JPH05127648A - Compression device for outline font data - Google Patents

Abstract

PURPOSE:To decrease the capacity of a memory for holding outline font data. CONSTITUTION:A shared pattern extracting means 11 extracts the number of a character, which can be considered to be synthesized from a registered character list and extracts the character numbers of the main body and crown part constituting a composite character and a restoration information calculating means 12 expands the composite character to obtain the base points of the main body and crown part, expands the main body to obtain its base point, and expands the crown part to obtain its base point. The offset values of the main body and crown part are calculated to correct their offsets. An information adding means 13 additionally stores the offset values in the table when coordinate data on the main body and crown part of the composite character match each other.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for compressing outline font data which constitutes print data of characters.

[0002]

2. Description of the Related Art Generally, some problems remain when constructing a Japanese desktop publishing (DTP) system, and one of them is the font capacity. In DTP, it is indispensable to handle fonts of many typefaces and character sizes, but outline fonts have the advantage that the data capacity is smaller than bitmap fonts when attention is paid to character size, and scaling and rotation are also important. When doing, there is an advantage that the character quality can be maintained. When this outline font is composed of Beze curves as shown in FIG. 39, a start point S, an end point E, and a total of four points C1 and C2 form one curve.

[0003]

However, although the outline font has a smaller data capacity than the map font, one typeface in Japanese is composed of about 8000 characters according to the JIS 1st and 2nd levels. So
The amount of data becomes huge. In addition, as shown in FIG. 39, the outline font has a large amount of data in the case of a complicated Chinese character because one curve is composed of four points, and the amount of data in one typeface is about 4 Mbytes. Therefore, in the DTP system, a method is adopted in which outline font data for several typefaces which are frequently used are built in a ROM (read only memory) and outline font data of other typefaces are held by a hard disk device or the like.

In view of the above conventional problems, it is an object of the present invention to provide an outline font data compression apparatus capable of reducing the capacity of a memory for holding outline font data.

[0005]

In order to achieve the above object, the first means is to extract the outline font data of a graphic in which a part of a character can be shared, and the graphic extracting means. Restoration information calculating means for calculating information for restoring the original character by the drawn graphic, information adding means for adding information indicating that the character is composed of the graphic extracted by the graphic extracting means, A graphic unit extracted by the graphic extraction unit; the restoration information calculated by the restoration information calculation unit; and a compression unit configured to compress the information added by the information addition unit into a predetermined format. Characterize.

The second means is an extracting means for extracting data exceeding 1 byte from the outline font data of the character, and an X-adic number for preliminarily calculating an X-adic number such that the spread of the distribution of the appearance frequency with respect to the data width of the character becomes small. It is characterized in that it is provided with a base number calculation means and a compression means for compressing the data extracted by the extraction means with the X base number.

A third means is a hierarchizing means for hierarchizing the character outline font data for each information, and a calculating means for calculating the appearance frequency of the data of each hierarchy hierarchized by the hierarchizing means, And a compression unit that creates a compression replacement table for each layer based on the appearance frequency of each layer calculated by the calculation unit and compresses each layer.

The fourth means is a grouping means for extracting the curve data of the original outline font data in advance and grouping the outline font data,
It is characterized by comprising compression means for calculating and compressing a difference between the original outline font data and the grouped outline font data.

The fifth means is characterized by further comprising means for further compressing the code compressed by the compression means of the first to fourth means by reversible encoding.

[0010]

According to the first means, the outline font data of the figure capable of sharing a part of the character is extracted and coded by the above-mentioned structure, so that the character can be decomposed and compressed. The capacity of the memory for holding the font data can be reduced.

In the second means, an X-adic number is calculated in advance so that the spread of the appearance frequency distribution with respect to the character data width is reduced, and data exceeding 1 byte is compressed with this X-adic number. The efficiency can be improved.

In the third means, the outline font data of characters is hierarchized according to its information to calculate the appearance frequency of the data of each hierarchy, and the appearance frequency of each hierarchy is used to determine the hierarchy by the compression replacement table for each hierarchy. Since each is compressed, the compression efficiency can be improved.

In the fourth means, since the difference between the original outline font data and the grouped outline font data is encoded, the compression efficiency can be improved.

In the fifth means, the code compressed by the compressing means of the first to fourth means is further compressed by lossless encoding, so that the original data can be restored by lossless decoding.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. 1 is a block diagram showing an embodiment of an outline font data compression apparatus according to the present invention, FIG. 2 is an explanatory diagram showing a main body and a crown of a character, FIG. 3 is an explanatory diagram showing an offset of outline font data, and FIG. Is an explanatory view showing the outline of the restoring operation, FIG. 5 is an explanatory view showing the overall structure of the outline font data, FIG. 6 is an explanatory view showing the base point of the outline font data, and FIG. 7 is the outline font data of the conventional example and this embodiment. 8 is a flow chart for explaining the compression operation, FIG. 10 is a flow chart for explaining the decompression operation, and FIG.
1 is an explanatory diagram showing lossless encoding.

In order to facilitate the explanation, a European typeface will be described below as an example. The major difference between the European typeface and the Japanese typeface is the difference in data volume due to the complexity of the characters, but since the typeface data format is the same for both, the Japanese typeface is compressed by the method of compressing the European typeface. can do. First, in a European typeface, there are characters with graphics added to the upper part of the alphabet.
There is a character called "a umlaut" in German etc. as shown in. It should be noted that kanji and the like are configured by more complicated combinations.

