing characters and symbols, and the designs to be prototyped, are expressed as unique indices independent of the scaling ratio. Furthermore, not only the difference in the enlargement / reduction ratio, but also a universal index that can be used freely between different series of characters, symbols, and designs is required. The present invention has been made in order to solve such a problem, and provides a group of indices having a universality unique to a figure independent of size or series, and using this to objectively characterize shape characteristics of a figure or the like. An object of the present invention is to obtain a shape characteristic analysis method that can be analyzed.

[0005]

According to a first aspect of the present invention, there is provided a method for analyzing a shape characteristic of a graphic or the like according to the present invention. a database seeking index four items, using all parts of these indices, letters, symbols or fabrication of the graphic, sequence, characterized in that to perform design, shape characterization for differential .

Aspect ratio FX / FY Expression (1) Area ratio A / (FX.FY ) Expression (2) Dose ratio 2APM / (FX + FY) Expression (4) or 2 PM/ (FX + FY) Expression (5) Fine ratio LNG / WA Expression (6) or (LNG / WA) × (PM / APM) ² Expression
(7) where A is the actual area of the object, FX is the Feret diameter in the X direction (distance between two vertical tangents), FY is the Feret diameter in the Y direction (distance between two horizontal tangents) ) , A PM is the perimeter of the object, PM is the sum of the perimeters of the object (sum of the perimeter of the object and the perimeter of the internal hole), and LNG is the length (= APM /
2), WA is the average width (= 2A / APM ) .

According to a second aspect of the present invention, in the method of analyzing a shape characteristic of a graphic or the like according to the first aspect, for a character, a symbol or a graphic composed of a plurality of separated elements, the plurality of elements are connected by a thin line. Then, the shape characteristics are analyzed after being converted into one character, symbol or figure.

[0008]

According to the shape characteristic analysis method for a graphic or the like according to the first aspect of the present invention, the size of a character, symbol, or other graphic to be analyzed (hereinafter simply referred to as a graphic or the like) is similar if the shape is similar. based on the exponent value of the four items having a practical problem no accuracy of repeatability regardless to the analysis of the shape characteristics is performed using the entire part.

In the method for analyzing shape characteristics of a graphic or the like according to the second aspect of the present invention, even if the graphic or the like is composed of a plurality of separated elements, the shape characteristic is analyzed as being regarded as one graphic or the like. .

[0010]

Embodiments of the present invention will be described below in detail.

[0011] In this case the determination of the specific index is described first method of determining the index value having practical problem no precision of repeatability for use in shape characterization.

As a method of simply and quickly geometrically measuring a figure and providing various types of numerical information on the figure, image processing by a computer is usually used. It varies depending on the size of the sample used for analysis. Therefore, in order to perform comparison, extraction, and the like between figures having different sizes, a pattern obtained on the basis of the original values (ie, the original values) rather than the numerical values obtained by image processing by a computer (hereinafter, referred to as original values) is used. It is necessary to use a numerical value unique to a figure shape. Therefore, the unique numerical value (hereinafter, referred to as a unique index) must have reproducibility regardless of the scale of the symbol and the system of the symbol, and have a precision that can withstand practical use.

Therefore, in the present invention, a large number of original values are obtained by using the patterns and characters of each system as materials, and from each of the indices obtained based on these values, a unique index having practically accurate and reproducible values is obtained. Work was done to determine the index. In particular,
20 to 30 times (one-dimensional magnification) with the starting point as 1
The images were magnified sequentially, and the images at each magnification were processed by computer image processing to calculate the respective indices from the original values, and work was carried out to find an eigen index with the accuracy and reproducibility that can withstand practical use.

Since image processing by a computer uses the mechanism of an optical system, the mechanism of electronics, and the like, there is a limit to the resolution of small and complex symbols. When the display uses, for example, 520 pixels in the X coordinate and 480 pixels in the Y coordinate, even if the resolution of the lens system is increased, the number of pixels in the X or Y direction on the display is 80 pixels or less. It is not preferable to analyze by size, and especially when the number is about 10 to 20 pixels, the numerical value becomes inaccurate. This is because the details tend to be crushed and rounded, and the line length tends to be small. Therefore, adjustment was made so that the measurement could be performed with at least a size of 80 pixels or more, and among the indices obtained by variously changing the measurement conditions, those whose variation from each average value was within ± 3% were accepted. . The superior one exhibited reproducibility with an accuracy of about 0.2%.

When there are about 50 original numerical values, the combination calculated numerical values (candidates of the unique exponents) considered from these original numerical values are 200 to 300 or more. From these results, it was concluded that the following nine types of indices are important as proper indices that meet the above-described condition of accuracy and are suitable for expressing the geometric properties of the shape. The calculated values other than these are ± 10
There were not a few indices showing large errors of more than%.

[0016] ^{(1) FX / FY (2} ) A / (FX · FY) (3) CYV / (FX 2 · FY) (4) 2APM / (FX + FY) (5) 2 PM/ (FX + FY) (6) LNG / WA (7) (LNG / WA) × (PM / APM) ² (8) (CGX-X) / FX (9) (CGY-Y) / FY Next, the original numerical values used in the calculation of the index group Will be described. In general, there are three types of numerical values (original numerical values) obtained by computer image processing: binary measurement values, black and white ratio measurement values, and hand measurement values. In the present invention, among these, the following four values are used as the original values obtained by the binary measurement.
Five types were adopted, and the black-and-white ratio and manual measurement were performed as necessary. In the following original numerical value groups, those relating to length are expressed in units of pixels.

