DE2503185A1 - METHOD AND ARRANGEMENT FOR MINIMIZING MEMORY SPACE OF COMPUTER-CODED DATA DETERMINED FOR HALFTONE IMAGE PRINTING - Google Patents

Description

r .. (3b ±? .nren ₃ de α I ¹ J. January 197: 5

Aiiiiielderin: IBM Germany GmbH

7000 Jtatt part 30 FasealstraLe 100

Official file number: iieuaniaeldun ?. File number of the applicant: GIC 9 ~ 75 ~ Oü3

Method and arrangement for storage space rdnii.iieran ^ · of comp ut ere ο dl er th for the halftone image print b es t imten pat en

The ^ rfindun; relates to methods for minimizing storage space: of coiiijjuter coded data intended for calf tone image printing.

The digital lay-tone printing technique (in J ^ ri-£ .lisch: halftone printing) makes use of the fact that, given a sufficient distance, the human eye can see small, digitized pattern elements does not recognize as such, but rather several of these model elements integrating as a continuous female tone .

Usually an original halftone image is screened in this sense j that original halftone image area assigned to each raster point the associated roughness value is determined. This, initially still analog gray value, is assigned to an interval of the gray value scale divided into ρ intervals. The bigger the Gray value, the larger the diameter can be, one at the point of intersection black dot appearing in print. This known method (DA3 1959 459) is, however, in terms of print quality limited.

To achieve a finer print quality and the associated coi: .puter-oriented storage space problem for the image data to be printed, it is the object of the invention to provide a

609831/0534

drive to the ■ storage of computer-coded files for
to specify certain data in the color tone image printing.

According to the invention, this task is carried out in an advantageous manner thereby
solved that an analog 3ri ₀ irial halftone image of an equidistant
Casting is subjected to the fact that for each original halftone image area assigned to a raster point (picture element) the; analogue roughness value is determined and an interval of in ρ intervals
(including the pure black value) divided gray value scale
is assigned that each point from (· -) matrix formip; arranged bright and / or dark pixel elements is formed, that determines the number (the length value) of similar (hotter or darker) adjacent picture elements for the totality of all picture elements of all picture elements in rows, ^{1 or columns.} it becomes that all possible length values of a probability '

grounding the distribution of subordinates, and that according to this; Probability distribution of frequently occurring length vertices

a shorter, coded to be stored indication associated v / ith the i

rarely occurring length values. I.

According to the invention, the fanning takes place _; : · The pixels in a! individual picture elements, with the same picture quality , in an advantageous manner
Way according to the ochema. \

15 4 2 G877 i

3 7 35 6555;

or
6 8 7 3 4 4 3 3

2451 2211

where the numbers given mean that the gray levels I2-I3
₃ are formed by the fact that for the stage IÖ the; .iit δ indicates-; are black, for level 17 those with
7 elements are black, for level 16 the elements indicated with 5 are also white, etc.

_3E 9-75-003 60983 1/0534

In the first case of the first scheme, according to the invention, a storage space optimization can advantageously be implemented; to the effect that a Kuffman code optimization uni- 'for all possible length values that occur. is applied. This results in an aoiapression factor for the image print data to be stored of 1.4; in the case of the second scheme, a Koniprimi he ungs factor of 5 ₃ 45 i ^m compared to a stored code without Optimierunp ;.

In the case of pure text representations, it is advantageous according to the invention to if all possible length values of an optimization by a 'be subjected to logarithmic code.

For code optimization; of VJetter maps or line drawings, which is not pronounced in its distribution of the individual length values Maxima as in the case of pure! "Ext information leave the country, is according to the invention for a code optimization; a linear code 'advantageous.