Here, the alphabet parts "A" and "a"
Is called the main body, and the figure added to the upper part of this main body is called the crown. If the coordinate data excluding the offset of the crown is the same, the uppercase crown shown in FIG. It can be used for lowercase letters shown in b). That is, the offset will be described with reference to FIG.
When the uppercase crown shown in (a) is used for the lowercase letters shown in FIG. 2 (b), they do not match unless they both move in the negative direction in the xy directions.

By the way, as shown in FIG. 3 as an example of the format of the outline font, the base point (xs, ys) is used as a reference, and unless this reference is moved, the relative value to the previous point is coded. In some cases, the difference in offset means the difference in this base point. When characters of only a crown are registered as shown in FIG. 4A, the crown and the main body shown in FIG.
The uppercase letters shown in (a) can be restored, and similarly, the lowercase letters shown in FIG. 2 (b) can be restored by this crown and the main body shown in FIG. 4 (c). Further, in the case where the characters of only the crown shown in FIG. 4 (a) are not registered, the crown data is extracted from the capital letters shown in FIG. 2 (a) and registered.
The original data shown in (a) can be deleted.

In the present embodiment, as shown in FIG. 1, the outline extracting font data of a graphic capable of sharing part of a character is extracted by the graphic extracting means 11, and the restoring information calculating means 12 is used.
The information adding means 1 calculates the information for restoring the original character by the graphic extracted by the graphic extracting means 11.
3, information indicating that it is a character composed of the figure extracted by the figure extracting means 11 is added, and the figure extracted by the figure extracting means 11 and the restoration information calculated by the restoration information calculating means 12 are added. , Information adding means 13
The information added by is compressed into a predetermined format by the compression means 14 and compressed. Therefore, the larger the number of characters having such a common crown, the smaller the amount of data. In the present embodiment, the compressed data is further compressed by the lossless encoding means 15 and recorded on the medium. When output, the lossless decoding means 16 decodes the compressed data and the character expanding means 17 expands the original outline font data. Is configured to.

Next, when the crown and the main body are combined, if the relative data other than the offset in the coordinate data are not equal, the original character cannot be constructed, but the outline font data for one character is Figure 5
As shown in (b), character-specific data is added in addition to the coordinate data. Note that FIG. 5A shows all the data of one typeface. Typeface information such as typeface name, creation date, number of registered characters, etc., and character index for each registered character, character reference such as data reference location, etc. It is composed of information and character information which is outline font data for each registered character. Then, FIG. 5B shows a detailed structure of this character information, which is composed of unique data “0” “1” for character correction and coordinate data. That is, unlike the coordinate data, this unique data cannot be combined, so
In this embodiment, only the coordinate data is compressed.

Next, the actual offset will be explained. As shown in FIG. 6, the base point of the outline data is
Since there are almost a plurality of characters in one character, a base point for calculating the offset between the crown and the main body is required. Here, as in the case of the characters shown in FIG. 6, the data storage order is overwhelmingly started from the main body, and this synthetic character is the base point s0 →
The main body is registered in the order of s1 → s2 → s3, the main body is registered in the order of base points s0 → s1, and the crown is registered in the order of base points s2 → s3. Therefore, the offset can be calculated even if these base points are out of order. In addition, as an exception, since the writing start coordinates of the crown may be different between the combined character and the character of the crown alone, the offset of such a character is not calculated.

Specifically, the data between each base point is developed, and the minimum rectangular coordinates (for example, two coordinates, lower left and upper right) that can surround the figure are obtained. When the coordinates of the base points s0 to s3 in FIG. 6 are represented by the following formula (1), sn: lower left coordinate (xln, yln), (upper right coordinate (xun, yun) (where n = 0, 1, 2, 3) (1) The rectangle of the base point s0 of the main body becomes the maximum, and the rectangle of the base point s1 is included.Therefore, the figure included in this maximum rectangle is determined as the main body, and the base point s0 is the offset of the main body. Used to calculate

Further, the main body and the crown can be distinguished by considering all the rectangles existing outside the largest rectangle as the crowns, and therefore, in the example shown in FIG.
Each rectangle of 3 can be discriminated as a crown. Note that FIG.
When the crown is composed of a plurality of parts as shown in, the leftmost rectangular base point s2 is used to calculate the offset of the crown. Therefore, each offset of the main body and the crown of the character shown in FIG. 6 can be calculated as follows.

Body: (xl0, yl0) (2) Crown: (xl2, yl2) (3) When the characters shown in FIG. 2 are combined, the characters shown in FIG.
By calculating the coordinates of a similar rectangle from the data shown in (b), the offset of the body is (xl0 ', yl0'),
Since the offset of the crown can be calculated as (xl2 ′, yl2 ′), the offset amount for the parallel movement can be calculated by the equations (4) and (5).