1. A [Actual area] (FIG. 4): The actual area of the object.

2. 2. NHL [number of holes] AIH [area including hole] (FIG. 4): This is the area including the hole of the object having the hole.

4. APM [perimeter of object] (FIG. 5).

[5] PM [sum of perimeters] (FIG. 5): Sum of the perimeter (APM) of the object and the perimeter of the hole (HPM) (A
PM + HPM).

6. X [minimum X coordinate]: In the case of a small computer, for example, 0 ≦ Y <512 (pixels).

7. Y [minimum Y coordinate]: For a small computer, for example, 0 ≦ Y <480 (pixel).

8. FX [X-direction Feret diameter]: a distance between vertical lines when an object is sandwiched between two vertical lines (FIG. 6), for example, 1 ≦ FX ≦ 512. In addition,
"Ferret" will be described later.

9. FY [ Y- direction Feret diameter]: Distance between horizontal lines when an object is sandwiched between two horizontal lines (FIG. 6), for example, 1 ≦ FY ≦ 480.

10. CGX [X-axis gravity center coordinate]: For example, 0 ≦ CGX ≦ 512.

11. CGY [Y direction gravity center coordinate]: For example, 0 ≦ CGY ≦ 480.

12. HD [circle equivalent diameter] (FIG. 7): The diameter of a circle having an area equal to the area of the object.

13. RND [Roundness]: RND = 4π
A / (APM) ² .

14. LNG [length]: LNG = (AP
M) / 2.

15. WA [average width]: WA = 2A / A
PM.

16. ODL (long axis corresponding to ellipse) (Fig. 8)
: The length of the major axis of the ellipse obtained from the moment of inertia around the center of gravity.

17. ODS (short axis equivalent to ellipse) (FIG. 8) OLS [tilt of elliptical long axis] (FIG. 8): 0 ° ≦
OLS ≦ 180 ° 19. OSF [elliptical long axis ratio]: OSF = ODL / O
DS 20. LG [Projection length to long axis] WD [projection length to short axis] 22. PMH [Envelope circumference] (FIG. 9): This is the circumference when connecting the convex portions of the object.

23. ACV [convex hull area] (FIG. 9):
This is the area of the part enclosed by the envelope. 24. ML [maximum absolute length] (FIG. 10): The maximum length of the distance between any two points on the circumference of the object.

25. MW [Maximum absolute width] (Fig. 10)
: The distance between two straight lines when the object is sandwiched between two straight lines parallel to the absolute length ML.

26. MSL [ Slope of maximum absolute length] (Fig. 1
0): An angle between the X axis and the direction of the absolute maximum length ML, where 0 ° ≦ MSL ≦ 180 °.

27. WAX [average width in X direction]: Average value of the length of the cross section in the X direction.

28. CYV [cylindrical volume] (Fig. 11):
Assuming that the length (X direction) of each cutting line when the object is cut at a minute interval Δd in the Y direction is represented by the following equation.

[0038]

(Equation 1) 29. EDH (horizontal equal diameter) (FIG. 12): The length of a cut line cut by a horizontal line that bisects the area.

30. EDV (vertical equal diameter) (Fig. 12)
: The length of an incision cut by a vertical line bisecting the area.

31. NCG (shortest center of gravity distance) 32. NOB (shortest particle label): a label of an object to which NCG is given.

33. CRL [maximum distance from center of gravity] CRS (minimum distance from center of gravity) 35. CRA [average distance from center of gravity] CR2 [mean square radius]: expressed by the following equation.

[0042]

(Equation 2) 37. RLT (direction of maximum distance from the center of gravity): 0 ° R
LT ≦ 360 ° 38. RST (minimum distance direction from the center of gravity): 0 ° ≦ R
ST ≦ 360 ° 39. LC1 [four-direction diameter 1 passing through the center of gravity] (FIG. 13) LC2 [four directions diameter 2 passing through the center of gravity] (FIG. 13) LC3 [four directions diameter 3 passing through the center of gravity] (FIG. 13) LC4 [four directions diameter 4 passing through the center of gravity] (FIG. 13) THL [maximum thickness in eight directions] (FIG. 14): This is the maximum value of the thicknesses TH1 to TH8 in each of the eight directions.

44. THS [minimum thickness in 8 directions] (Fig. 1
4): The minimum value of the thicknesses TH1 to TH8 in each of the eight directions.

45. THA [8 direction average thickness] (FIG. 1
4): Average value of the thicknesses TH1 to TH8 in each of the eight directions.

The above original value group is an example of an image processing process by a widely used computer, but another process similar to this can be applied. These processes have their uses and characteristics, and there are some differences in the types of calculated values, but they can be applied to any of the processes.

Although the above-mentioned original numerical value group generally differs in numerical value depending on the size of the analysis sample, some of the original numerical value group (for example, 2: NHL, 13: RND, 18: OL)
S, 19: OSF, 26: MSL, 37: RLT, 3
8: RST), which are ratios to other original values and those indicating angles should be the same if the shapes of the figures are the same. I have it. Therefore, it is also effective to use them together with the above-mentioned nine types of unique indices.

Although various tests were performed as described above for errors caused by the enlargement / reduction, two samples are shown in FIGS. 15 and 16 to show the actual test situation. Here, FIG. 15 shows a kanji character “A” (excerpted from a new square typeface published by Marl), and FIG. 16 shows a mark (mark mark) with the theme of the capital letter “T” of the alphabet.
Symbol-3 Kashiwa Shobo No. 1437). It is considered that the larger sample (( A ) in both figures) can obtain more accurate values than the smaller sample (( B ) in both figures). Was tested.

