JPS6075975A - Processing method of character picture data - Google Patents

Abstract

PURPOSE:To obtain the compression data faithful to character contours by dividing the contours into blocks to obtain a vector along the contour of a character picture for each block and performing the coding process with approximation of a section polynomial according to said vector. CONSTITUTION:The supplied character images are decomposed 30 into dots and undergo the contour extraction 31. The extracted contour receives the linear approximation 32 with a vector and is identified 33 between a linear part and a curved line part. Then the dividing points of the contour are detected 34 while obtaining the cross angle of adjacent vectors. At the same time, the inclination of each contour point is detected 35 for approximation of the curved line. Thus an n-degree polynomial is obtained from the both-end coordinates of a section receiving the approximation of curved line and the sloping of said section. Then the sample sections are successively decided 36 and the inclination of each contour point is calculated 37 again. An n-degree polynomial which is fathful to the contour is calculated 38 from the inclination of each contour obtained newly and the coordinate value. Then these coefficient, degree numbers, etc. are coded 39 and stored 40. This compression data is decoded 41 by a decoder and a desired character contour is restored 42.

Description

[Detailed Description of the Invention] 1 Komyo's Technical Field 1 The present invention relates to the universal expression of character/image (hereinafter referred to as character) data, and in particular, to approximating the outline of a character with a function curve or a set of straight lines. , its outline specific information 'tfA 1L! The amount of data is compressed by memorizing the data, and the compressed data is converted to VN to regenerate the character image. (be,.

1. Komyo's So-Notice Technique 1. The 2-ship data obtained by decomposing the characters into 4-C redundancy 1-1 is a high A'-Yu (this is well known).

Therefore, in order to reduce and overcome this redundant fMiff, various data compression formulas have been proposed. It has been proposed.

One of the data compression methods is the so-called contour method, which compresses the amount of data by grasping the shape of a character by its outline and storing the contour specifying information.

As a data compression method using this contour θ, a straight line (vector) approximation method as shown in FIG. 1 and an r)-order curve approximation method as shown in FIG. 2 have already been proposed.

The straight line approximation method illustrated in Fig. 1 is
This method is disclosed in Japanese Patent Application Laid-Open No. 522-79154, etc., and its outline is to approximate the contour 1 of an arbitrary character indicated by a dotted line with a set of two-dimensional vectors 1 to 2 indicated by straight lines. The specific information (starting point position, length and (
By using the horizontal and vertical direction components as stored data, data Ff compression can be overcome with 711 functions.

In addition, the n-th curve approximation method illustrated in FIG.
No. 39963), it is a method that has already been applied for, and
11N The point is to set a point p a appropriately on the outline of an arbitrary character.
By memorizing the y coordinate, the amount of data can be reduced, and by connecting arbitrary consecutive (n + 1) points, n
This is a set of curved elements 3 that overlap in order to approximate a desired contour.

When reproducing a character image by converting the compressed data of these contour methods into text images with various magnifications, interpolation processing or vector magnification conversion processing is necessary. :3-' It has the characteristics of being able to deal with image reproduction.

1 Problems with 114 techniques] However, on the other hand, these conventional methods, for example,
One endpoint of each solid circle in Figure 43L), or
It is clear from the fact that the angle δ, which is the slope of each line segment on the left and right, centering on the connecting point p of each r)-order curve 3 in Figure 2 c134J as the center, is b'4 continuous in any case. Like,
Regarding the 811 Rakaza (continuity of the slope of the contour) of the contour,
Eliminate water quality deficiencies that do not guarantee optimal results.
() (Iko.

On the other hand, in general, the contour shape of a character is just a series of two contour white bodies, and furthermore, at the intersection of character strokes, the tip of the wing 1, etc. 0 except Tatsumi Chosada
In the case of a child, its first derivative (slope of the contour) changes continuously! I have lzf symptoms.

Therefore, with the data rI = reduction method based on the conventional contour method, it is difficult to obtain compressed data that is faithful to the character contour, and it is difficult to obtain compressed data that is faithful to the character contour. The problem was that it could not be removed.

