JPS6075976A - Processing method of character picture data - Google Patents

Abstract

PURPOSE:To reduce the compression data quantity by obtaining an n-degree polynomial which gives approximation to a prescribed section from the inclinations at both ends of a contour section of the character picture data and the coordinates of a terminal point and storing both the coefficient and the order number. CONSTITUTION:The supplied character images are decomposed 30 into dots, and a contour is extracted 31. This contour undergoes the linear approximation 32 with a vector and discriminated 33 into a linear part and a curved line part. Then the dividing points of the contour are detected 35, and an n-degree polynomial is obtained from the both-end coordinates of a section to receive approximation of curved line and the sloping of said section. The sample points are successively decided 36, and the inclination of each contour point is calculated 37 again. An n-degree polynomial which is more faithful to the contour is calculated 38 from the calculated 37 inclination and the coordinate value of each contour point. These coefficient and degree numbers are coded 39 and stored 40 as the compression data. This compression data is decoded 41 by a decoder and a desired contour signal is restored 42.

Description

Detailed Description of the Invention [Technical Field of the Invention 1] The present invention (.11, Compression of character/image (referred to as "character") data in one system in the Buno style, 1J[, character outlines between functions (Or approximating it with a set of n lines and distributing its contour identification information.) Perform fE compression of data M, decode the compressed data, and reproduce the character image. 71g
, JuruJ, C on processing method of character image data
It's from Jurub.

Invention of Back Fan Technique 1 It is well known that the data that is created by decomposing a character into two pieces is highly redundant data.

Conventionally, various schemes have been used to reduce this redundancy.
A new compression method has been proposed.

Determine the shape of the letter 11 by its outline, and specify the outline information by h.
As I recall, one of the methods for compressing data h1 is the data Y line method, which is called the one-contour method.

This ring! The most recent data compression method is
A white line (vector) approximation method as shown in FIG. 1 and an n-th order curve approximation method b< as shown in FIG. 2 have already been proposed.

The straight line approximation method illustrated in Figure 1 is
This is a method disclosed in Publications No. 22, Ff-155-79154, 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 I to 2 indicated by straight lines. , the specific information of each vector (starting point position, length and inclination, or horizontal/vertical direction components) is stored in the stored data.
This makes data compression possible.

In addition, the n-th order curve approximation method illustrated in Figure 2 is applicable to 111g of the present case.
There is a method that has already been applied for in A.
By memorizing the coordinates of the group, data can be compressed, and the desired contour can be approximated by a set of n-dimensional curve elements 3 connecting continuous J-(n + 1) points. ,
It's very similar to that.

These contour method-based data compression methods perform interpolation processing or vector magnification conversion processing when decoding the compressed data and reproducing character images.J, -, > -C,
4: It has a Qb characteristic that can cope with character image reproduction at a cost of 1 degree.

1. Problems with backward technology] However, on the other hand, these conventional methods, for example,
(+i I:a P for l' of each vector in 1), or the connection point 1 of each r)-dimensional curve 3 in Figure 2)
Regarding the continuity of the slope of the smooth contour of VaOR, as is clear from the fact that the inclination angle δ of each line segment to the right of C is discontinuous in each case. However, the woody defects that do not guarantee optimal results are ignored.

On the other hand, in general, the shape of a character's ring r71 not only has a continuous outline, but also has special points such as the intersection of character lines and the tip of a "spring" ( When viewed with the exception of !, its first derivative (the slope of the contour) has a continuous foundation characteristic.

Accordingly, with the conventional data compression method based on the contour method, it is difficult to obtain compressed data that is faithful to the character wheel Co1t.
However, there was a problem in that it was not possible to accurately remove unnaturalness (discontinuity t!I of 1p) from a character image reproduced based on the data.

In order to solve problem A1, the applicant has already applied for the method of Japanese Patent Application No. 57-16 B84 (Japanese Unexamined Patent Publication No. 58-134745).

However, since the data compression method disclosed here attempts to approximate the entire contour in one nail, in a contour part where a straight line part and a curved part are continuous, unevenness that deviates from the contour at the connection point. is unlikely to occur.

If you try to avoid these problems, you will have to set more Mukumoku, divide it into many polynomials, and make an approximation of 4 lines, which will increase the amount of compressed data. The problem became clear.