Body: xh = xl0-xl0 ', yh = yl0-yl0' (4) Crown: xc = xl2-xl2 ', xc = yl2-yl2' (5) Next, this movement amount and combination Change the area in which the characters to be stored are stored and register again. Here, FIG.
To explain the coordinate data shown in (b) in detail, the coordinates of outline font data are generally 1000 × 100.
It is stored based on 0. Therefore, since the maximum relative value exceeds the number "255" that can be expressed by 1 byte, 2 bytes may be required. Also, since one of the relative values of the straight line in the xy direction is “0”, the capacity can be reduced by encoding the coordinates with a variable length in the range of 0 to 2 bytes. In the case of (1), a flag that determines the length is required. That is, the coordinate data repeats the information shown in Expression (6), but the format of Expression (6) is converted into Expression (7).

Coordinate data: flag + coordinate + correction information (6) Flag + xh + yh + reference location + flag + xc + yc + reference location (7) The flag shown in equation (7) is used for recognizing the special format, but variable. Since length determination and other information are set by turning bits on and off, it can be realized by a combination of vacant bits. Although it is necessary to change the data length by this process, in this case, the new data length is overwritten in the area where the old data length is stored by counting the number of bytes while rewriting with the new format. be able to.

Further, since the data amount is reduced and the data reference location is changed by the new format, it is necessary to change the reference location data of the character index information shown in FIG. 5A, but all unnecessary data is deleted. A new font data can be created by shifting. That is, although the synthetic character is constructed in a completely different format, the character expansion means 17 shown in FIG. 1 reads the main body data from the reference location of the main body, expands it by the offset movement amount, and similarly processes the crown. As a result, data equivalent to the original font data can be obtained.

Here, the difference between the conventional example and this embodiment will be described. In the conventional generation process, as shown in FIG. 7A, a character is generated by moving character information once to a reference place. In this embodiment, as shown in FIG.
A character is generated by moving the information of the composite character three times.

Next, the compression operation will be described with reference to FIG. 8. First, in the routine for creating the table of composite characters shown in FIG. 8, character numbers which are considered to be compositable are extracted from the registered character list ( In step S1), the character numbers of the main body and the crown forming the composite character are extracted (step S1).
2) Create a character number table for the number of characters N, the composite character PN, the main body MN, and the crown CN (step S3).

Then, after the counter i indicating the number of characters N is cleared (steps S4 and S5), the composite character PN of the number indicated by the counter i is expanded as shown in FIG.
The base points of N and the crown CN are obtained (step S6), the main body MN is expanded to obtain the base point (step S7), and the crown CN is expanded to obtain the base point (step S8). Then, the offset values of the main body MN and the crown CN are calculated by Equation 2 (step S9), and the offset is corrected by Equation 3 (step S10).

Next, the main body MN of the composite character PN and the crown CN
It is determined whether or not the coordinate data of the two match (step S11), and if they match, the offset value is added and stored in the table (step S12). Deleted from (step S1
3). Then, the counter i is incremented (step S14), the process returns to step S5, the same process is performed for the next character, and when the process for all the characters is completed, this synthetic character table creation routine is ended (step S).
15).

In the new font file creation routine shown in FIG. 9, first, the character index information of the uncompressed portion is read from the original file RF to the end, and after the number of registered characters M is obtained (step S21), that information is newly created. Copy to the file WF (step S22). Then, after clearing the counter i indicating the number of registered characters M (step S
23, S24), information such as a reference location is read from the i-th character index portion (step S25), and the character exists in the above-mentioned composite character table (step S2).
In step 6), the data of the original file RF is replaced with the new file WF if it does not exist.
(Step S27).

In step S28 and thereafter, the unique data as shown in FIG. 5B is read from the afterglow location of the original file RF, copied to the new file WF (step S29), and the new file WF is updated as shown in Formula 5. Create a format and write it (step S30), and write change data such as data length and reference number (step S30).
31). Then, the counter i is incremented (step S32), the process returns to step S24, the same process is performed for the next character, and when the process for all the characters is completed,
It branches to step S33. In step S33 and thereafter, a compression conversion table for each character is created by the reversible code, the reference location is rewritten (step S34), and this new font file creation routine is finished (step S35).

Next, the restoring operation of the compressed font data will be described with reference to FIG. First, the new file W
When the typeface in F is designated (step S41) and the designated character i of the typeface is input (step S42), the data of the character number i is read from the character index portion (step S41).
43), the data at the reference location is replaced by the compression conversion table and written in the buffer (step S44), and the unique data is sent to the character expansion unit (step S45).
Then, in step S46, if this character is in the special format, the process proceeds to step S48 and below, and if it is not the special format, the process branches to step S47 to send the remaining data to the character expanding unit, and the process is ended.

In step S48 and subsequent steps, the reference location data is completely replaced by the conversion table and written in the buffer, the offset of the coordinate data of the main body is corrected and sent to the character expanding section (step S49), and then the reference location data is stored. The data is completely replaced by the conversion table and written in the buffer (step S50), the offset of the coordinate data of the crown is corrected and sent to the character expansion unit (step S51), and this restoration processing is ended.

Here, since the outline font data compression in this embodiment needs to be completely restored when decoding, lossless encoding is used as shown in FIG. In this case, even if the compressed data is completely different from the current data, if the data seen in the direction of the arrow in the figure from the device for expanding the outline font as shown in FIG. 11 is not the current data, the compressed data is completely restored. be able to. Further, when the typeface information and the character index portion are compressed in FIG. 4A, the character information cannot be retrieved unless all of this information is decoded, so that in this embodiment only the character information is compressed. ..