In order to solve this problem, the applicant of the present invention filed Japanese Patent Application No. 57-16884 (Japanese Unexamined Patent Publication No. 58-13/1745).
An application has already been filed for method No.

However, the data compression method disclosed here approximates the entire contour at once, so in contour parts where straight and curved parts are continuous, unevenness that deviates from the contour occurs at the connection point. It's easy to do.

Also, to avoid these problems, set more sample points, divide into many polynomials, and perform approximation (
As a result, the problem that the amount of Fi reduction data increases has become clear.

(2) Furthermore, since compound interest calculations are required to store compressed storage data, there remains the problem that it takes time to create storage data.

[Object of the Invention] An object of the invention is to provide an improved contour method data compression method, and another object of the invention is to reduce the calculation time when completing calculations on compressed data. Nevertheless, it is an object of the present invention to provide a method for processing character image data that is faithful to the character contour <I: rl reduced data, and furthermore, another object of the present invention is to I have developed a method for processing character and art quotient data that can obtain a sufficiently high data compression rate while faithfully memorizing the smoothness of 1.
Another object of the present invention is to decode the data compressed by the present invention and reconstruct the character image.
When creating a character image, character images with round outlines at various magnifications are faithfully reproduced.

I +11 Essentials of the Invention The method of the present invention is related 1) II ■ Essentials of data processing related to 3
The explanation will be given according to Flowch Oni-1 shown in Fig. 3.

Fumiko manually made (αZ is l:x, y171 helix shape 1
-t is decomposed (30), and the ring ν1 of the decomposed character image is extracted (31). 4f so that
Linear approximation J-ru (32) using vectors. A contour approximated by a straight line is divided into a straight part and a curved part based on the quality of the solid surface (33), and on the other hand, the dividing point of the contour is determined by determining the intersection angle of adjacent vectors. (34).

Then, for a curved part, the contour shape of the section is expressed by an n-degree polynomial (n=1), and for a curved part, the contour shape of the section is expressed by an n-degree polynomial (n-2, 3>). Curve approximation. In order to approximate the curve, find the slope of a3 +) at each contour point forming the contour (35), and use the coordinates of both ends of the section to approximate the curve and its slope to calculate the n-th degree polynomial. Ru. and,
While extending the section, find the longest section (lean pull section) within the range where the offset position between the n-th degree polynomial that approximates the section and the contour is within the tolerance tolerance, and sample both end points of the section. It is difficult to determine by setting <<3G). In this way, once we have found the n-dimensional polynomials that approximate each nipple interval by sequentially determining the ripple intervals, we have determined the first-order derivative of each n-many In expression and re-extracting the slope of each contour point ( 37). Then, based on the slope and coordinate values of each newly determined contour point, the rimple points are determined in the same manner as described above, and a 13 nth degree polynomial that is more faithful to the contour is created (38). In order to code the coefficients, degrees, etc. of the n-dimensional polynomial thus obtained,39) and to efficiently restore the character wheel Of+, 1
(distance between the starting point and the ending point in the direction of the
Each block data is stored in the order of 1 and 0. 1
I made a sea urchin (/10). It is memorized by the letter #Akatsu 6
The ffl1ii data is distributed among multiple f-1-ders and the desired character 91 is restored based on the decoding result L↓. J, -) t
Koshi lko (42).

[-Embodiment 1 of the invention Next [Details about the data processing of each IS]
Ill'Iru.

Input 1 image (30>') Raster 5L of 19hi η: =)
Then, the image of Bunzai 1111 is interpreted as ['x, y171 ~ dotbun], and j, is requested, and l is pip1.
Hepattern Data/JXk! As the A-riginol character data that becomes the L logic (Jξ supply 3).

(Contour extraction (31> 1 The 2 1iQ data corresponding to the decomposition driver 1- is Dot position that changes to (outline 7)
gtV) to cap the contour. Divide the contour obtained by J into blocks of monovalent functions with X as a variable and make a set of multiple blocks.

Figure 4 shows how to extract a contour from character data decomposed into matrix-like
(dli function block (('0]h\ et al. '0.1
It is divided into 2 sections).