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

[Object of the Invention] It is an object of the invention to provide an improved contour method data 1f format;
Even though the calculation time was shortened when calculating the compressed data, the outline of the characters remained faithful. Another object of the present invention is to provide a method for processing character image data that can obtain 1 data, and it is another object of the present invention to provide a method for processing character image data that can obtain data with a sufficiently high degree of compression while faithfully storing the smoothness of character contours. It is another object of the present invention to provide a method for processing character image data that yields a
A method for processing character image data that faithfully reproduces character images with smooth contours and magnifications per second when decoding the data and displaying the character images. Is Rukoto.

[Ghost of the invention 1 Data processing related to the method of the present invention 1! Summary of Ij in Part 3
(shown in the figure) [-1-Cyto-1~]

The manually generated character image is decomposed into dots in the form of a [k, V] matrix (30), and the outline of the character image is extracted (31). The extracted contour is linearly approximated by <1 vector] so that the amount of deviation from the contour is less than the allowable error (32). The linearly approximated contour is classified into a straight part and a curved part based on the length of the vector (33),
On the other hand, dividing points of the contour are determined while determining the intersection angle of vectors tangent to V4 (34).

Then, for a straight section, the contour shape of the section is expressed by an n-degree polynomial (n=1), and for a curved section, the contour shape of the section is expressed by an n-degree polynomial (n=2.3>). Approximate a curve.

In order to approximate the curve, at each contour point forming the contour,
Find the slope at (35), find the coordinates of both ends of the section to be approximated by the curve, and its (n-th degree polynomial by σ1). Then, while extending the section, calculate the r-th degree polynomial and the contour to approximate the section. The longest section (sample section) within the range in which the deviation r is within the allowable error is determined, and both end points of this section are determined as sample points (36). In this way, once the ripple intervals are determined and the n-degree polynomial that approximates each ripple interval is determined, the first derivative of each n-degree polynomial is determined, and the slope of each contour point is recalculated ( 37). Then, sample points are determined in the same manner as described above based on the newly determined slope and coordinate values of each contour point, and an n-th degree polynomial that is more faithful to the contour is calculated (38). In order to encode the coefficients, degrees, etc. of the n-dimensional polynomial obtained in this way39), and to efficiently restore the character contour later, the starting point and Name block data are stored in order of distance between end points (40). The compressed data of the Iζ character contour that is stored is a plurality of 5? 41), and the desired character outline is restored based on the result of the deciphering.

1 Examples of the invention] Next, data processing in each part will be explained in detail.

[Image input (30)] Characters/images are separated into [X, Vl] helix shape by raster scanning of S-I+, Su-shouchiku, and this is J-shaped. The pattern data is processed! It is supplied as original character data.

[Contour extraction (31) i Determine the position (contour point) where the 2 candy data corresponding to the disassembled driver 1~ changes from ``0'' to ``1'' or from ``1'' to 101 in the X or X direction. The outline can be obtained by Then, the requested contour is divided into blocks of monovalent functions with x as a variable to form a set of blocks.

Figure 4 shows how the contour is extracted from the character data decomposed into [x, y] mamatrix-like dora], and the contour is divided into monovalent function blocks (interval from [○] to "○"). This is what I did.

[Line approximation (32)] Any one block on the contour [In one block, the amount of deviation from the contour can be set within a range that is within a predetermined tolerance. Make an approximation.

FIG. 5 shows an arbitrary block [P+, Pnl,
A character contour 50 indicated by a dotted line is linearly approximated by a set of two-dimensional vectors 51.

Connection points PI, P2, - of each flat 1 to 51
Pn-r.

Pn is a sample point stored in a sample point coordinate storage unit 60, which will be described later.
to be memorized.

[Identification of straight and curved parts of character outline (33)] As mentioned above, character outlines generally have straight parts and curved parts, and conventionally (-+, I, Approximate the entire contour at once and use
l'J 43; Set many sample points near the connection point of A so that one curved part is continuous, and perform VJ by dividing into many approximation formulas. ri, l
：'<K, the amount of data increased, and the compression ratio increased by 1 (1.J).

In this section, we will identify whether the contour shape is a straight part or a curved part based on the length of the vector determined by the surface line approximation described in 1. J, which solves the above drawbacks by processing each separately.

FIG. 6 is a block diagram showing an embodiment of 4I+ which realizes a method for identifying straight and curved portions of a character outline.