Here, when the characters of a plurality of typefaces are mixed and output, the development time becomes a problem, and this problem is
This becomes noticeable when Japanese has a large number of registered characters. In addition, the compressed character information must be extracted as one block, but since the compressed data is encoded in a variable length, the code that is less than the last 1 byte of one character is converted into byte units by adding bits. Must.

FIG. 12 shows an example in which all sides are composed of 1-byte units and divided into 8 bit units. When all characters are encoded at once, as shown in the upper part of FIG. Will not do. Therefore, as shown in the lower part of FIG. 12, the character unit corresponding to the data length can be read from the reference location without omission by arranging to be able to cut out in byte units.

Next, a description will be given of a second embodiment in which this 1 byte is encoded as a unit. FIG. 12 is an explanatory diagram showing a difference in data between the conventional example and the second embodiment, and FIG.
3 is a block diagram showing the second embodiment, FIG. 14 is an explanatory diagram showing data in byte units, FIGS. 15 and 16 are flow charts for explaining the compression operation of the second embodiment, and FIG. 17 is a decompression operation. It is a flow chart for explaining.

Outline font data is generally
The length is recorded in units of 1, 2, and 4 bytes, but in order to form a font, it is necessary to read data in accordance with a certain format, and therefore the byte length at the time of reading is known. Therefore, if all the font data is assumed to be in 1-byte units, the 1-byte has a data width of 0 to 255 without a code. Therefore, it is necessary to reduce the data width in order to reduce entropy. That is, when the horizontal axis is the data width of 0 to 255 and the vertical axis is the distribution of the appearance frequency (number) of the data, a condition that a normal distribution concentrated on arbitrary data is formed and the spread of the distribution is small, That is, the compression efficiency can be improved as the condition that the standard deviation is small is satisfied.

Therefore, since the minimum unit of 1 byte of information cannot be manipulated by itself, the second embodiment extracts data exceeding 1 byte from the outline font data of a character as shown in FIG. Data extracting means 21, an X-adic number calculating means 22 for pre-calculating an X-adic number such that the spread of the distribution of the appearance frequency with respect to the data width of the character becomes small, and the data extracted by the data extracting means 21 with this X-adic number. Having compression means 23 for compressing,
By manipulating information in units larger than 1 byte,
It is configured to reduce entropy. Further, this compressed data is further compressed by the lossless encoding means 24 and recorded on the medium, and upon output, it is decoded by the lossless decoding means 25 and expanded by the character expanding means 26 into the original outline font data.

As shown in FIG. 14, the 2-byte information is in the range of 0 to 65535 without a code, and the number of each upper and lower 1 byte is counted at the time of encoding.
Entropy can be reduced by bringing the values of the upper byte and the lower byte close to similar values. Therefore,
In order to replace information exceeding 1 byte with X-adic data, the optimum X-adic value is first obtained. Note that any value can be expressed by equations (8) and (9).

AX + b = c (8) a <X, b <X (9) (where a: upper byte value, b: lower byte value) Here, the basic unit of the outline font is "100".
Since there is no coordinate-related data that deviates significantly from this value, the above values a and b are set to "100" without a sign.
The maximum value is around "0". Similarly, for other data lengths, the value per character is less than "1000". According to equation 4, the coordinate data includes positive / negative information even if the relative value is negative or the base point starts from negative.
It can be handled as unsigned information.

Note that negative data exists in a small part of the character information shown in FIG. 5, and when there is a sign, the most significant bit becomes "1", so that the data width widens. Therefore, in order to replace these small numbers of negative information with positive information before obtaining the X-adic value, the minimum value of the negative information is extracted and an offset is added so that this minimum value becomes "0". .. The character information that is actually compressed is expressed in units of 1 and 2 bytes, and data in units of 4 bytes does not exist in the compression area.

Next, the maximum value is extracted from the 2-byte information, all of which are unsigned. If this maximum value is set to the value c of Equation 6 and a = b = X−1, X can be obtained by the minimum value that satisfies the expression (10).

(X−1) · (X + 1)> c (10) If, for example, c = 1050 in the equation (10), X =
33, and normally a and b have a maximum value of "255" in a 256-ary number, but the maximum in this X-ary number is X-1, so the spread of the distribution can be suppressed.

Next, the compression operation will be described with reference to FIG. 15. First, the X value and the negative correction value are calculated as shown in FIG. That is, the registered character number N is obtained from the original file (step S61), and the counter i indicating the registered character number N is cleared (steps S62, S).
63). Then, the index information of the i-th character is read from the search unit (step S64), and the process moves to the reference location to obtain the maximum negative number and the maximum value of 2 bytes (step S65). Then, the counter i is incremented (step S6
6) Return to step S63, calculate the maximum negative number of each character and the maximum value of 2 bytes (step S67), then use the absolute value of the maximum negative number as a positive conversion value, and convert the maximum value of 2 bytes to X. The value is calculated (step S68).

Then, in FIG. 16, the character index information of the non-compressed portion is read to the end from the original file RF, and after the registered character number N is obtained (step S71).
Copy that information to the new file WF (step S
72). Then, the counter i is cleared (step S7
3, S74), information such as a reference place is read from the i-th character information section (step S75), and 2 in the character information section is read.
The byte data is replaced with an X-adic number (step S76).
Then, the counter i is incremented (step S7
7) Returning to step S74, each character is replaced with an X-adic number, and then a compression conversion table for each character is created by the lossless code described in the first embodiment (step S7).
8) The reference location is rewritten (step S79), and the new font file creation routine is finished (step S80).