[Line approximation, (32) Co-wheel! For any one block of /l1-1- (13, the large number σ is set to be as long as 4'J as long as the deviation from the contour is within the predetermined r1 volume error)
Perform linear approximation using vectors.

FIG. 5 shows a linear approximation of the character outline 50 indicated by the dotted line by a set of two-dimensional vectors 51.

Connection point P1 of each vector 51. [)2. . . , Plow 1°F)n is stored as a sample point in the sample point coordinate storage unit 60, which will be described later.

[Identification of the straight and curved parts of the letter ring (33)] As mentioned above, generally speaking, the outline shape of a literary figure (which has a straight part and a curved part) is J'3. Tsu,?+U coming (
Because we tried to approximate the body at once, we created a ring 1ll (with cl
(Yes, there are +5 near the 5 points and many 1!
(This is done by setting up the main idea and dividing it into many approximations:
L'4: Jre+, r<r<r<<, the data IIl increases, resulting in a disadvantage that the reduction rate decreases.

(The present invention is based on −.) [By linear approximation,
Depending on the length of the solid line, the outline shape is a straight line or 11
105) The above-mentioned drawbacks are resolved by identifying the straight portion and the curved portion, and processing the straight portion and curved portion separately.

FIG. 6 is a block diagram showing an example of a configuration that realizes a method for identifying straight and curved portions of a character outline.
1 -C160 is a sample +j that stores the sample points obtained by J, -) in the linear approximation (32) in units of blocks.
, ' are coordinate storage units, 61 is a straight line section identification vector length setting unit for setting the identification vector L-, 62 is a curve division point identification angle setting unit for setting the curve division point identification angle tUθ, 631
J: Coordinates of i+1st sample point P L+1 (x l
, +l, 'J 1. +1), and 64 is the next sample point coordinate register for holding the coordinates (+1) of the first sample point Pc.
×C.

current sample point coordinate register for holding yL), 65
is the coordinate of the i-1st sample point P1-1 (X t-1゜y
t-+), a previous sample point coordinate register for holding 6
6 +J, +if from each coordinate in the sample point coordinate register 63 to the current sample point coordinate register 64.
A vector length calculation unit 67 calculates the vector I to length β of X 14 ], and 67 calculates the coordinates J: of the next sample point coordinate register 63, the current sample point coordinate register 64, and the previous sample point coordinate register 65. A vector-to-vector angle calculation unit 68 that excludes the vector I and the angle θL at the sample point PL (also calculates the identification vector length set by the straight line part identification vector length setting unit 61 and the muk 1 to line lengths); 69 is a curve division point identification angle θ set by the curve division point identification angle setting unit 62, which is a curve division point identification angle θ and nlr vector. An angle comparison unit 70.71 compares the angle θ between the vectors sieved by the sieve 7, and 70.71 compares the straight section or the curved section based on the comparison result of the vector length comparison section 68. Straight line part memory i'jl
, III Kl 7'+li storage section 1,72tj
, is a curve division point coordinate storage unit that stores the first sample point 1] where θ>OL as 11 division points in the six angle comparison units.

Next, regarding the operation, F, +2 light J-ru. First, the straight line segment identification I\r1-ru angle setting unit 61 sets the identification vector angle, and the curve segment N+'1 point identification angle setting unit 62 sets the curve 《) division point identification angle 0, respectively. Next, specimen Sada P16
4 sects [from 200 million parts 60, 1! [For the start point of the first block] Zu1() is sent to the next sample point coordinate register 63.At this point, the current (♂(for the starting point of the block 0/l) does not store anything, so the In Muku 1
- You cannot do 51a). Next, the sample point coordinates stored in the next sample point coordinate register 03 are stored in the current Mukuki point coordinate register 6'l, and the next sample point coordinate register 63
3, the next sample point coordinates are newly stored. From now on, the sample point coordinates of the current stock point coordinate register 64 are stored in the Maegura main point coordinate register 65, and the sample point coordinates of the next sample point coordinate register 63 are stored. The coordinates of the next sample point are stored in the coordinate register 64, respectively, and the coordinates of the next sample point are stored in the coordinate register 63 of the next sample point.