In the figure, 60 is 41 by the linear approximation (32).
1, the sample point is stored in block units.Sample Pisces 4, 1jji storage section, 61 is the identification vector l~l length setting section, 62 is a curve segment'
+'5111t Set waste separation angle 1αθ J-ru curve dividing point identification angle wave setting section, 63 is 1st sample point P LSI
The next sample point coordinate register for holding the coordinates (x1.+I, yt+1>), 64 is the first sample point PL
The coordinates (Xt, .

yL), the rounded current Muku tree point coordinate register, 65 is i-I Tr [sample point j of 1] L-1 locus 4Gi (
x t, -+.

The main point coordinate register 66 holds the sample point coordinate register 63 and the current sample point coordinate register 6/l.

A vector length calculation unit 67 calculates the vector length of the sample point PL from the coordinates of the next sample point coordinate register 63, the development freezing point coordinate register 64, and the Maegura main point coordinate register 65. Angle 0L (aZ
A final overlapping vector angle calculation section 68 indicates the identification vector length 1-d II set by the straight line section identification vector length setting section 61.
IJ record vector] ~ Vector length J calculated by locking out the long camphor
A solid 1 to 2L comparison section 69 is set by the curve division point identification angle setting section 62 and calculated by the 1st curve division point identification angle θ and the solid 1 to 2L angle hemp output section 67. An angle comparison unit that compares the angle θL between the vectors, 70.71
is a straight line part storage part 9 curved part storage part which stores the section of the straight part or the curved part, respectively, based on the comparison result of the vegetal length comparison part 68;
-2・θL 1st sample point 1〕〕〕〕〕〕〕〕〕〕〕〕〕〕〕

Next, let's explain the operation. First, Jj to the straight part identification l\ri1-ru long button part 61 and identification mark 1 hell & L,
llIth ≦ (minute ^) 1 fish identification angle setting f'fll 62
The curve dividing point identification angle 1 and θ are respectively set in . Next, from the sample point coordinate storage unit 60, 41-1 block [1 block!
/11 Point coordinates are sent to the next marker tree 1 - coordinate register 63. At this point, it is not possible to set a solid value in the vector 1゛・le length calculation unit 66, which will be described later. Not 31, then the next sample point) 1q! >, The sample point stored in the first sample point 63 (now stored in the current sample point coordinate register 64, and the next sample point position is newly stored in the next sample point coordinate register 63.) J, , ]J, j sample point P
Specimen Pisces in the Imo register 64 is stored in the previous sample coordinates register 65, next sample point Li+sample point J in the register 63!
! Set the target to the current point coordinate register 6/If this each shift 1
.

Solid 1 to sieve extractor 81 (66 is the 1+1st sample point P of the above-mentioned next point coordinate register 63, LSI coordinates (country x, country y) and development freezing point coordinate register 6401st sample point l) of t. A vector is calculated from the coordinates (xL, yc >. The calculated solid line length β is determined in advance by the straight line portion identification solid line length setting unit 61 in the back line length comparison unit 6ε3. Compare the set identification number 1.Here,
! sheet>, the section l:x (,, country
(Stored in the storage unit 70. Also, the interval β(≦1-throat, CXL, XI,I+] is blank)
-2, the section is stored, and the next section is identified. Here, if the next section is a straight section, the above-mentioned 1 is omitted. ,
I remember it. 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 i(
When the section changes to a straight section, the continuous curved section until then becomes one continuous curved section, and its 917
The point coordinates and end point coordinates are stored in the curve section storage section 71.

[Divide the contour (34) 1 - h1 vector 1 helta square mill Q output part 67, next block Mt. Pl
n Mukushi Jista 63. From the current sample point coordinate register 64° and the previous sample point coordinate register 65, sample points Pc-+ and 1, respectively.
Two L, one (+1 (1) seat 4M (Xt-+
, Vt-+), (XL.

Vt), (Country

(> The angle OL between Muku 1 and Hel that is output is compared with the curve dividing point identification angle θ set in advance by the curve dividing point identification A78 setting unit 62 in the angle comparing section 69. Here, when 0) θ. , set the sample point PL to a new curve dividing point and (), 111
Eye≦1 The coordinates are stored in the division point coordinate storage section 72.

7i, there are many sample points near this curve dividing point.1
However, by dividing the contour at the curve dividing points obtained above, the curves are approximated separately before and after the curve dividing points, which reduces the processing time. Approximate curves can be easily drawn.