Next, the operation of restoring the compressed font data will be described with reference to FIG. First, as in the case shown in FIG. 10, when the typeface in the new file WF is designated (step S81) and the designated character i of the typeface is input (step S82), the data of the character number i is read from the character index portion ( In step S83), the data at the reference location is replaced by the compression conversion table and written in the buffer (step S84). Then, the data is taken in according to the format, the 2-byte data is converted into a 256-ary number (step S85), and when the conversion of all the character data is completed, the data is sent to the character expansion section (step S86).

Next, a third embodiment will be described. FIG.
Is a block diagram showing the third embodiment, FIG. 19 is an explanatory diagram showing the appearance frequency of outline font data, FIG. 20 is an explanatory diagram showing the appearance frequency of flag hierarchies, and FIG. 21 is the compression operation of the third embodiment. FIG. 22 is a flowchart for explaining.
Is a flow chart for explaining the restoration operation.

Referring to FIG. 5 and the equation (6) described in the first embodiment, the compression area is 1) unique data “0”, 2) unique data “1”, 3) flag, 4) coordinates, 5 ) It can be classified into five layers of correction information. The distribution of the appearance frequencies in which the five layers are not classified does not attenuate like a normal distribution as shown in FIG. 19, and conversely there is a whisker-like portion that increases as indicated by the symbol T. Since the whiskers are abnormal in appearance frequency on both sides, it is considered that the feature of one of the layers appears.

Therefore, since it is not possible to realize a high compression rate by utilizing this distribution, in the third embodiment, as shown in FIG. 18, the layering means 31 sets the outline font data of the character for each information. And the calculating means 32 calculates the appearance frequency of the data of each hierarchy hierarchized by the hierarchizing means 31, and the compressing means 33 calculates by the calculating means 32.
A compression substitution table for each layer is created based on the appearance frequency of each layer calculated by the above, and compression is performed for each layer. Then, the compressed data is further compressed by the lossless encoding means 34 and recorded on the medium, and when output, decoded by the lossless decoding means 35, and expanded by the character expanding means 36 into the original outline font data.

This layer has a strong correlation for all characters and is 1
Although the distributions of the characters show similar tendencies, the whiskers T have a great influence in the case of a character having a large amount of data in the five layers. That is, when the ratio of the hierarchy itself to the whole is small, a method in which this embodiment is not applied can be considered, but in this embodiment, it is applied to all five hierarchies. If you take the distribution for each of the five layers, one of them should show an abnormal distribution that is completely different from the normal distribution, but due to the nature of the data,
The hierarchy of 3) flags and 5) correction information indicates the abnormal distribution. The reason is that it is derived from the nature of the flag and forms data by bit manipulation.

As shown in FIG. 20 (a), these data have a comb-like distribution that is missing in some places. Therefore, in this embodiment, the distribution of each layer is sorted and as shown in FIG. 20 (b). A replacement table is created, and the source data of the layer is rewritten by this replacement table. Then, by using this replacement table, the whiskers T shown in FIG. 19 are eliminated and corrected to a distribution close to a normal distribution, and after the replacement of all layers is completed, the entire distribution of the compression target range is taken again and encoded.

Next, this compression operation will be described with reference to FIG. First, in FIG. 21, as in the case of FIG. 15, the number N of registered characters is obtained from the original file (step S91), the counter i indicating the number N of registered characters is cleared (steps S92, S93), and the index of the i-th character is obtained. Information is read from the search unit (step S94). Then, the data is read according to the format, and the histogram of the data is obtained byte by byte for each layer (step S
95), this process is performed for each character (step S9).
6 → S93). Then, the respective histograms are sorted in descending order, and the maximum data is set to "0" (step S97) to create this hierarchical replacement table (step S98).

Then, in FIG. 22, as in the case shown in FIG. 16, the character index information of the non-compressed portion is read from the original file RF to the end, and after the registered character number N is obtained (step S101), that information is read. New file W
Copy to F (step S102), clear counter i (steps S103 and S104), and read information such as a reference location from the i-th character information section (step S1).
05). Then, the data is read according to the format and replaced by the replacement table of that layer (step S
106), this process is performed for each character (step S
107 → S104). Then, similarly to the case shown in FIG. 16, a compression conversion table for each character is created by the lossless code (step S108), the reference location is rewritten (step S109), and this new font file creation routine is finished (step S110). ..

Next, this restoration operation will be described with reference to FIG. First, as in the case shown in FIG. 17, the typeface in the new file WF is designated (step S11).
1) When the designated letter i of the typeface is input (step S
112), the data of the character number i is read from the character index portion (step S113), the data at the reference location is replaced by the compression conversion table and written in the buffer (step S114). Then, the data is taken in according to the format and converted into the original data by the replacement table of the layer (step S115), and when the conversion of all the character data is completed, the data is sent to the character expansion section (step S116).