The 1st to 3rd calculation unit 66 calculates the coordinates (X 1.
+I, V+, +1 > and current sample point coordinate register 6
Coordinates (XL, Vt,
), vector I'i +2 +, = F〒:r〒
-150 ■ Go + yt -y and +1>2 is impaired. The vector length βL extracted from the camphor is Muku 1 to l1 comparison part 0
8, the curve section identification vector length setting section 61 sets the length in advance! , compared with identification length. here,! 〉1. In this case, the section [Xl, country
Ru. ;l: Ta-! When ≦1-8, +, Il, l' X
t, , Here, if the next section is a straight section, the above-mentioned memorized [XL,
X1. +1 ] as a curved section, and its 1 point coordinates and end point coordinates are stored in the curved section storage section 71. On the other hand,
If the next section is also a curved section, the next section is further identified, and the section to be identified is curved; [4 consecutive spaces: as one curved section, its starting point] J! ! The target and end point coordinates are stored in the curve section storage section 71.

[Divide the contour (34) 1 1, vector] In the square melon sieve section 67 between
Jiusa Zhang Register 63. Current sample coordinate register 6/l.

Previous school: From the main point coordinate register 65, each sample point l'
1lt-+ , F:'t, , P t, +1 coordinates (xt-t, yt-+), (xL.

Vt ), (XI, ÷l, y1.+1) are read out, and the angle θ ((θ is an acute angle) between the vectors 1 and 1 shown in FIG. 5 is calculated.

C) The angle θ (between the curved lines 1 and 2) is set in the angle comparison section 69, and the curve dividing point identification angle setting section 62 (preliminarily set). Here.
C1θ, · OL, new sample point PL to curved part 21
11 points and store the coordinates iid in the curve dividing point coordinate storage section 72.

Normally, there are many sample points near this curve formula IJ point, and it takes time to process the approximate curve.
By dividing the contour at the curve division points obtained, the curve approximation is performed separately before and after the curve division point, so the processing time is quick and the approximate curve can be easily obtained.

FIG. 7 shows the character outline data shown in FIG.
Above [linear approximation < 32)]-F contour division (34)
] An example of data processing is shown below.

In the figure, ◯ indicates the start and end points of one block, △ indicates sample points obtained by linear approximation, and ◯ indicates curve dividing points. J
, Ta, * indicates a straight part, no mark indicates a curved part.

[Calculating the slope of each contour point (35)] The contour shape of the curved part obtained above is calculated using an n-dimensional polynomial (where n=
When performing approximation using 1.2.3), the n-th degree polynomial is determined in a simple manner once the values and slopes of the two points are determined. In this case, it is first necessary to give each contour point on the contour an inclination of C5.

In the present invention, the slope at each contour point is determined by extracting a predetermined number of contour points before and after each contour point, and connecting the contour point whose slope is to be calculated and each extracted contour point by a line segment ( 9) Find the slope at the desired contour point.
I did it.

A method of calculating the slope of a-3 at each contour point according to the present invention will be explained in accordance with FIG.

First, the outline of Figure 8 (a) - 12 Q1 block [
1 tsuk 1. To add an inclination of 1.1 to the starting point P+ (contour point Q1) of P+, Pnl, extract an arbitrary number of contours with (f) after the starting point P+, and connect the starting point P1 and each contour point. Calculate the slope of the line segment connecting t), and use the starting point 1]1 from each slope.
The slope tl of C5 can be calculated by 9fjt-J from the formula 1-, which will be described later. .

For example, when extracting two contour points and calculating the slope, the slope rnl of the line segment Q1Q2, the slope rnl of the line segment 6-
Set T12.

Line segment (trisa rTl, CJ:, two points & (
C can be found from XI + yl) + (X2, V2). Incidentally, the case of C2 is also concluded in the same manner.

When the slopes m, , m, , of the line segments are completed, the slope 1 of the contour point Q1 is determined by 3.Even if the end point is C3, it can be determined in the same way.