The invention also relates to an arrangement for generating | of dotted dot element oysters from digital pixel gray values j for carrying out the method according to the invention.

i -

According to the invention, this arrangement is advantageously thereby characterized in that a rotating shift register for Recording of the digital gray values is provided for N line pixels of an original halftone image, and that from a gray value the associated pixel element pattern can be generated by accessing a memory containing all the pixel element patterns. ! and that the generation of the pixel element pattern for the! II cable pixels must he line-by-line by ρ-fold rotation of the Slider register content is feasible, and that the generated ! Bilpunktelementmuster computer program controlled by a Codeopti ™ can be subject to

A particularly favorable method for simplifying the through the second scheme given is more advantageous

"Else according to the invention characterized in that a circulating shift register of a Originalhalbtonbildes is for receiving the digital gray scale values for ii line image points, and in that from a gray value of the corresponding pixel elements pattern is generated by accessing a all Bildpunlctelementenmuster containing memory, and that the generation of the Lildpunktelementenmuster for ii line pixels can be carried out in a pattern line by ρ-fold circulation of the slide register contents ₅ and that the generated pixel element can be subjected to code optimization under computer program control 3 ind.

An arrangement for carrying out this simplified method is advantageously characterized according to the invention; that a circumferential shift register comprising 3-bit-wide II stages is provided for the successive recording or digital gray values II-I6 for .J line pixels of an original blue-tone image, and that the associated pixel element pair pattern can be derived from a gray value by providing an adder δ is üurch to the gray value of a Bildyunktes for Bildpunktelexüentenpaar a dor Fourth "0, for the LiIdpunkteleuentenpaar L of I \ ^r ert I ₃ for the öildp oint element enp aar C uer i'ert 2, for the Bi ldpunkt element enp aar D the value 3, for the pixel element pair E the value 4, for the pixel element pair F the fourth 5 ₃ for the pixel element pair G the value G and for the pixel element pair H the value 7 can be added and that the addition result S. .Comparison circuit can be examined to determine whether it is smaller than the number of grayscale levels I. to I _n plus 1 (ie 8 ■ + 1 = 9), with the signal for an affirmative comparison The result is a black pair of pixel elements, which can be marked by a negative pair of pixel elements, and that the generation of the pixel element pair pattern for the pixels can be carried out in pattern lines by circulating the shift register four times, xv this can be acted upon by the adder before each cycle, and that the generated image

ge 9-75-003 6Q9831 / Ö534

point element pairs computer program controlled by a code opti g are subject to.

The invention is explained with reference to the drawings and tables

demonstrate:

^! Figure 1

Figure 2

a pseudo representation of a rriatrix-shaped Fanning out individual pixels into black and white pixel elements for the individual ones Areas of a gray scale according to a first scheme,

a schematic representation of a luatrix-shaped Fanning out individual pixels into black and white pixel elements for the individual ones Areas of a gray scale according to a two-fold scheme.

table 1

a schedule for a Huffman code optimizationj

a iacramm to a probability distribution P (RL) = f (RL) for the "length values" (RL) of Text information,

Table II

a list for a logarithmic code,

Table III

a diagram for a probability distribution P (RL) = f (RL) for the "length values" of Line drawings,

a statement for a linear code.

GE 9-75-UO3

a schematic representation of a simplification of the matrix-like fanning out of the individual

609831/0534

Image points according to FIG. 2 in black and white image point element pairs.

Ii ¹ IgUr ό a schematic circuit diagram of an arrangement for

Identifying the black and white. Pairs of pixel elements according to FIG. 5 from digital gray; \ rerten.

To explain the process according to the invention and the invention ^ According to the arrangement, the prior art will first be explained, 'on which the invention is based.

Previously, digital printing of halftone images required starting : the following process steps from an analog original halftone image

1. The analog Orinal halftone image becomes an equidistant one Subject to rasterization. Each raster point (image point) is a specific area of the original halftone image a assigned.

2. For each pixel, the analog gray value of the associated Orignal halftone image area determined and an interval of the original value scale divided into ρ intervals assigned.

In the illustrated example, it is assumed that the drauwertskala insgesar.it ρ = δ intervals I1-I3 (black, gray levels 7) contains.