Next, the contents of nine types of unique indices used in this embodiment will be described.

First, as a first index group, there is an index relating to the area of the figure. The indices belonging to this group are
It indicates what area the object (symbol, character, graphic, etc.) requires on paper or spatially. There are various ideas about that area. For example, in the case of printed characters, it is assumed that the characters are arranged horizontally if they are alphabetical characters, and that they are arranged vertically or horizontally one by one if they are Chinese characters.

Now, a rectangle or a square surrounded by FX and FY shown in FIG. 6 is considered as the area, and this area is temporarily referred to as a Feret area (aspect ratio) . The shape of the feret region is expressed by the following equation. FX / FY (1) Considering the importance in the field of printing and printing, the Feret area shown by the expression (1) is considered to be the most important and has high practicability.
In addition, even from the accuracy of the reproducibility of the measured values, the index indicating this Feret area shows the best result, and the maximum error with respect to the average value in the test on the enlargement / reduction ratio and the sample of various figures is It was ± 0.5% or less.

The real area (shown by the original value A) and the area containing holes (shown by the original value AIH) shown in FIG. 4 can be considered as a kind of region, or the above-mentioned ODL- A rectangle defined by ODS (FIG. 8), a convex hull ACV (FIG. 9), or a rectangle defined by ML-MW (FIG. 10) can also be considered as a region. However, among these, the ML-MW region and the ODL-ODS region have special meanings and may be measured as needed, but are unsuitable for use as universal indices in the reproducibility and accuracy of the numerical values. Met. On the other hand, A, AIH, A
CV is considered to be useful as representing a graphic area. Hereinafter, these original numerical values will be described with an example.

FIG. 17 shows an uppercase letter "A" in the alphabet.
Is represented. The original value A is a value indicating the substantial area (black portion) of the character “A”, and the original value AIH is obtained by adding the area of the hole 51 indicated by the oblique line to the original value A. The original numerical value ACV (convex hull) is the area obtained by adding the concavities 52 to 54 with horizontal lines to AIH. As the area of the character “A”, in addition to the Feret area indicated by the above-mentioned FX-FY, the original numerical values A, AIH, and ACV can be said to have significance, respectively. From the usefulness, the index (area
Ratio) .

A / (FX / FY) (2) Note that AIH / (FX / FY) and ACV / (FX / F)
Y) also has a meaning as a reference index, but in some cases the accuracy is unstable (especially ACV / FX · FY).

The second index group relates to the amount of lines used for forming symbols and figures. When drawing a figure using a line in a predetermined area, the end and end of the line are connected to form a shape, but if the inside is painted black, the internal area of the line will be the original value A Corresponding. However, even if the front end and the end of the line are connected, if the inside is blank, it becomes a large hole. The original value A in this case is only the area of the line around the outer periphery, but the original value AIH is the same as the original value A when the inside is blacked out.

On the other hand, when looking at the dose, APM = PM when blacked out, but PPM when the inside is blank.
M = APM + HPM (FIG. 5 ). Perimeter dose APM
As the number increases in the area where is determined, the shape becomes complicated and the actual area A decreases. Further internal structure dose HPM
When A increases, a hole is formed inside, and A becomes smaller.

That is, the dose has a very large effect on the shape of the figure, and particularly has a very close relationship with the actual area A. Therefore, in order to make the doses APM and PM into indices unique to the figure shape, they are normalized (normalized) by an average value (FX + FY) / 2 of Feret diameters FX and FY as a unit of the figure length. The obtained index (dose ratio) of the following equation was adopted.

APM / [(FX + FY) / 2)] = 2 APM / (FX + FY)... (4) PM / [(FX + FY) / 2)] = 2 PM / (FX + FY) (5) APM and PM The ratio PM / APM is considered to be one of the basic indices that hardly fluctuates depending on various types of scaling of a large number of samples, and is used for calculating an index represented by the following equation (7).

The ferrite region (FX-F) is used as the region.
It has been experimentally confirmed that Y) is optimal and accurate.

As described above, various original numerical values were normalized using (FX + FY) / 2 as a unit of length representing a figure area, and reproducibility and accuracy were calculated. The exponents of the equation and the equation (5) were within an error of ± 1.5 to ± 2.0% with respect to the average value in repeated experiments, confirming their usefulness.

The third index group indicates the relationship between the dose and the actual area A of the figure composed of the dose, and is a combination of the above-mentioned second index group and A. .

As an original value of computer image processing, there is LNG (= APM / 2) as described above. This is half the length of the line forming the periphery of the figure, and is the length when the entire length of the APM is folded in two and stretched to make it straight. The original value WA (average width) is represented by the following equation.

[0062] Therefore, the exponent (fine
(Close ratio) .

LNG / WA (6) The larger the APM, the larger the LNG, and the smaller the actual area A, the smaller the WA. Also, as LNG is larger and WA is smaller, the index shown by the equation (6) becomes larger. That is, the fact that the exponent of equation (6) is large means that
A large number of lines are used and the area cut off is small.

Although this index is based on the APM, the index given by the following equation is adopted in consideration of the total dose PM.

(LNG / WA) × (PM / APM) ² (7) From this equation, when PM / APM = 1, that is, PM =
In the case of APM, Expressions (6) and (7) have the same value. That is, the exponent of the equation (6) relates to the outer circumference, and the index (7)
The exponent of the expression indicates a shape characteristic which should be called the fineness of the figure with respect to the total dose obtained by adding the dose of the outer line and the internal structure. The larger the value, the higher the fineness.