Next, in Fig. 8(b), J: a lifting platform that calculates the inclination tl at the sea urchin next contour point Q2. :Find the slope of the line segment connecting each contour point, and add C5 to the contour point Q2 (set tl).

For example, when extracting two contour points before and after (σ1), only the front one point is extracted at contour point Q2, so
In this case, specify 7 to extract 1 point before and after.
．． - If there are no contour points, extract the maximum number of contour points within the specified range and smooth the slope.

Next, when adding a slope T3 of J3G to the contour point Q3 as shown in FIG. 8(C), at the contour point Q3, 111 white 1
．． A predetermined number of contour points exist before and after. j; Tsute outline L
;j Q 3. Slope ml of each line segment connecting each contour point.

Find mz, m3, rT14, and calculate J at contour point Q3. , and calculate the slope at each contour point of 1 C in the same way while sequentially passing the contour points.

1 Ripple Point Determination (36)'] Once the slope at each contour point is equalized as described above, then
An approximate curve determined by two contour points of 1 monme is determined, and the amount of deviation between the approximate curve and the contour is determined one by one for each contour.

Each 1 of Zoshi-(,t) in the approximation curve section
If the error is less than 1ga, move the contour point forward by one more point, repeat the same process again, and determine if the deviation IJ is the longest within the n'B error range. While determining the interval (hereinafter referred to as the -lean pull interval),
Divide the contour into rimple sections. Then, the contour is approximated by an n-dimensional polynomial determined by this rimple interval.

Hereinafter, the calculation method of the approximate curve of the n-th degree polynomial and the calculation method of the amount of deviation will be explained using FIG. 9.

First, it is determined whether the approximate section is a straight section or a curved section, based on the coordinate values stored in the straight section storage section 70 and the curved section storage section 71. Then, if the section is a straight section, calculate the starting point coordinates of the section and a linear equation (linear equation) indicating the section,
After) Convert to a host.

Next, a case where the section is a curved section will be explained.

In FIG. 9(a), 5L (Ql') is the starting point of the first sample section of interest, and the starting point SL (
Ql > and 4 non-pull candidate points S L+1 ((,) n
7. Sequentially draw an approximate curve between the two points in ) as one non-pull candidate section. B is the contour point QJ, C is the intersection of the curve approximation J-, and intersects with the straight line 92, and the point B (QJ)
13Δ is the deviation amount in the X direction between the contour 90 and the approximate curve 91 at the contour point QJ, BC is the deviation amount between the contour 90 and the approximate curve 91, BD is the deviation Ml in the X direction of the contour 90 and the approximation J3iiDI.

Also, 1. ' Q + , Q 2 ] and ['
Since the nth degree polynomial of J is lazily determined to be a straight line, an approximate curve of the next interval [Ql, Q3], J, sr) degree>term is found.

The following will explain the case of a cubic curve as an example. The already calculated starting point 5c (Ql) and sample candidate point Qn (Z1kit
L, t: l and coordinate values (XL, Vt >, (
By substituting Xn, Vn) into the following cubic equation, an approximate curve 1<x) of the section is obtained.

j' (x)=yt (bt, (x-xc) ICt,
(x - X L ) 2 10 dt (x - XL): 1 1) (, -1,.

After determining the approximate curve f(x) as described above and summing the overlapping coefficients, the amount of deviation ε between the approximate curve f(x) and the contour is calculated.

The amount of deviation ε is determined by each contour point QJ existing in the section (however, 2
≦j≦n−1).

In the figure, the amount of deviation ε(=B
In calculating C), the present invention calculates the deviation amount ε in a pseudo manner, thereby enabling high-speed processing.

FIG. 9([)] is an enlarged view of the error portion between the contour 9o and the approximate curve 91 in FIG. 9(a).

In the figure, it is assumed that the line segment plane is parallel to the straight line 92,
Let the slope m of the line segment CΔ be the slope of the straight line 92. At the contour point B (QJ), the distance εχ from the approximate curve 91 in the X direction and in the X direction.

By setting ε〃, the desired deviation amount ε can be approximately determined as follows.