Further, as shown in FIG. 19, the first to third embodiments may be combined to form a block, an X-adic number, or a hierarchical permutation. In addition, in FIG.
If it is converted into X-adic numbers after permutation according to hierarchy, it is meaningless to sort it, and the blocking shown by the broken line should not be performed with a typeface that cannot be easily blocked.

Here, if the histograms are sorted and replaced for each layer as in the third embodiment, the entropy of the distribution decreases, but in order to further improve the compression efficiency, 4).
Focusing on the coordinate data part, the coordinate data does not show an abnormal distribution unlike the other layers, but shows a normal distribution. Therefore, by predicting the coordinate data and encoding the prediction error, the amount of coordinate data can be reduced and a high compression rate can be realized.

Next, a fourth embodiment will be described. Figure 25
26 is a block diagram showing a fourth embodiment, FIG. 26 is an explanatory diagram showing a Beze curve, FIG. 27 is an explanatory diagram showing a unit pattern, FIG. 28 is an explanatory diagram showing each unit pattern of FIG. 27, and FIG. 29 is a first diagram. 30 is an explanatory diagram showing a unit pattern, FIG. 30 is an explanatory diagram showing a second unit pattern, FIG. 31 is an explanatory diagram showing a third unit pattern, and FIG. 32 is an explanatory diagram showing a combination example of first to third unit patterns. 33 is a flow chart for explaining the first stage of pattern integration in the fourth embodiment, FIG. 34 is a flow chart for explaining the second stage of pattern integration and pattern file creation, and FIG. 35 is for the source font file. FIG. 36 is a flowchart for explaining a rewriting operation, FIG. 36 is a flowchart for explaining a compression operation, FIG. 37 is a flowchart for explaining a decoding process, and FIG. Is an explanatory diagram showing a difference calculation process.

In this embodiment, as shown in FIG. 25, the grouping means 41 selects, for example, the Mincho font and the textbook font as the first and second patterns, and patterns close to the respective curves are selected. It is configured to encode the difference from the pattern. First, the Beze curve is determined at points P0 to P3 as shown in FIG. 26, and when the differences dxi and dyi (i = 0 to 2) from the previous points are encoded,
Normally, when dxi * dyi = 0, the horizontal line and the vertical line form a peculiar pattern. Therefore, if this peculiar pattern is not taken into consideration, the amount of information increases. That is, in the case of a horizontal line (dy0 = 0), the x coordinate information is 1 to 2 bytes and the y coordinate information is 0 bytes, so that the y coordinate information increases.

Further, in order to pattern a curve, the pattern must be known in an arbitrary curve, so in the present embodiment, bits "0" to "5" of 1-byte pattern information are set as pattern numbers. , Bit “6”,
"7" is used as information of the first to fourth quadrants of the xy coordinates. That is, the maximum number of pattern numbers of the bits "0" to "5" is 64. Therefore, step S in FIG.
As described in 120, as the first step of pattern integration, the number of pattern classifications is increased in consideration of the characteristics of the curve, and then, as shown in FIG. 34, the maximum number of patterns (= 64) is set as the second step of pattern integration. ) Integrate the patterns so that: For example, using the unit pattern shown in FIG. 27, the first unit pattern is dx0> 0, dy0 =
0, the second unit pattern is dx1> 0, dy1 = 0, and the third unit pattern group is 10 patterns (dy2 <0, dx2 = 0) (dy2 <0, dx2> 0, | dy2 |> dx2 ) (Dy2 <0, dx2> 0, | dy2 | = dx2) (dy2 <0, dx2> 0, | dy2 | <dx2) (dy2 = 0, dx2> 0) (dy2> 0, dx2> 0, dy2 <Dx2) (dy2> 0, dx2> 0, dy2 = dx2) (dy2> 0, dx2> 0, dy2> dx2) (dy2> 0, dx2 = 0) (dx2, dy2 other than the above) ..

If the curve used in the typeface exists in the group and is small in number, it is integrated into the pattern having the maximum number in the group. In this case, a one-sided test (statistics) is used for a small number of judgments, and for example, the rejection range is 5%.
At this time, the threshold value th for the small number determination and the total pattern number T are th = 0.05 * T / 0.95. Then, 1) register a pattern that is equal to or larger than the threshold th, 2) register a pattern that does not satisfy 1) but shows the maximum pattern in the group when a small number of patterns exist, and 3) register the number in 2)> 2) is deleted when 64, and 4) 1) is reintegrated when the number of registrations> 64 in 3).

This unit pattern will be described in detail. In this embodiment, as shown in FIGS. 28 to 31, one curve is set as a set of three unit patterns, and a certain condition is set.
Classify the curve into a reasonable number. For example, point P shown in FIG.
The straight line consisting of 0 and P1 is used as the first unit pattern in FIG.
It is assumed that 0 ° ≦ ang <90 ° in the first quadrant as shown in FIG. The reason for this is that when the first unit pattern is in the second quadrant of 90 ° ≦ ang <180 ° as shown in the figure f, the y-axis target position is set, and 180 ° ≦ ang <as in the figure g.
Rotate according to the placement information by setting the origin target position when in the third quadrant of ° 270 and the x-axis target position when in the fourth quadrant of 270 ° ≦ ang <° 360 as shown in FIG. Because you can.