The analysis of the indices of the equations (6) and (7) is as follows.
The test was performed under the condition of 50 to 80 or more pixels. As a result, it was confirmed that the index is an index showing accuracy that is practically usable without being affected by the magnitude of the scaling ratio.

The fourth index group indicates how much the substance of the symbol or figure fills the area.

When viewed two-dimensionally, the ratio of the real area A to the feret region is represented by A / (FX × FY) in equation (2) (see FIG. 18).

By the way, among the above-mentioned original numerical values, there is a numerical value called a cylindrical volume CYV. In this case, as shown in FIG. 19, it is necessary to consider what should be called a three-dimensional ferrite area for storing a figure. Since the size of this three-dimensional ferrite is FX2 · FY, an index of the following equation obtained by normalizing CYV is adopted. Again, this index has been proven accurate by numerous repetitive experiments.

CYV / (FX ² · FY) (3) The fifth index group relates to the center of gravity of symbols and figures. It is clear that the center of gravity is extremely important among the geometric properties. The feret area is used to display the coordinates of the center of gravity. In the case of computer image processing, the upper left corner is the zero point of coordinates, and the X coordinate is taken to the right and the Y coordinate is taken down. The coordinates of the center of gravity of the figure using these coordinates are as follows.

(X coordinate) = CGX-X (Y coordinate) = CGY-Y These numerical values are normalized by FX and FY, respectively, to obtain an index indicating the position of the barycentric coordinate.

(CGX-X) / FX (8) (CGY-Y) / FY (9) If this is shown in the Feret area, for example, FIG.
The center of gravity of the Gilsan Bold Italic capital letter "B" in the feret region shown in FIG. 18B shows various areas in accordance with the display method shown in FIG. 18 with respect to the Feret area of the character "B" in FIG. By enlarging or reducing various characters and figures, the above (8),
The reproducibility of the indices of the formula (9) was examined, and they showed extremely excellent precision.

By the way, when simply obtaining the barycentric coordinates,
A particular problem arises when a figure or the like is composed of a plurality of separate and independent portions (labels). In such a case, the computer image processing calculates the barycentric coordinates, the area, and the like for each label. Then, in order to determine the center of gravity of the entire symbol, it must be calculated using the coordinates of the center of gravity of each label and the actual area (A). For simple shapes or when the number of labels is small, such a method is not a problem.However, in the case of a complex shape or a shape composed of many labels, it is not easy to calculate the center of gravity as a whole. Nearly impossible. Therefore, in this embodiment, a simple method for obtaining a value that does not substantially hinder such a case is adopted.

First, each label of a graphic consisting of four or more labels is connected by a possible combination method using a line as thin as possible and analyzed as one label, and the entire figure obtained for each connection method is analyzed. Is compared with the theoretical value of the entire figure obtained from the analysis value of each label.

For example, FIG. 21 (A) or FIG. 22 (A)
As shown in Fig. 7, consider a figure composed of four or five labels that can be theoretically calculated to determine the respective centers of gravity. FIG.
In No. 1, the method as shown in FIGS.
In FIG. 22, each label is connected by a straight line and analyzed as one label by the method shown in FIGS. 22B to 22G, and each analysis value is analyzed as an analysis value for each label. Is compared with the theoretical calculated value obtained from.

In this case, it is the dose of APM, PM, and the like that changes due to the addition of wires for connection. Although the value relating to the area slightly increases due to the addition of the lines, these values can be easily obtained by summing the values of the respective labels separately. Regarding the center of gravity, make the thickness of the line as thin as possible,
In addition, when the connection is made as short as possible, in each of the examples of FIGS.
The error was within%.

As described above, the error caused by the addition is small because the index used in the present invention does not indicate a simple dose or area but focuses on the ratio. In this way, it has been found that an index calculation can be performed extremely easily by converting a figure or the like having a plurality of labels into one label and then measuring it.

Next, a description will be given of an example in which the above method is applied to kanji in a square style as shown in FIGS. 23 (A) and 24 (A). As shown in these figures, each kanji is composed of a plurality of labels. In such a case (in the case of kanji), the following two connecting methods are particularly common.

(I) The shortest distance between labels is connected (FIG. 24B).

(Ii) Considering the stroke order of kanji or other writing techniques, linking is performed in a manner that is not unnatural even when linking (FIG. 23B).

For example, in the case of the “picture” in FIG. 24B, the following result is obtained, and it can be seen that the center of gravity is slightly lower and rightward.

(CGX-X) /FX=0.527 (CGY-Y) /FY=0.517 In the case of "published" in FIG. 23B, the following result is obtained, and the center of gravity is obtained. Turns out to be significantly higher than right.

(CGX-X) /FX=0.574 (CGY-Y) /FY=0.442 Experiments were conducted on a number of examples, but the center of gravity should be as short as possible with a line as thin as possible. Even if the connection method is slightly different, the same value can be obtained very accurately.

Note that indices other than the center of gravity in a figure or the like having a plurality of labels may be obtained by simple addition. For example, the ferrite region (FX-F
Y) is unaffected by the consolidation and APM,
PM and A can be easily obtained by adding values for each label. Thereby, LNG and WA are also obtained. However, AC
V, AIH, and CYV are often affected by line addition.

In consideration of the above, even in the case of an irregularly arranged figure composed of a large number of labels, which is generally extremely difficult to calculate, when calculating the index according to the present invention, the analysis value of a figure connected to one label is particularly required. Was found to be used, and in this case, practically no trouble was caused.