FIG. 7 shows the character contour data shown in FIG.
Above mentioned “Line approximation (32) 3 to [Contour division (34)
'] shows an example of data processing.

In the figure, O is the 917 points and the end point of one block, Δ is a sample point obtained by linear approximation, and fist is a curve dividing point. Also, * indicates a straight part, and 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 the r11th order term (where n
-1, 2, 3>, the n-th degree polynomial is uniquely determined once the coordinate values and inclinations of the two points are determined. Therefore, in this case, first, a3 is applied to each contour point on the contour.
There is a need to reduce the inclination to

In the present invention, the slope given to each contour point is determined by extracting a predetermined number of contour points before and after the contour point, and then connecting the contour point whose slope is to be calculated with each extracted contour point by a line segment Lυ. (1f1
Go ahead and move J> 4j to the desired contour point.
Calculate the value J, sea urchin.

A method of controlling the inclination of each wheel I; II+ in the present invention will be explained with reference to FIG.

Any one section of the contour as shown in Figure 3(a).
Tsuk 1. 'P+, Pn] starting point 1) 1 <contour point C)+
) in a month (to set the starting point 1), extract the contour points that exist r1 after 1 (extract the number of breaths, and set the starting point 1)
1 and the famous contour point, and calculate the slope tl of each class J, start J: (IP + ) from the formula described later. .

For example, λ-11 (extract points 25j, Insert 1 Sago T1□.

wa;
+, Vt>.

< x2 , y2 ), it can be found. Note that the same procedure can be applied to the case of rTl'a.

When the slope of the line segment m, , rTl 2 is rounded, the contour point Q1, .

Next, when calculating the slope L2 that falls to the next contour point Q2 as shown in Fig. 8 (i), extract a predetermined number of contour points that exist before and after the contour point Q2, and calculate each slope from the contour point Q2. Determine the slope of each line segment connecting the contour points, and set a to the contour point Q2.

For example, when extracting two contour points in the front and rear to calculate the C slope, the ring 91 (if J3 is located at the point Q2, only the front one point can be extracted, so in this case, by extracting the front and rear points, the slope t2 can be calculated using the following formula. Jζrimel.

If this J: does not have the specified number of contour points,
In order to extract the maximum number of contour points within the specified range and tilt J8°, then, as shown in Fig. 8 (C), if you set the slope t at contour point Q3, At the contour point Q3, there are a predetermined number of contour points 1° before and after it. Therefore, the ring ν1 is 03
and the slope ml of each line segment connecting each contour rib.

Find rn2, rTIJ, m, respectively, and find the contour point Q3.
Then, calculate the slope at each contour point in the same way while sequentially looking at the contour points. To go.

“Determination of Rimple Points (36)] When a rectangle of JjL is drawn at each contour point as shown in below-1-,
Next, an approximate curve determined by the contour points of the contour 1-02 fish is determined, and the amount of deviation between the approximate curve and the contour is sequentially determined for each wheel Yu1 (point).

Therefore, we now have the problem and the approximation curve section i15 It
) is less than the allowable error, the contour point is moved forward one more time and the same process is repeated again, and the deviation h1
The contour is divided into υ-numble sections while determining the longest section (hereinafter referred to as υ-numble section) within a 621-dimensional error range. Then, the ring 91ζ is approximated by the r11th order term determined by this ripple interval.

Hereinafter, the calculation of the approximate curve of the n-th degree polynomial and the method of determining 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 coordinates 1111 stored in the straight section storage section 70 and the curve section storage section 71. Therefore, if the section is a straight section, the beginning and end of the section and the section are excluded from the linear equation (linear equation), and encoding, which will be described later, is performed.

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

Figure 9 (a>, a3-c, 5L ((1) l> is the starting point of the first rimple section of interest, and the 913
Take point 5L (Ql) and select candidate point St ÷ (0 to 1)
-C
go. B is the contour point QJ, C is perpendicular to the straight line 92 connecting the sections covered by the curve approximation;
iJ3 G) The amount of deviation in the x7j direction between the contour 90 and the approximate curve 91, ■ is the amount of deviation between the contour 90 and the approximate curve 91, BD is the amount of deviation between the contour 90 and) ) is the amount of deviation.