This J, the approximate deviation of the contour 90 and the approximate 1fi191) °! If φε is within the range of F7/A difference υ11, extend the section using the next ring Of+ +:, IQ no as the rimple candidate point, and create a new section [Q +, Q no 1, put 1 and fit an approximate curve, and the ring Of+ +:, i
For Q J, increase the deviation amount ε by J1'1.

If the h'1 capacity error is exceeded at any contour point, the previous 1-pumple candidate point Qn is changed to 4 non-pull points S country and the first lean-pull section [SL].

8 countries and 1 will be decided.

Thereafter, the rimple point country S will be designated as the starting point of the next rimple section.
, In the same manner as described above, the sample points are determined one after another and the ripple interval is determined from L.

In this way, the contour is divided into multiple sample intervals by N+
11, and the outline is approximated by approximating each rimple section with an approximate curve.

In addition, in the explanation in L, the following υ sample candidate interval [(1) +
, if the deviation amount ε in Q no l exceeds the n-volume error, one lean-pull candidate point Qη is used as a sample point to determine a sample interval.

However, in Sen's example, when W [exceeds the tolerance error,
Q is the lean pull candidate point at that concave point. , 1 is temporarily memorized, and several points ahead (Q n+z + Q I
The value ii'f' of the deviation amount ε is incremented by i for the sample candidate section up to (l+:1, . . . ).

Then, even if a predetermined number of points are reached, if the error exceeds the Y[tolerance error, the sample section may be determined as a non-pull point.

Of course, in this case, r
``When the tolerance condition is satisfied, the candidate section is updated to a new section [Q+, Qn], and from there, the above evaluation process is repeated while moving forward by 4 sample points. By implementing such look-ahead, the sample interval can be further reduced, and the data compression rate can be improved.
] As mentioned above, when the contour is approximated by a cubic approximate curve in a3 in a plurality of sample sections, the starting point coordinates of the quantity sample section and the order and coefficient of the approximate curve approximating the four non-purple sections can be calculated as follows. Contour data R1: By memorizing it, it is possible to obtain 1 [old man 6 and l ko 2] -do.

However, the ratio of each contour point determined by the above [Inclination of each contour point (3!+)'l is the average value of the slope of a predetermined number of contour points before and after it, and in reality There are cases where (G1 SATO is human < V4. Therefore, in the present invention,
Furthermore, in order to faithfully approximate the contour, (
] Find the -th derivative f'-(x) of the approximate curve f(x),
In the non-pull section, each wheel was set at 0ft5, and the slope at the end of the wheel was made to be more consistent.

[Redetermination of Rimple Points (38)] 1-Redetermine the slope of each contour point (37)].
:Depending on the ship, MoejiL4 [Summer Ik] decision (
36) In the same manner as above, determine the sample interval while determining the severe lean pull point. This J: point obtained in the above manner is approximated by an approximation curve f(x) that is faithful to Niriwa ν1; in the sample section A.

[] - J3 to the straight and curved parts that were claimed like soil.
Coordinates of the starting point of each Zamble section, approximate curve f(x
) are converted into :I-codes and further stored as a set of block data organized in units of blocks, it is possible to obtain compressed data that is faithful to the outline of an arbitrary character.

FIG. 10 is a diagram illustrating an example of an arbitrary block (preferred data storage of J-block data) A-Matsu 1 to be applied in carrying out the present invention as described above.

In the figure Ichino A-Matsu 1, the block header is:
The end coordinate of one block and the number of "non-pull sections existing in one block" are memorized, and the segment cutter memorizes the x and Y coordinates of the starting point of one re-pulse section and the order of the approximate curve "<x". , the information for legmen 1 stores each coefficient of the approximate curve f'(x) based on the above-mentioned order.

Then, one lean pull section data is constituted by legmen 1 to hengu and leg information, and furthermore, lc clean pull section data is sequentially arranged according to the number of V pull sections, and the whole constitutes one block data. There is.

[Storage of compressed data (40)] Each block [l
In order to efficiently restore the character contours described later, you can use the block data by pressing ◇ and ◇ for each block.
The decoding process is performed by ∆ huffing on r:l and is memorized in the order of the spacing.