As described above, according to the present invention, the aforementioned
Of the eigen exponents defined by equations (1) to (9),
Aspect ratio, area ratio, dose ratio, and fineness ratio
The feature is that the shape characteristic analysis is performed using all the unique indices.
I do. Of the above four characteristic indices, the dose ratio is calculated by the formula
It is possible to use either (4) or (5).
And the fineness ratio is given by the formula (6) or (7).
Either one can be used. In addition, as described later, an index group showing these characteristics can be used for a search according to purpose by storing a large amount of the index group with a classification item as a database of a computer. That is, if the shape characteristics of a figure are quantified into several types of indices, an extremely large amount of material data can be easily classified and accumulated. It is also possible to obtain character codes and the like. Also, by comparing newly created characters and designs with many existing characters in the database, it becomes easy to develop and anticipate new typefaces and logos.

As described above, in addition to the indices (1) to (9), the original values obtained by computer analysis (NH
L, RND, OLS, OSF, MSL, RLT, RS
T) and various calculated values (for example, AIH / FX · FY,
ACV / FX / FY, A / AIH, A / ACV, etc.) are more effective.

Next, an example of the actual analysis will be described.
An example showing the utility of the expression (9) will be described using various symbols, characters, and designs. In addition, figures and characters were used by extracting various designs such as various typeface typography published by Marl Company, "World Characters" published by Ryo Nakanishi, Shokodo, and "Mark Symbol 1-3" of Kashiwa Shobo.

The index ((1) to (9)) used in the present invention
Equations) have the property of being able to be compared with each other as they are, but various methods as shown below can be considered to display the calculated index values.

(I) Method of listing numerical values as they are (FIG. 25) (ii) Form (FX / FY) of feret region in equation (1) and (2)
A method of expressing A / (FX.FY) in the expression together with the values of AIH / (FX.FY) and ACV / (FX.FY) in a rectangle having the shape of a Feret area (FIG. 18). Within the region, (CGX-X) / F of the equation (8)
X and (CG−Y) / FY in equation (9) are shown together with lines that equally divide the region itself vertically and horizontally (FIG. 20 (C)). (Ii) and (iii) are shown in FIG. Other exponents may be listed as numerical values.

(Iv) A method of arranging mutually related indices using a radiation graph (radar chart) and visually patternizing the figure shape (FIG. 26). The following can be considered.

First pair: A pair of indices indicating the occupancy in the area Two-dimensional A / (FX · FY) (2) Three-dimensional CYV / (FX ² · FY) (3) 2 pairs: An index indicating the dose that constitutes the figure vs. the outer dose ratio 2APM / (FX + FY) ... (4) The inner / outer whole dose ratio 2 PM/ (FX + FY) ... (5) Third pair: ... Figure LNG / WA ... (6) Base on the dose and actual area of the outer circumference LNG / WA ... (6) Based on the total dose and the real area (PM / APM) ² ... (7) Various patterns can be created by combining these six indices. The scale of each pair in the radiation graph shown in FIG. 26 is assumed to be the same.

FIGS. 27 to 33 show examples in which the expression by the radiation graph is applied to a simple figure. In each of these figures, (A) shows the shape of the object, and (B) shows a radiation graph displaying the index of the object.

In FIG. 27, since PM includes the inside of the line, 2 PM/ (FX + FY) is 2APM /
(FX + FY) is twice as large. Further, in FIG. 28, A / (FX · FY ) is very close to the theoretical value "1", CYV / (FX 2 · FY) is very close to the theoretical value of "0.78". In FIG. 29, the target object is merely a polygonal line, and the original value A is extremely small. APM, PM
Both have an inner side and an outer side, which are almost twice as large as those in FIG. In FIG. 31, the target object is a maze formed of one line, and LNG / WA and (LNG / W
A) · (PM / APM) ² is extremely large.

Next, an example will be described in which the above-mentioned exponential expression in the present invention is analyzed with a printing character group that has been used for a long time as an object. Here, as the target character,
Examples of the alphabet, kanji, Hebrew, and Arabic are shown.

When considering characters, it is necessary to consider not only individual differences between the characters but also typeface differences between the same characters. For example, when considering uppercase letters of the alphabet, there is a difference of 26 characters from A, B, C,... Z, and there is a typeface that has been used for each character since ancient times.
Here, for each of the following six types of fonts shown in FIG.
An exponent was determined for each letter from uppercase A to Z. In this case, a typeface sample of "Alphabet" published by Marl was used.

Optima (1001) Optima Italic (1003) Optima Bold (1005) Gilsan Bold Italic (1015) Cloista Black (1087) Vivaldi (1102) The seven types of Chinese characters shown in FIG. We analyzed the typefaces (square typeface, Song Dynasty style, Gothic style, slave style, Mongolian style, old seal style, Shinkantai style). Kanji character types are
There are 2000 to 3000 or more for each typeface, but here 90 types were randomly selected from the common kanji of the Marl script. In the analysis sample, the same characters were used for each typeface.

Further, as characters having different series, FIG.
Hebrew and Arabic characters are also analyzed. As for these character samples, all characters are analyzed and analyzed by Shokodo Shuppan published by Ryo Nakanishi, "The Characters of the World".
Calculated.

FIG. 25 shows the results of the above analysis as numerical values. These numerical values are the average value of the character group of each typeface, and are distributed with a width around the average value. In some cases, the distribution is sharp, but in others, it is slightly wider.

Characters belonging to the same typeface are designed so that any combination of characters can achieve good harmony and consistency in shape. Accordingly, the pattern of the average value indicates the center of the distribution of the shape characteristics of the typeface. These average patterns are shown in FIGS.