41, section IQ+, Q2]
ff 24.1 It is suddenly decided that it is the 1st white line, so the next section IQ
+ , Q:] ] to find the approximate curve of the n-th order ¥/jj'4 equation.

The following will explain the case of a cubic curve as an example. Already t>out()
Iw e t,L,1. of the starting point SL (Ql) and one non-pull candidate point Q++. n to 11 (vote
A1 white (×(,,V+,),(Xn,Vn) as below 3
By substituting into the following equation, ) Li quasi-concave curve 1 of the section
X) is round.

r (x) -yc +b (x-xt) ICL (
X - ``Amount of deviation ε between (X) and the contour
I put it on.

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

In the figure, the amount of deviation ε(=B
When considering C), the present invention provides 1. The i-applied amount ε is artificially excluded and covered to enable high-speed processing.

FIG. 9 ([)] is an enlarged view of the error portions of the contour 90 and the approximate curve 9'1 in (a>).

At J3 in the figure, it is assumed that the line segment plane is parallel to the straight line 92, and the slope r of the line segment CA is the slope of the straight line 92. Ishi-
At C1 wheel 91 (point B((,)J), approximate curve 91
What is the distance ε for which x direction and y direction?

By setting ε〃, The desired deviation b1ε is It can be approximated by .

If the IL-like deviation tail ε of the contour 00 and the approximate curve 91 is within the allowable error range, the next contour point 0 is set as the Linsol candidate point in the interval , and in the same manner, create new sections LQ+ and Qnnl (
Including the approximate curve, all (contours +:, 1
(, l J, examine the deviation mε.

Then, if any ring *1 (point exceeds 6'F tolerance error), the previous rimple candidate point Q n is changed to 1 numble 1j, L S 1.+1, etc. i?f'rth rimple section "SL.

8 countries and 1 will be decided.

Thereafter, the Rinsol point S country is set in the same way as point 9r'i of the next rimple section, and Moe 611, and the rimple section is sequentially requested from (J) without determining the rimple point.

In addition, the contour is divided into a plurality of electronic number sections, and J: ring νIX is approximated by approximating each sample section with an approximate curve.

In addition, Juki explanation C-iJ, the next simple candidate section [('
) + , Q nul in Ni A3 kel (la (Q
The rimple section is determined as the previous rimple candidate point Q n @ 'J'.

However, as another embodiment, when the allowable error is exceeded, Q nll is formally stored as a sample candidate point (Q nll) at that point (Q n + z , Q n
For the sample candidate section up to s, .

Even if the IJ reaches a predetermined number of points ahead, the J exceeds the allowable error.
:The non-pull section may be determined by using the above Qn as a sample point.

Of course, in this case, when a3 is in the υ sample candidate interval several points ahead and the tolerance regulation is satisfied, the candidate interval is updated to a new interval [(:) + , Qn ], and one non-mple candidate interval is added to the history from there. Repeat the above evaluation process while moving the points forward. By performing such pre-reading, the sample interval can be further reduced, and the data compression rate can be improved.

[Recalculating the slope of the 91ζ point of each ring (37)] As described in 2, when the contour is a3 in multiple sample sections and each is approximated by a cubic approximate curve, the starting point coordinates of the lean pull section and the sample By calculating the degree and coefficient of the approximating curve that approximates the section by i+l-j m according to the contour data, it is possible to obtain the contracted 2]-de. !71; Only for Sada's inclination (:l!+),
1 The QIX points for each wheel that you requested
tf because of the ring l71 In order to approximate it faithfully, we take the -th derivative 1-(x) of the approximation curve 1(x) in each ring layer [interval, I'm going to re-introduce the inclination of
It was to so.

Shirimple fish -ttj decision <38)] Recalculation of intelligence of 1 person outline (37)] J, Ru data processing l! 1) For each contour point requested by Then, determine the sample section while determining the re-rimple point. 1'(X)-("Approximate.

[L-do conversion (39)] Below-1 J, Jj is added to the straight and curved parts that were claimed in the above way.
Starting point coordinates of each non-pull section, approximate curve 1'
By converting the coefficients and orders of (x) into =1-code and storing them as a set of block data organized in units of blocks, compressed data that is faithful to the outline of an arbitrary character can be obtained.

FIG. 10 is a diagram showing an example of the most preferred data storage format for any one block data, which is applied in implementing the present invention as described below.