For example, in Fig. 11(a), the character 1'AY>, the 8th mark is ticked for 1.
We accept requests for compressed data for each number (indicated by number).

For example, the amount of time it takes to decipher the
For 1113, each part is memorized in order from the starting point of 1st tsuku and the distance of the final J qi distance b). Figure 11 (
E]) aj, the time required to decipher 1 block [1 block is the best is 1] block 12, and the following block [1 block 10. Block 16 follows, and block 3 is the most M4 <
<I'm here. Therefore, the final character wheel Of
The compressed data is stored as a set of h block data as shown in FIG. 11(b).

[Dispersion Denied] (41>) What has been discussed so far is related to data compression of character contours. Next, a method for restoring the original contour based on the contour f compressed by the data processing method described above will be explained.

In this case, in order to restore the contour quickly, the block data of one block angle is sequentially transferred to multiple data points for processing, and the processed data is transferred to the next data point. Processed block data. Therefore, the time required to decode one block mentioned above becomes a problem.

In other words, if the time required to decode one block is approximately tJ, then the processing time for restoring the contour usually becomes longer. J:
So, when block data that takes a long time to process comes later, that decoder () continues processing (), and all other decoders finish processing, which reduces the efficiency of decoder usage. Decline cover.

Therefore, the present invention stores compressed data in descending order of the time required to decode one block, as mentioned above.
By processing with multiple decoders, the decoders can be used efficiently and contours can be restored at high speed.

12 is a block diagram showing an embodiment for optimally carrying out the contour restoration according to the present invention.

In the 1st block 1, 120 is a compressed data storage unit that stores each block of data in descending order of the time taken to read 1 block of M data, and 121 is a compressed data storage unit that stores each block of data in descending order of the length of time it takes to read 1 block of M data. The SE1 node 1.122 stores the desired magnification that is input separately.
) decoder IIY, 1''24 selects the decoder for which processing has been completed from among the decoder group 123, and transfers the decoded and requested contour pixel data to the one-character storage unit described below. The cell 1 noctor 2.125 describes the contour f-ta obtained by jx from the 1? rector 2 (124>).
1 character storage unit, 126 is +1ij 1 character storage unit 1
This is an output device that prints or displays a character based on the outline pixel data for one character that has been stored in the memory 25.

Next, O's work will be explained.

From the compressed data storage unit 120, in order of the time required to decode one block (for example, in FIG. 11, block 12
．． Block 10. block 16...) is transferred to the director 1 (121). The selector 1 (121> selects and transfers the transferred block data to decoders in order from decoder 1 of the decoder group 122.The decoder that has finished transferring the block data transfers the transferred block data to the 4F< *storage unit 122 Desired A8 separately memorized in
After reading the ratio data, the contour restoration process begins.

Selector 2 (124> has completed the restoration process) Data 1-
The characters are sequentially selected and the restored outline pixel data is transferred to the single character storage section 125. The decoder that has completed the restoration process for 1 block [1 block] requests the selector 1 (121) to transfer the block data, and the block data for the next 1 block (1-1 block) is transferred.

FIG. 13 is a timing chart showing an example of the processing state of each decoder group 123 shown in FIG. 12.

In the figure, -'1 is from selector 1 (121> to each -f
-7) - Taheb [Time to transfer one piece of data, l-,
,4;L The contour pixel data decoded by each denigger is se1
1! to the 1 character storage unit 125 via Nokta 2 (124).
/: There is some time to send T3 to each decoder ().
The processing time for decoding the 1-1 piece of data will vary. jkota, - "4 indicates one character's worth of processing Jj1 hour.

'j' -l -'l 07179 minutes of block data/J' is transferred to 1-n and processing starts.
The first 11 block data is transferred to the shortest data T1 between processes 0 and r (in this case, decoater 3), and continues to process J (11 block 41).

However, each block of data to be transferred 61 times is arranged in the same order as described above, with the 5 questions gradually becoming >, i7<, so each ductor, as shown in FIG.
1 sentence 2 before Habo Dou? Ends processing for the minute.