While so far the various gray expressions have been replaced by an in its diameter is variable black point in the image grid were reproduced, is now after the invention Process such a pixel into individual pixel elements' fanned out. "

According to FIG. 1, this fanning out takes place in the way that one an image point of bright and /

609831/053 4 GiS 9-75-003

or dark pixel elements. The ilatrix has an extension of (vy) rapid elements. In Fi.p ·. 1 "are shown for the individual gray levels - starting from a dui * chwe ^ black pixel - the corresponding Λ on acnerunr: s of the pixels in individual image elements. The ochwar-sstüfe and the gray levels are designated with the numbers II to 13. The graduation i of the gray levels 12 to I'J ci ^o fol _f "; t in a sequence of decreasing arautones. (For reasons of Χοχ. A black egg; ipunktelenient is not flat black, sop-.lern only rasterized black ι j reproduced in the drawings). '·

A lighter shade of gray is achieved by adding more white images

j point elements reached in the image point. The order; the \ ischwarzen and white pixel elements for each 3raustufen ßeht from the following scheme riervor:

j Ί ο G ο J 7 3

2 Ί 5 1 -

Scheiaa contains as many numbers in lines as in

^ like the pixel niatrixförmi,.; contains arranged image elements, usually 4x4 image elements. ochvrarze t-picture dot lines appear for the first time in the first gray level 13 on the in; ocheiua positions of the pixel element matrix identified by the numbers 0. (Pay attention to the direction of the gray scale from right to left, i.e. IO, 17, Ij ... II) »

! • lit other words: The number 8 in the scheme means that for the ; Gray value level 18 in FIG. 1 the corresponding ilatrix pixel element is black. For the following gray value level 17 , are in addition to the black pixel elements of the IS level , also the pixel elements indicated by 7 in the scheme 'black. For the gray value level 16, the pixel., ! elements, which are marked with β in the scheme, black; Etc.

ge "9-75-003 gust 9 B3 1/0 5 3%

For storing the print data for any one of many Image composed of pixels, in which, in turn, the pixels are composed of individual pixel elements arranged in the form of a matrix exist, it is necessary to obtain the pixel element print data to save black or white according to their statement.

ι In this case, storage according to the known

j th so-called '' run-length ^ -procedures. JJanaeh would be the number j (length value) of all adjacent (or superimposed) similar pixel elements - also across the border of a ! Pixel away - be saved. The length of the image point elements II, 13 and 15 next to each other, so that would be line by line Storage for similar adjacent pixel elements take place as follows:

k, 1-, 2, 2, 1, 2

8, 1, 3

11, l _y i

4, 1, 2 _S 3, I ₃ 1!

The third row of numbers 11, 1 is explained as an example: In the third row of adjacent pixels II, 13, 15, there are 11 black pixel elements next to each other (hence the number 11), followed by a white pixel element (hence the number 1 after j 11). The numbers mentioned would be stored in coded form in a computer. However, such storage jwäre still too bulky, and are for that reason! Each so-called "runs", ie, the number at the same ^1-like, adjacent picture elements, hereinafter referred Länigenwerte subjected to a probability distribution! Be. Such code optimizations are known per se. A code optimization according to Huffman is chosen as an example of a coding of the possible length values.

It is assumed that the individual length values are in the range from 1 to 12 8 with the appropriate image size. For such a length distribution is now shown in Table I.

GE 9-75-003 6Ü383T / D534

Probability distribution according to Huffmarin before ^ e _t ; even.

The probability with which a certain length value RL occurs is indicated in Table I with ρ (RL).

(This designation ρ (RL) is also used for the Si ^ o and Fi,:;. ¹ ^ designation ; the probability).

609831/0534

TABLE I.

Lan jenwert 3 (RL) O P (RL) Binary ( 1 O , 10 Oil 2 ο , 18 001 , 40 1

0.05 0.0ό 0.10 0.0ό 0.03

12 6

0.02

000010

0101

0000

0100

00011

1000111001

GE 9-75-003

609831/05

According to the Liuffman distribution, Bi 1 de learn ten lengths a long code word (not so nüufig) assigned with low probabilities, lengths with greater probability a shorter code word. If, as can be seen from Table I, the length value 4 has a probability of 0.05., Then becomes ^ this length value is binary coded as 00010, the length value 3 with On the other hand, a probability of 0.40 receives the binary! coding 1

This Huffman code optimization, when applied to an

fanning of the pixels according to Fig. 1 a Kornprimi erung factor of 1.4 compared to a length value storage without code optimization.