In the format shown in the figure, the block header ~ stores the ending 7' coordinate of one block and the number of sample sections existing in one block, and the segment header stores the X, Y coordinates of the starting point of one sample section and approximate The order of the curve 1(x) is stored, and the coefficients of the approximate curve f(x) based on the information 1'J and the above order are stored.

Then, one (non-pull section data) is organized by segment header and segment information, and
Rimple section data corresponding to the number of Rimple sections are arranged in sequence and constitute one block data as a whole.

[Storage of compressed data (,40)] Each blower/7 coded from 1.A-7 1.
j-yu is effective when restoring the outline of the 16 characters described below. Let's do it! Then, blocks are stored in each f1-] block in descending order of decoding time. .

Example
1 tsuku〈explanation〉-1 for each block [] tsuku 1-. 20 numbers a, l 1. / /, : ) If: 1Ji to flj
The case where i data is collected is shown below.

For example, if the code is difficult and takes a long time to decipher,
'II'! 11 ivy glue for each l? The distance between the point and the end point is memorized in order, starting with the distance between the two points. 1st
In Figure 1(a), the block that takes the longest time to decode one block is block 12, followed by Block 1G, followed by Block 1G. Tsuku 3/
J (I know the most. Therefore, the final character ring νIX) 1reii J'□-Yu is shown in Figure 11 (b)
As shown in 5/,;-1F], the data is stored as a set of data1, :/,j.

Distributed data = + - code (41)] What I have just described is related to data compression of character contours. Next, a method for reproducing the original contour based on the contour data compressed using the IC data processing method described above will be explained.

In this case, in order to restore the contour at high speed, the block data for each block is sequentially transferred to multiple decoders for processing. The block will process one block of data. (Previously) The time required to decipher one small block is a problem. In other words, the longer the time required to decode one block, the longer the processing time required to reconstruct the contour. Therefore, if the data with the longest processing time comes later, the data will continue to be processed, and all other data will finish processing, resulting in 10 days. , the efficiency of use of the device decreases.

Therefore, as mentioned above, the present invention has developed a method for converting 11 compressed data into one block.
Memorize the key parts that take the longest time to decipher one tsuku,
By performing processing with multiple decoders, the decoders can be used efficiently and contours can be restored at high speed.

Below 1., Explanation 14 with reference to Figures '12 and 13,
.

FIG. 12 (31) is a block diagram showing an embodiment for optimally performing contour restoration according to the present invention.

In the figure, 1 block data is stored in 1 block (120 is the solution of 1 block), and 14 blocks are compressed in order of order. Selector 1 selects and transfers 1 block data of
The magnification storage unit 123 stores the previous hi block data 4J'
Restored to ◆Q-/B corresponding to l, rate n l hard day 1-
Data consisting of data: 21-data group, 12/l is the data]-
From the data group 123, select group 4 for the end of the process. Seat I contour side rope
Selector 2 which transfers the data to the 1-character storage unit described below.
125 kt: the selector 2 (124) J: 1 character storage unit that stores the contour data of the selected character; Printing characters 1L/; 2 is an output device for displaying, etc.

Next, the operation will be explained.

Pressure 1? For example, in FIG.
Tsuku 12. Block 10. Block 1 (in order of 3...)
is transferred to selector 1 (121). Selector 1 (1
21) transfers the transferred block data by sequentially selecting decoders from the decoders 1 of the data group 122. The decoder that has completed the block data ψcrowding process the contour restoration process Il based on the block data and the desired (g<ratio data) separately stored in the magnification storage unit 122.
Start l. The selector 2 (121) sequentially selects the decoders that have completed the restoration process, restores them, and transfers the restored lζ contour pixel data to the one-character storage unit 12F). The decoder that has completed the restoration process for 1 block [1 block] selector 1 (12
1) is requested to transfer block data, and one block of data is transferred to the next one block of data.

FIG. 133 shows each decoder +! shown in FIG. Y 12
An example of the processing status f of No. 3 is shown below.

In the figure, T+'t selector 1 (121) to 81
'-1-k/11 time to transfer data, 1. f
J, each character storage section 1251\Φ1 stores the contour search data read by M through 1 node 2 (12/l).
．． There is C between 1+iy7 to send, and 1j is J' to each decoder.
Nful/'IItsukj'-Evening solution, processing time <)
. Maruko, 1, accepts processing time for one character.