Next, an example in which the same character is analyzed between typefaces will be described. Radiation graphs showing the results of these analyzes are shown in FIGS. 52 to 63 listed below.

FIG. 52: Optima A FIG. 53: Optima E FIG. 54: Optima Italic A FIG. 55: Optima Italic E FIG. 56: Optima Boldoku A FIG. 57: Optima Bold E FIG. 58: Figure 59: Kanji / Song Dynasty “sama” Figure 61: Kanji / Song Dynasty “sama” Figure 61: Kanji / square “yo” Figure 62: Kanji / Gosic “yo” Figure 63: Kanji / Song Dynasty “Yo” As can be seen from these figures, pattern similarity is seen between the same typefaces rather than between the same characters. Of course, some characters show special patterns, but in the case of kanji (especially, “sama” with a large number of strokes and “gi” with a small number of strokes) show a strong similarity in typeface.

FIG. 64 shows the analysis result of the Hebrew character as a special character. In addition to the character group, there are four types of designs with the theme of the capital letter "A".
The results of the mark analysis are shown in FIGS. All of these samples are excerpted from “Mark Symbol” published by Kashiwa Shobo. Many of the above character groups have similarity.

Description of Graphic Shape Analysis System Next, a graphic analysis system to which the above-described graphic shape analysis method using a unique index is applied will be described. This is 1
This is a specific example, and various applications are possible, and the type of each device is not limited to this.

FIG. 1 shows a schematic configuration of this graphic analysis system. This system has a system bus 1
1 are provided to connect between the following devices.

(I) Central processing unit (CPU) 12: a processor for controlling the operation of the entire system,
Various application programs are executed under an OS (operating system) loaded in the memory 13.

(Ii) Memory 13: a RAM (randomly readable and writable memory) for storing an operating system loaded from the hard disk 14 at the time of startup and storing necessary application programs and various data required for its execution, and a basic input / output system It is composed of a ROM (Read Only Memory) for storing programs. The application program includes, for example, an editing program for inputting new binary image data, correcting the input binary image data or the binary image data read from the image database 24, and calculating a unique index of a figure or the like. ,
It comprises an analysis program for performing analysis and various searches, and an input / output program for inputting and outputting various data.

(Iii) Hard disk 14: Stores an index database 23 constructed with unique indices relating to the OS, various applications, various figures and the like, and an image database 24 constructed with binary image data of various figures and the like.

(Iv) Image input device 15 (image scanner, CCD camera, etc.): A figure or the like drawn on a medium such as paper is read as binary image data,
4 is stored in the image database 24.

(V) Floppy disk device 16: Used to read various data or application programs from the floppy disk medium 22 as required. The hard disk 14 and the floppy disk device 16 can be replaced with other media. (For example, various magneto-optical disks, laser disks, CDs, etc.) (vi) Display device (CRT) 17: If necessary,
Displays various numerical data and image data.

(Vii) Output device 18: Prints out various numerical data, image data, and the like as necessary.

(Viii) Keyboard 19: Used for inputting processing instructions and necessary input / output parameters.

(Ix) Mouse 20: Used for inputting a processing command and necessary input / output parameters, as well as for inputting a binary image.

The operation of the graphic shape analysis system having the above configuration will be described. Here, as an example, a character production / modification process for producing or modifying a character in a typeface and a search / analysis process for judging the consistency between the produced or modified character and another existing character are shown in FIG. , And FIG. There is no limitation on the process as long as the process is capable of judging a figure by effectively using the above-mentioned index indicating the shape of a figure, and this index can be searched from a database and directly compared. Of course.

Production / Correction Step (FIG. 2) After the system shown in FIG. 1 is started, when “character production / modification” is selected from a processing selection menu (not shown) on the screen of the display device 17, editing in the memory 13 is performed. The program starts, and an input selection menu is displayed on the screen. Here, when a character already represented on a medium such as paper is used as a base (step S101; Y), a character image on the medium is read by the image input device 15 (step S10).
2). To create a completely new character (step S
101; N, step S103; Y), manual input is performed with the mouse 20 while looking at the screen. When correcting a character already registered in the image database 24, the character image data is read from the image database 24 (step S105).

The character image input or read in this way is displayed on the screen of the display device 17 (step S106), and if necessary (step S107; Y), the character image is corrected by the mouse 20 (step S108). ).

When the correction is temporarily completed (step S10)
7; N), the CPU 12 calculates the unique indices (1) to (9) based on the character image data (step S1).
09) In step S110, the unique index is output / displayed according to the output method (screen output or printer output) and display format (numerical table format, radiation graph, or Feret region format) designated by the screen menu. Here, the numerical table format is the format shown in FIG.
20 (B), (C)
The format is as shown in the figure.

When the character image is further corrected by looking at the analysis result of the unique index output and displayed in this manner (step S112; Y), the character image is corrected on the screen by the mouse 20 (step S108). ), Step S1 again
Return to 07. Then, redesign is performed by repeatedly performing the procedures of steps S107 to S112 until a target unique index value is obtained.

On the other hand, when there is no room for correction (ie,
If the character is completed up to a state where a satisfactory unique index is obtained (step S112; N), the unique index value at that time is stored in the index database 23 of the hard disk 14, and the binary of the completed character is stored. The image data is stored in the image database 24. At the time of storage in these two databases, the same identification code is assigned to one character image.