Another improvement in the compression factor of 1.4 t 5.45 with the same image quality is achieved according to the same procedural principle, but with the use of a fanned-out image point according to FIG. 2. This fanning out of the image point takes place in analogy to FIG ider mentioned to Fig. 1, according to the scheme

8th ο
υ 7th 7th 6th β j 5 4th 4th 3 3

2 2 11

This fanning out of the pixels has the advantage that for adjacent Pixels of different gray levels on average i fewer, but longer "runs" occur. For 3 side by side- '.

lying pixels of gray values II, 13 and 15 would therefore be without;

! Code optimization to save only the following length values: j

: 12

i 8, 4

! . 4, 8

6 0 9 8 3 1/0 5 3 4. ._

GE 9-75-003 ■

, 'j ^ ine modification of the method previously described then seheint ι I attached when it is print information * the pure from j ¹ Text (without Halbtoraltollälei *) Ibestehvfc .. lexten In Feimeni check results by the iluchstafoenabstand, Cuir-da the otrlehdleke typical for a certain £ j, Jchrlftart, due to the cell level, etc., ₃ an e-functional distribution of the ⁵ Iiiingemiwerte (RL) with typical kaxliua. For such a distribution; ρ (PUL.) = f '(KL) as shown as an example in Flg. 3 wlederce-ysben Is, let be ¹ a code optimization ί mi hilfc a so-called en lo j code dorclifuhren.

■ As an example, such a logarithm is more likely · Code is shown in Table II ¹ .

Iu binary code of each Jlnärstufe a so-called "color bit" is assigned to enable \ xui a unique decoding to black and white pixel elements.

These color bits are not listed in the binary code of Table II.

609 83 1/05 34

GE 9-75-003

T A ΰ λ LLK II

Length value (RL) Binary 0 code 1 0 1 2 1 0 3 0 1 4th 0 0 5 1 0 6th 1 1 7th 0 .1 ϋ ο 0 ■ J 1 9 0 ü 0 10 0 1 1 11 1 1 0 12th 1 ΰ 1 13th 1- 0 0 14th 1 1 13th 0 0 0 16 0 0 1

GE 9-75-003

609831/0 534

- lh -

In the first column the length values (RL) for similar, side by side (or one below the other) pixel elements are listed j in the second column the associated binary code for these length values. Four similar adjacent pixel elements are encoded by the binary code 0I ₃ ten similar adjacent pixel elements by the binary code 011. Such logarithmic codes are particularly used for a Cοde optimization ₃ when 'a distribution is present corresponding to a falling exponential function with each i-iaxima.

other code optimization is available for weather maps, line drawings or the like. In Fig, H the linear probability distribution curve ρ (RL) = f (RL) of the length values (RL) is shown for a typical weather map. The diagram shows a falling e-puncture (without maxima as in FIG. 3). A so-called linear code has proven to be particularly favorable for a code optimization j for such a distribution curve. In Table III the example of a linear code is recorded.

6 0 9 8 3 1/0534
9-75-003

TABLE III

Length value (RL)

1 2

i 7

9:10

11 '12

ί 13

14th

! 15th

! 16

GE 9-75-003 binary code

0 0 1 0 10 Oil 10 0 10 1 110 111 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 10 0 0 0 0 0 0 000000001 00000 0

609831/0534

j In the first column are the length values (RL) for similar ! next to one another (or one below the other) pixel elements

, notes in the second column the associated length values for these

Codes. _{Accordingly, the binary code 10O 5} is used for a length of four similar adjacent pixels, for a length of 10, the binary code 000011 is used. Such a code is particularly suitable for optimizing probability distributions with an exponential descending course.

Representations that contain both textual information and halftone image representations should only be processed with one code for the sake of simplicity. For pure texts there is a logajrithmic Code that could be modified depending on the font etc. 'optimal. However, such a logarithmic code is for halftone images with a fanned-out image point according to Pig. I worse, but not for halftone images with a fanned-out image point According to FIG. 2, the latter would be carried out particularly in connection with additional text information which is also to be edited a logarithmic code optimization experience an optimal solution.

In Fig. 5 is a schematic representation of a simplified one The form of the fanning out of a pixel into individual pixel elements according to FIG. 2 is shown.