Next, the straight line formed by the points P1 and P2 shown in FIG. 26 is used as the second unit pattern, and its existing location is shown in FIG.
It is limited to the first and fourth quadrants as shown in. Also, point P
A straight line composed of 2 and P3 is used as a third unit pattern, and the existence prediction range of this third unit pattern is made to depend on the angle of the second unit pattern. For example, when the thick line shown in FIG. 31 is the second unit pattern, it is assumed that the third unit pattern is within a range of ± 90 ° with respect to the incident angle of the second unit pattern.

FIG. 32 shows an example of a combination of unit patterns, which can be classified into about 360 types according to the above rule, and in this case, a curve that does not satisfy the above assumption is used as an exclusion pattern. Further, when the classification is performed, the number of pattern classifications, the sum of the distances between two adjacent points, and the sum of the angles are held (steps S121 to S128 in FIG. 34). Then, in the one typeface, after the completion of the first step, the pattern having the corresponding number of patterns is patterned. When the number of lines i and j is very small in FIG. 32, they are represented by adjacent thick line patterns. This process is performed so that the number of patterns is 64 or less and numbered (second step), and after the final pattern is determined, the average distance and average angle of each unit pattern, that is, polar coordinate data are calculated (step 129 in FIG. 34). ~ S131). here,
Assuming the average distance Li, the average angle angi, and the coefficient k, the statistical xy lengths dxi ′ and dyi ′ of the unit pattern are given by the equation (11).
Can be obtained by

Tan ⁻¹ (angi) = dyi ′ / dxi ′ (11) dyi ′ = kdxi ′ (12) dxi ^'2 + dyi ^{' 2} = Li ² (13) Therefore, according to the above process, (first) It is possible to form a pattern file of (x, y values of unit pattern + x, y values of second unit pattern + x, y values of third unit pattern) * the number of patterns (step S132).

Next, the case of expanding the original source outline font data will be described with reference to FIG. 35. First, as shown in FIG. 23, in the curve format, the point P0 is 1) motoro 2) curveto 3). lineto, points P1 and P2 are control points, and point P3 is cur
Since it is veto, it is easy to extract data only from the curve.

In the rewriting process, first, the original curve is expanded and the curve is translated so that the point P0 becomes the origin (steps S141 to S145), and the first from dx0, dy0.
First, the angle of the unit pattern is calculated to determine the placement information.
Rotate the pattern so that it is in the first quadrant (step S1
46), the pattern number is determined by the first to third unit patterns (step S147). Then, the difference between the xy value and the original pattern is calculated, the pattern information is added before the data of the point P1 (step S148), the difference from the original is encoded (step S149), a pointer indicating the beginning of the character data, The data length and the like are rewritten (step S150). When this process is performed for each character, the file is closed (step S151) and rewritten with a new file (step S152).

Further, as shown in FIG. 36, similarly to the third embodiment (FIGS. 21 and 22), the layers are grouped and replaced by the replacement table (steps S161, S).
162), the compression efficiency can be improved by the lossless encoding. Next, the decoding process will be briefly described with reference to FIG. 37. When a character number is designated (step S171), lossless decoding is performed (step S17).
2) After decoding for each layer by the replacement table (steps S173 to S175), the additional information is read (step S176) and the designated pattern is rotated (step S).
177), by obtaining the xy coordinate values of the source data (step S178), the data is expanded to outline font data at the new coordinate values (step S179).

Here, there are two methods of calculating the difference, depending on how the reference is taken. That is, as shown in FIG. 38, after the start point (thin line) of the original curve and the pattern start point (thick line) are made to coincide with each other, a method of handling the entire pattern as shown in FIG. It is conceivable to move the reference point for each point as shown in FIG. Further, one pattern has three unique distances L0.
Since it has ~ L2, it may differ greatly from the original, but by scaling the second and third unit patterns according to the ratio with the distance of the original first unit pattern,
This problem can be solved. In addition, the compression efficiency can be further improved by combining the third embodiment and the fourth embodiment.

[0072]

As described above, according to the first aspect of the invention, the figure extracting means for extracting the outline font data of the figure capable of sharing a part of the character, and the figure extracted by the figure extracting means. Restoration information calculating means for calculating information for restoring the original character by means of, information adding means for adding information indicating that the character is composed of the figure extracted by the figure extracting means, and the figure extracting means Since the graphic extracted by the above, the restoration information calculated by the restoration information calculation means, and the compression means for compressing the information added by the information adding means in a predetermined format are compressed, the character is decomposed. It is possible to reduce the amount of memory for holding the outline font data.

According to the second aspect of the invention, the extraction means for extracting the data exceeding 1 byte from the outline font data of the character, and the X-adic number which preliminarily calculates the spread of the distribution of the appearance frequency with respect to the data width of the character are calculated in advance. Since the X-adic number calculating means and the compressing means for compressing the data extracted by the extracting means with the X-adic number are provided, data exceeding 1 byte is compressed with the X-adic number, and therefore the compression efficiency is improved. Can be improved.

According to a third aspect of the present invention, a hierarchizing means for hierarchizing the character outline font data for each information, and a calculating means for calculating the appearance frequency of the data of each hierarchy hierarchized by the hierarchizing means. And a compression unit that creates a compression replacement table for each layer based on the appearance frequency of each layer calculated by the calculation unit and compresses each layer, compression based on the appearance frequency of each layer. The compression table is used to compress each layer, and thus the compression efficiency can be improved.