By repeating such operations, the index database 23 and the image database 24 are constructed in association with each other. In constructing these databases, classification and storage are performed using a method such as a tree structure or a hierarchical structure, for example, so as to be advantageous for later retrieval. As the classification, for example, it is preferable to roughly classify into kanji typefaces, general characters, special characters, general symbols, general figures, and the like, and further subdivide each of them, but it is not limited to this. Instead, use the most appropriate classification method as needed.

Character Retrieval / Analysis Step (FIG. 3) Next, processing for judging the consistency between a character in a certain typeface and another existing character and obtaining auxiliary information effective in designing the character. explain. The characters to be compared may be newly created characters, or both may be existing bypasses. Alternatively, one of the characters may be a newly created character and the other may be an existing character. Here, for the sake of explanation, a case will be described in which consistency between a newly produced character (for example, “A”) and an existing character (for example, “B”) is checked.

First, after activating the system of FIG.
A new typeface is produced for the capital letter "A" of the alphabet (step S201), and a unique index is obtained from the produced character image (step S20).
2).

Next, when "character search / analysis" is selected from a process selection menu (not shown) on the screen of the display device 17, an analysis program in the memory 13 is started, and a waiting for input of an analysis target is displayed on the screen. A message (not shown) is displayed.

Here, when the character "B" to be compared is input, one typeface character registered for the character "B" is selected from the exponent database 23, and a set of unique exponent value data for the character is registered. The data is read into the memory 13 (step S203).

Next, the CPU 12 compares each unique exponent value of the character image "A" obtained in step S202 with each read unique exponent value, and calculates a matching degree indicating the consistency between the two. (Step S204).

Various methods can be considered as a method of calculating the matching degree. For example, the ratio of the difference between the eigen index value and the eigen index value is obtained for each of the eigen indices (1) to (9). A method of calculating an average value of nine difference ratio data is considered. Alternatively, it is also possible to specify only particularly important indices among the nine unique indices, and take an average value of the difference ratio data only for these indices. According to these methods, when the two are completely matched, the matching degree becomes “0”, and the lower the matching, the larger the value.

Note that the method of calculating the matching degree is not limited to the above, and other methods (for example, fuzzy logic) can be used, or a plurality of matching degree calculation modes are prepared. It is also easy to realize selection and change as appropriate.

When the calculation of the degree of matching with one typeface character is completed, the exponent database 23
, Another typeface character registered for the character "B" is selected, the unique exponent value for that character is read, and the matching degree is calculated in the same manner as described above (step S).
203, S204). Hereinafter, such processing is performed for all character fonts of the character “B”.

Then, the smallest matching degree value is selected from the obtained matching degree values, and the binary image data of the corresponding character type is read from the image database 24, and the character " A "(step S206). In this way, the character "B" of the font that most closely matches the newly produced character "A" is automatically searched and displayed. In this analysis, the processing speed is extremely high because only the unique index is used without using any binary image data.
3, the working memory area can be reduced.

In this embodiment, the determination of the consistency of the character typeface has been described as an example. However, it is needless to say that the present invention can be similarly applied to a logo mark, a trademark, and the like.

[0131]

As described above, according to the first aspect of the present invention, if the shapes to be analyzed are similar in shape, regardless of their size, the reproducibility of accuracy does not hinder practical use. Since the analysis of the shape characteristics is performed based on the four characteristic index values having the following, the shape evaluation can be performed objectively. Further, in a system to which this analysis method is applied, analysis can be performed at high speed using these unique indices.

According to the shape characteristic analysis method for a figure or the like according to the second aspect of the present invention, even if a figure or the like composed of a plurality of separated elements is regarded as one figure or the like, the shape characteristic analysis is performed. The index calculation can be performed very easily as compared with the case where the search is performed for each individual element, and there is an effect that the processing speed of the analysis can be improved and the processing program can be simplified.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a graphic analysis system to which a graphic shape analysis method according to an embodiment of the present invention is applied.

FIG. 2 is a flowchart for explaining a character production / correction process in the operation of the system of FIG. 1;

FIG. 3 is a flowchart for explaining a character search / analysis step in the operation of the system of FIG. 1;

Fig. 4 Original value A [actual area] and AIH [area including holes]
FIG.

FIG. 5 is an explanatory diagram showing the original numerical values APM (perimeter of the object) and PM (sum of perimeters).

FIG. 6 is an explanatory diagram showing the original values FX [X-direction feret diameter] and FY [X-direction feret diameter].

FIG. 7 is an explanatory diagram showing HD (equivalent circle diameter).

FIG. 8 is an explanatory diagram showing ODL (elliptical long axis), ODS (elliptical short axis), and OLS (tilt of elliptical long axis).

FIG. 9: PMH [envelope circumference] and ACV [convex closed area]
FIG.

FIG. 10 is an explanatory diagram showing ML (maximum absolute length), MW (maximum absolute length width), and MLS (slope of maximum absolute length).

FIG. 11 is an explanatory diagram showing CYV (cylindrical volume).

FIG. 12 is an explanatory diagram showing EDH (horizontal equal diameter) and EDV (vertical equal diameter).

FIG. 13: LC1 [diameter 1 in four directions passing through the center of gravity], LC2
It is explanatory drawing which shows [four directions diameter 2 which passes through a center of gravity], LC3, and LC4 [four directions diameter 4 which passes through a center of gravity].

FIG. 14 is an explanatory diagram showing THL [maximum thickness in eight directions], THS [minimum thickness in eight directions], and THA [average thickness in eight directions].

FIG. 15 is an explanatory diagram showing a sample used for a test related to an error caused by scaling.

FIG. 16 is an explanatory diagram showing another sample used for a test related to an error caused by scaling.