The type of image point fanning according to FIG. 2 allows one Simplifying, more comprehensive representation, as it is line by line only multiples of two similarly neighboring pixels occur. For this reason, in the illustration according to FIG. 5 always combines two of the neighboring pixel elements in FIG. 2: The first and second pixel elements in the The first line of the pixel according to FIG. 2 becomes the pair of pixel elements A; the third and fourth pixel elements in the first line of the pixel according to FIG. 2 become the pixel element pair B summarized and so on. Accordingly, the picture element according to FIG. 5 consists of 4 lines with the picture element pairs

GE 9-75-003 609831 / 053A

A and B (first line), C and D (second line), E and P (third line) and G and II (fourth line) together.

An arrangement is shown in FIG. 6 which makes it possible to generate the corresponding pixel element pair patterns according to FIG. In dual encryption, the parallel transmission of a total of 8 possible digital gray values requires 3 lines 1, 2 and 3. An original halftone image is scanned line by line, the associated analog gray value is determined for each of the N line pixels, which is then digitized on lines 1, 2 and 3 is fed to a parallel shift register 5. Each stage of the shift register is 3 bits wide, since the 'representation of a gray value II up to 18 requires 3 bits. The gray values for successive line image points are fed one after the other to the itf shift register stages. The gray value 12 is dual encoded with 010; it is stored in the lowest position of the shift register 5. The gray values that are stored in: the! Individual levels 1 to 1-, in the shift register 5, • are indicated to the left of these levels. To the right of the individual ^: shift register stages is the ordinal number (1 - ΪΪ) of the individual line pixels for which a scan was carried out. The entry of the gray values in the individual shift register levels is arbitrary and should only be understood as an example. Then the fourth pixel has a gray value of level 15 and the second pixel has a gray value of level 1H. The pixel element pair patterns for adjacent line pixels should be generated line by line - i.e. in 4 lines according to FIG. For this reason, the shift register 5 must be designed as a circulating shift register, since the gray value of a pixel is required several times when generating the pixel element pair pattern: For the first time when the first image sample

GE 9-75-003 609831/0534

¹ row of element pairs is generated, etc, and to the last and

! fourth time when the fourth row of pixel elements generates! ι will.

Since the gray values are required to generate the pixel element pair pattern for each of its k lines, the contents of the shift register are circulated four times via lines 8, 9, 10 from the output of adder 6 to the input of circuit k. The circuit 4 CT ensures that the values fed back on lines 8, 9 and 10 are returned from the output of the shift register to the input of the shift register and that this process takes place a total of 4 times for each line of pixels. To generate the pixel element pair pattern according to Pig. An adder 6 and a comparator COMP 7 are provided from the predetermined digital gray values. The adder 6 ADD is supplied on the one hand with the content of a shift register stage and on the other hand with a generation-specific value on the line 11. Then different values are added to the gray value for a pixel, depending on which pixel element pair A, B, C, D, E, P, G or H is to be generated from the gray value of the pixel. The value 0 is added to the gray value of the pixel to generate the pixel element pair A, the value 1 is added to the gray value of the pixel to generate B, the value 2 is added to the gray value of the pixel to generate the pixel element pair C, and so on. The values to be added are set in brackets in Pig. 5 after the pixel element pairs labeled A to II, 1 times the addition of the gray value for a pixel with the special; generation value follows, effected by circuit 7, a comparison. This comparison circuit determines whether the addition result S is smaller than the number of possible amplitude intervals (II-18) + I ₁ in this case, that is, 8 + 1 = 9 ^. If the comparison result on line 12 is affirmative, the corresponding area of the pair of pixel elements is printed black, in other cases xieiss (line 13). In this way, i.e. by simple addition with subsequent comparison, a gray-

GE 9-75-003 60 983 1/0 5 34

I - 19 -

! value for a pixel in the context of a fanned-out pixel j are determined in individual pixel element pairs according to FIG. 5, whether the individual pixel element pairs A to II should be black or white. This information is stored and processed in accordance with the procedure subjected to a code optimization.

The generation of the pixel element pairs for adjacent Pixel is ensured that before each cycle of the content of the shift register 5, its content through Addition of the generation-specific values is updated.

When fanning out pixels z. B. according to FIG. 1 or in another of FIG. 2 in connection with FIG. 5 different ways, ! the generation of the individual pixel elements from the gray values for a pixel cannot be carried out in such a simple manner Manner as indicated in Fig. 6 list. For these cases, adder 6 and comparator circuit 7 would be closed by a circuit replace, which accesses a memory for each given gray value of a pixel, in which the individual pixel element patterns are stored for the individual gray values.