If the decoder 11 and 1 are used, each decoder operates evenly, and the decoding of the character data is completed as quickly as possible without wasting any wasted play.

Restoration of L contour ('12)] As mentioned above, one block of contour pixel data restored by each decoder is sequentially stored in the one character storage section 125, and the contour pixel data for one character is stored one by one in the one character storage section 125. Completed J-ru. and,
The outline pixel data for one character stored in the one character storage unit 125 is stored in a laser beam printer, CRT'Lj, for example.
- By supplying an output device 126, such as a true typesetting machine or a display device, the desired characters are generated in the same manner as in the prior art.

[Effect of the invention] The processing power method of the character image f-Yu according to the present invention was applied to the Mincho body deck name character [A, 1] consisting of 800 x 800 dots 1 to 1 As a result, if the π1 volume error for the desired character image is taken as 1 dot, it is 1.21.
The data pressure rate can be reduced to 7%.

Further, in the above explanation, μ2 has been explained using # and L "a" as examples, but it is clear that images of various marks, symbols, line drawings, etc. other than kanji and letters can also be handled in the same way.

As described above, according to the present invention, data that faithfully stores the smoothness of character width 91 can be obtained at high speed with a sufficiently high compression ratio, and when restoring to history, high-quality character images can be obtained at a magnification of F/(. It is possible to provide a processing method for a character image printer that can be reproduced at high speed.

[Brief explanation of drawings]

FIGS. 1 and 2 are diagrams explaining the conventional contour method data compression, and FIG. 3 is a diagram showing the essential points of (1λ) for data processing related to the method of the present invention. - Figure 4 shows the outline
Figure 5 is a diagram illustrating linear approximation, and the negative '1G diagram is an implementation of the method for identifying straight and curved parts according to the present invention. Example 1]
Figure 7 is a diagram showing the identification results of the identification method of the present invention, and Figure 8 is a diagram showing the identification method of the present invention. Figure 919 is a diagram explaining the method for determining the rimple point in the present invention.
Fig. 0 shows an example of block data storage A-Matsu 1~, Fig. 11 is a diagram illustrating a block data storage method according to the present invention, and Fig. 12 shows an example of block data storage according to the present invention. FIG. 13 is a timing chart showing the operation of FIG. 12. 1.50.90... Character outline 2.51... Two-dimensional vectors 1 to 3... m-th order curved element 60... Sample point coordinate storage unit 61... Straight line portion identification solid 1 to Line setting section 62...Curve dividing point identification angle setting section 63...Next sample point coordinate register 6/1...Normal sample point coordinate register 65...Previous sample point coordinate register 6G...Vector length tS Tabe 67...Angle closing part 68 between solids 68...Vector length comparison part 69...Angle comparison part 70...Straight part storage part 71...Curved part storage part 72... Curve division point coordinate storage unit 91...N-degree polynomial approximate curve 120...1F reduced data storage unit 121...Se1 nokta 1 122...A8 rate storage unit 12C3...decoder JiY] 2/I...Rector 2 125]...1 character storage unit 126...Output device surface Figure 3 Figure 9 (α) (b) Figure flO (0) (b) Figure 12

Claims (1)

[Claims] In addition to storing encoded data for specifying the Ic character image ring DIX by developing it into ten (+), X, and V coordinates, the 1'
In a method of processing character image data in which the original character image is reproduced and overwritten by decoding code 3 data, the contour is divided into blocks of 11II11 functions where x is a variable, and a set of multiple blocks is formed. 4" and a large number of vectors sequentially extending from each block of line segments λ to the outline of the character image; If the amount of deviation from
- The difference between the two vectors is calculated within a range of t1. A series of contours corresponding to a part where the length of each vector is shorter than a predetermined length is regarded as a curved part. A method of processing character image data in which encoding processing is performed using piecewise polynomial approximation. (2) When performing polynomial approximation J'' by regarding one contour as a curved portion, group the curved portion into multiple groups as appropriate based on the intersection angle of each of the obtained Muku 1 hels that touch each other by v4, Claim 1 wherein piecewise polynomial approximation is applied independently for each contour belonging to the same group.
Processing of character image data described in section