FIG. 17 shows the uppercase letter “A” in the alphabet.
It is explanatory drawing which shows the original value A, AIH, ACV, and Feret area.

FIG. 18 is an explanatory diagram showing a feret region.

FIG. 19 is an explanatory diagram showing a stereoscopic ferrite area.

FIG. 20: For the capital letter “B” of the alphabet,
It is explanatory drawing which shows the original value A, AIH, ACV, and Feret area.

FIG. 21 is an explanatory diagram showing an example in which figures composed of four labels are connected.

FIG. 22 is an explanatory diagram showing an example in which figures composed of five labels are connected.

FIG. 23 is an explanatory diagram showing an example in which kanji composed of a plurality of labels are connected.

FIG. 24 is an explanatory diagram showing another example in which kanji composed of a plurality of labels are connected.

FIG. 25 is a table listing original numerical values as they are.

FIG. 26 is an explanatory diagram showing a radiation graph.

FIG. 27 is a diagram showing an analysis result of a simple graphic.

FIG. 28 is a diagram showing an analysis result of another simple graphic.

FIG. 29 is a diagram showing an analysis result of another simple graphic.

FIG. 30 is a diagram showing an analysis result of another simple graphic.

FIG. 31 is a diagram showing an analysis result of another simple graphic.

FIG. 32 is a diagram showing an analysis result of another simple graphic.

FIG. 33 is a diagram showing an analysis result of another simple graphic.

FIG. 34 is a diagram showing six types of alphabets.

FIG. 35 is a diagram showing six types of kanji typefaces.

FIG. 36 is a diagram showing Hebrew characters and Arabic characters.

FIG. 37 is a diagram showing an average value of Optima (1001).

FIG. 38 is a diagram showing an average value of Optima Italic (1003).

FIG. 39 is a diagram showing an average value for Optima Bold (1005).

FIG. 40: Gilsan Bold Italic (101
It is a figure which shows the average value about 5).

FIG. 41 is a diagram showing an average value for Cloister Black (1087).

FIG. 42 is a diagram showing an average value for Vivaldi (1102).

FIG. 43 is a diagram showing an average value of a square typeface.

FIG. 44 is a diagram showing an average value for a Song Dynasty body.

FIG. 45 is a diagram showing an average value of a Gozic body.

FIG. 46 is a diagram showing an average value for a literary typeface.

FIG. 47 is a diagram showing an average value for Mongolian script;

FIG. 48 is a diagram showing an average value of an old stamp body.

FIG. 49 is a diagram showing an average value for a new flow.

FIG. 50 is a diagram showing average values for Arabic characters.

FIG. 51 is a diagram showing an average value for Hebrew characters.

FIG. 52 is a diagram showing an analysis result of Optima A.

FIG. 53 is a diagram showing an analysis result of E of Optima.

FIG. 54 is a diagram showing an analysis result of A of Optima Italic.

FIG. 55 shows a radiation graph for E of Optima Italic.

FIG. 56 shows a radiation graph for Optima Bold A.

FIG. 57 shows a radiation graph for E of Optima Bold.

FIG. 58 is a diagram showing a radiation graph related to “sama” of kanji and square characters.

FIG. 59 is a diagram showing a radiation graph related to “like” of kanji and gojikku.

FIG. 60 is a diagram showing a radiation graph relating to “sama” of the kanji and Song dynasties.

FIG. 61 is a diagram showing a radiation graph relating to “Yo” in kanji / square script.

FIG. 62 is a diagram showing a radiation graph relating to “gi” of kanji and gojik.

FIG. 63 is a diagram showing a radiation graph related to “Yo” of a kanji / Song dynasty.

FIG. 64 is a diagram showing an analysis result of Hebrew characters.

FIG. 65 is a diagram showing an analysis result of a design character with a theme of “A”;

FIG. 66 is a diagram showing an analysis result of another design character with a theme of “A”;

FIG. 67 is a diagram illustrating an analysis result of another design character with a theme of “A”;

FIG. 68 is a diagram illustrating an analysis result of another design character with a theme of “A”;

[Explanation of symbols]

 12 CPU 13 Memory 14 Hard Disk 15 Image Input Device 17 Display Device 18 Output Device 23 Index Database 24 Image Database

Claims (2)

(57) [Claims] 1. A character, a database seeking index following four items a symbol or figure shape from a computer image processing various measurement data obtained by measuring, by using the entire part of these indices, A shape characteristic analysis method for a figure or the like characterized by performing shape characteristic analysis for production, arrangement, design, and identification of characters, symbols, or figures: Aspect ratio FX / FY Expression (1) Area ratio A / (FX)・ FY) Expression (2) Dose ratio 2APM / (FX + FY) Expression (4) or 2 PM/ (FX + FY) Expression (5) Dense ratio LNG / WA Expression (6) or (LNG / WA) × (PM / APM) ² … Expression
(7) Here, A is the actual area of the object, FX is the Feret diameter in the X direction (distance between two vertical tangents), FY is the Feret diameter in the Y direction (distance between two horizontal tangents) ) , A PM is the perimeter of the object, PM is the sum of the perimeters of the object (sum of the perimeter of the object and the perimeter of the internal hole), LNG is the length (= APM / 2), and WA is the average Width (= 2A / APM) . 2. The method according to claim 1, wherein said character, symbol or graphic composed of a plurality of separated elements is converted into one character, symbol or graphic by connecting said plurality of elements with a thin line. A shape characteristic analysis method for a figure or the like, characterized by performing shape characteristic analysis.