GE 9-75-003 6098 31/0534

Claims (1)

PATENT CLAIMS Method for minimizing storage space of computer-coded data intended for halftone image printing, thereby marked, that an analog original halftone image of an equidistant Is subjected to rasterization, that for each original halftone image area assigned to a raster point (picture element) the analogue gray value and an interval of the gray value ska-Ia divided into ρ intervals (including the pure black value) is assigned, that each pixel of (E) is arranged in a matrix light and / or dark pixel elements is formed that for the entirety of all picture elements of all picture elements row or column the number (the length value) of similar (lighter or darker) adjacent ones Image element is determined, that all possible length values of a probability distribution to be subjected that according to this probability distribution, frequently occurring length values have a shorter one, which is to be stored in coded form Specification is assigned as rarely occurring length values. 2. The method according to claim 1, characterized in that on Huffman code optimization is applied to all possible length values. 3. The method according to claim 1, characterized in that the gray scale is divided into 8 intervals II, 12, ..., 18 becomes (7 gray values, 1 black value). GE 9-75-003 609831/0534 4. The method according to claim 3, characterized in that the fanning of a pixel into individual picture elements according to the scheme 3 7 8 6 6 8 7 3 takes place, the numbers indicated mean that the gray levels 12-18 are formed by the Level 18 the elements indicated by 8 are black, for level 17 the elements indicated with 7 are also black, for level 1.6 additionally those with 6 specified elements are white etc. 5. The method according to claim 3 _5, characterized in that the fanning of a pixel into individual picture elements according to the scheme 8 8 7 7 : 6 6 5 5 ..- "..- ■ 4 4 3 3 2 2 11 ; takes place, where the numbers given mean. That the Gray levels 12-18 are formed by the fact that for the j Level 18 the elements indicated with 8 are black, for ί the level 17 additionally the elements indicated with 7 \ are white etc. " ■. \ 6. The method according to any one of claims 1 to 5, characterized in that that ■ '' ■! ' for iext representations all possible length values of an Op-; timed by a logarithmic code i the. ■; _. I ₁ 603-8.3 170534- - ' GE 9-75-003 .7. Method according to one of Claims 1 to 5, characterized in that that for weather maps or line drawings all possible Length values are subjected to optimization by a linear code. 8. The method according to claim 5 _3, characterized in that two similar pixel elements lying in a line are combined to form a pixel element pair A, B, C, D, E, F, G and H. 9. Arrangement for performing the method according to claim 3, characterized in that a revolving 3-bit-wide shift register 5 comprising N steps is provided for the successive recording of the digital gray values II - 18 for Ii line pixels of an original halftone image, and that the associated pixel element pair pattern (Fig. 5) can be generated from a gray value That an adder 6 is provided through which the gray value of a pixel for the pixel element pair A the value O, for the pixel element pair, B the value I ₃ for the pixel element pair C the value 2, for the pixel element pair D the value 3, for the pixel element pair E is the value 4, for the pixel element pair F the value 5, for the pixel element pair G the value 6 and for the pixel element pair Ii the value 7 can be added and that the addition result S in a comparator circuit 7; It is possible to examine whether it is smaller than the gray level number IL to Ig plus 1 (i.e. 3 + 1 = 9), whereby the signal for an affirmative comparator result is a black pixel element pair, and a negative one, a white pixel element pair can be marked, and that the generation of the pixel element pair pattern; can be carried out for the N image points in pattern lines by circulating the shift register contents four times, this being acted upon by the adder ο 09 before each circulation GE 9-75-003 is, and that the generated pixel element pairs are computer program-controlled are subject to code optimization » 10. Arrangement for performing the method according to claim 1, characterized in that a rotating shift register 5 for recording the digital gray values for K line pixels of an original halftone image is provided, and that from a gray value the associated pixel element pattern by accessing all the pixel element pattern containing memory can be generated, and that the generation of the pixel element pattern for the N line pixels by pattern line by p- * times; Circulation of the slide register contents can be carried out,. and that the generated pixel element patterns computerpro can be subjected to a code optimization under program control. . _ _ ■ __6o9_83_r / _o_5_a4_ GE 9-75-003 Blank page