F 1-g 1'~・r) to r) b[-1 ts.
The 711-th 711-th check is transferred and processing starts!j(i-!If), then the n In this case, de=r-
The data is transferred to 3) and continues processing.

However, each block sent by
As shown in Figure 13, the processing time for each character is arranged in the order that the processing time gradually becomes C, ii <. The process ends.

Moko: -) "C each y=+-g operates almost evenly,
Waste/, I: Play (/ 11. '1- to mark between 41
<Complete reading of character data as soon as possible.

[Contour restoration (42) 1. As mentioned above, the 170 contour landscape data restored by each decoder are sequentially stored in the character storage section 12E5, thereby completing the contour search data for one character. Then, the contour search data for one character stored in the one-character storage section 12E5 is supplied to the output device 12G, such as a laser beam printer, CRT phototypesetting machine, display IN, etc. , 1'iL as in the prior art (the desired character is restored).

[Effect of the invention] Error-1-Enjoyment-(The place of character image data that becomes clear in the book 51 J! Ijp) method is 800 x 800
-/111ra Sei 6 Mincho 14VX11 Kana sentence'-j"l
- As a result of verifying the application to A1, it was applied to the desired character image. If the wear error is 1 ton and 1 is 1, then the data is 1.21% and 8I? I decided to increase the i rate to 11J.

Also, in the above explanation, ・ is explained as M2 using “a” as an example.
In addition to Taka and Ire, various marks other than b kanji and letters + 8
No. 1, Ming also said that images such as line drawings can be handled in the same way as bIIIJ.

As described above, according to the present invention, character contour d' can be compressed with a sufficiently high compression rate.
The data that is memorized is stored at high speed (Mi no), and when it is restored to history, it is necessary to
It is possible to provide a character image that can be reproduced during the day in (i+""-c").

[Brief explanation of drawings]

Figures 11 and 2 are diagrams explaining conventional contour method data compression, and Figure 3 is a diagram showing data compression related to the fJ method of the present invention. -1~, Figure 4 is a diagram showing an example of dividing the contour into blocks of a single-valued function that takes X as a variable, Figure 55 is a diagram explaining linear approximation, Figure -'16 (
1. Figure 7 shows an example of the discrimination method for white line parts and curved parts of the present invention. 9 is a diagram illustrating the method of calculating the DI of the slope at each contour point according to the present invention, FIG.
Figure 0 shows an example of the block data storage A-1~, Figure 11 is a diagram illustrating the method of storing data according to the present invention + FIG. 13 is a block diagram of an embodiment of the configuration in which block data is restored by the following, and there is a timing block t=-l-C showing the operation of FIG. 1, bO, 90...Character outline 2.51...2-dimensional vector 3...m-th curve element 60...Sample point coordinate storage unit 61...Row 1 line section identification vector 1~L Length setting section 62...
Curve dividing point identification angle 1 setting unit 63...Next sample point P1-marker register 7'l...Current sample point coordinate register 65...Previous 4 votes book Ia I soil register 6G...Bec) Line drawing part 67...Flat 1 to closing angle extraction part 68...Flat 1 hell shape comparison part 69...Square polishing comparison part 70...Straight line part storage part 71...Curved part Storage unit 72...Curve division point coordinate storage unit<)1...r1-order formula J''C approximate curve 120...
1 block data storage section 121...Rector 1 122...I15 seat''8L!4ml11233・
・・De]-da l+Y 124・−・Rector 2 12 h・・1 character t, i memory section 126・Output! 8. Neck Figure 3 Figure 9 (0) (b) Figure hO (a) (b) rill

Claims (1)

[Claims] (+1.
In the method for processing IJ character image data in which the original character image is reproduced by decoding the 1-q'l When the section is completed and 7j is reached, J-34 is placed at both end points of the section.
Find the coordinates (x+.
'I +1 and locus (vote b' 6) Find the n-degree polynomial MX that approximates the predetermined interval of R+J (where n-1, 2, 3), and calculate the coefficient and degree of the 111th-degree term for each of the calculated l. Processing power d1゜(2) of character image data stored as contour data specifying the predetermined section; Character image data processing method according to claim 1. (3) The above 1(x> is expressed as 1' (X) -V + + b 1(X -X +
) 10c, (x-x, )2 10d+ (x Xi )3, the coefficient is calculated from [), = 1゜n-xl. request