According to a fourth aspect of the invention, grouping means for extracting the curve data of the original outline font data in advance to group the outline font data, the original outline font data and the grouped outline font data. Since the compression means for calculating and compressing the difference is provided, the compression efficiency can be improved.

The invention according to claim 5 is the invention according to claims 1 to 4.
Since the code compressed by the compression means is further provided by means of lossless encoding, the original data can be decoded by lossless decoding.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an outline font data compression apparatus according to the present invention.

FIG. 2 is an explanatory diagram showing a main body and a crown of a character.

FIG. 3 is an explanatory diagram showing an offset of outline font data.

FIG. 4 is an explanatory diagram showing an outline of a restoration operation.

FIG. 5 is an explanatory diagram showing an overall configuration of outline font data.

FIG. 6 is an explanatory diagram showing base points of outline font data.

FIG. 7 is an explanatory diagram showing a difference in outline font data between the conventional example and the first embodiment.

FIG. 8 is a flow chart for explaining a compression operation of the first embodiment.

FIG. 9 is a flowchart for explaining a compression operation of the first embodiment.

FIG. 10 is a flow chart for explaining a restoration operation of the first embodiment.

FIG. 11 is an explanatory diagram showing lossless encoding.

FIG. 12 is an explanatory diagram showing a difference in data between the conventional example and the second example.

FIG. 13 is a block diagram showing a second embodiment.

FIG. 14 is an explanatory diagram showing data in units of bytes.

FIG. 15 is a flowchart for explaining a compression operation of the second embodiment.

FIG. 16 is a flowchart for explaining a compression operation of the second embodiment.

FIG. 17 is a flow chart for explaining a restoring operation of the second embodiment.

FIG. 18 is a block diagram showing a third embodiment.

FIG. 19 is an explanatory diagram showing the appearance frequency of outline font data.

FIG. 20 is an explanatory diagram showing an appearance frequency of a flag hierarchy.

FIG. 21 is a flow chart for explaining a compression operation of the third embodiment.

FIG. 22 is a flow chart for explaining the compression operation of the third embodiment.

FIG. 23 is a flow chart for explaining a restoration operation of the third embodiment.

FIG. 24 is a block diagram showing an example in which the first to third embodiments are combined.

FIG. 25 is a block diagram showing a fourth embodiment.

FIG. 26 is an explanatory diagram showing a Beze curve.

FIG. 27 is an explanatory diagram showing a unit pattern.

28 is an explanatory diagram showing each unit pattern of FIG. 27. FIG.

FIG. 29 is an explanatory diagram showing a first unit pattern.

FIG. 30 is an explanatory diagram showing a second unit pattern.

FIG. 31 is an explanatory diagram showing a third unit pattern.

FIG. 32 is an explanatory diagram showing an example of a combination of first to third unit patterns.

FIG. 33 is a first diagram of pattern integration in the fourth embodiment.
6 is a flowchart for explaining steps.

FIG. 34 is a second diagram of pattern integration in the fourth embodiment.
6 is a flowchart for explaining steps and pattern file creation.

FIG. 35 is a flow chart for explaining a source font file rewriting operation in the fourth embodiment.

FIG. 36 is a flow chart for explaining a compression operation in the fourth embodiment.

FIG. 37 is a flow chart for explaining a decoding operation in the fourth embodiment.

FIG. 38 is an explanatory diagram showing a difference calculation method.

FIG. 39 is an explanatory diagram showing points of a Beze curve.

[Explanation of symbols]

 11 shared figure extracting means 12 restoration information calculating means 13 information adding means 14 compression means 15 lossless encoding means 21 data extracting means 22 X-adic number calculating means 23 X-adic number compressing means 31 layering means 32 appearance frequency calculating means 33 layer-wise compression Means 41 Curve grouping means 42 Error compression means

Claims (5)

[Claims] 1. A figure extracting means for extracting outline font data of a figure capable of sharing a part of a character, and restoration information for calculating information for restoring an original character by the figure extracted by the figure extracting means. Calculating means, information adding means for adding information indicating that the character is composed of the graphic extracted by the graphic extracting means, graphic extracted by the graphic extracting means, and calculated by the restoring information calculating means An outline font data compression device comprising: the restored information and a compression unit configured to compress the information added by the information adding unit into a predetermined format. 2. Extraction means for extracting data exceeding 1 byte from character outline font data, and X-adic number calculation means for pre-calculating an X-adic number such that the spread of the distribution of appearance frequency with respect to the character data width is reduced. An outline font data compression apparatus comprising: a compression unit that compresses the data extracted by the extraction unit using the X-adic number. 3. A hierarchizing means for hierarchizing the character outline font data for each information, a calculating means for calculating the appearance frequency of the data of each hierarchy hierarchized by the hierarchizing means, and the calculating means. An outline font data compression apparatus, comprising: a compression unit that creates a compression replacement table for each layer based on the calculated appearance frequency of each layer and compresses each layer. 4. A grouping means for extracting curve data of original outline font data in advance to group the outline font data, and calculating a difference between the original outline font data and the grouped outline font data. A compressor for compressing outline font data, comprising: compression means for compressing. 5. The outline font data compression apparatus according to claim 1, further comprising means for further compressing the code compressed by said compression means by reversible encoding.