DE2851481A1 - ENCODER FOR IMAGE SIGNALS - Google Patents

Description

- 7 - 285U81

description

The invention relates to a coding device for image signals, which is particularly advantageous for coding standing pictures is used. To reduce the cost of image transmissions have been previously z. B. in "A Survey of Digital Picture Coding" by A, Habihi and G. S. Robinson in Computer, - 7, 5, pages 22-34 (May 1974) proposed coding systems such as DPCM, / \ M, and the like. These known systems are called predictive coding systems. With them there is a predictive signal which is based on a location preceding a reference point and is used for transmission an error signal which corresponds to the difference between the digit to be coded and the prediction signal, quantized. According to the property of the image signal, this prediction error signal is often in a range small amplitudes, so that even if it is only roughly quantized, the picture quality does not suffer significantly. As a result, the number of bits necessary for coding can be compared to the normal PCM, in which each digit is so as it is, is coded, reduced.

In this usual prediction Kodiersy star, however, must the error signal can be transmitted for each individual point

^1/2 909822/0858

and the number of quantization levels for the error signal can cause other disturbing influences on the image quality, such as B. granular noise, false contour usv /. cannot be extremely reduced. For these reasons, the number of bits used is on the whole relative high.

Has a novelty search of the European Patentante no specific literature pertaining to this invention has been provided, however, U.S. Patent Nos. 3,403,226, 3,940,555 and 3,984,626 cited as prior art. US-P 3,403,226, in which an image is divided into sections will be described here. In this patent, within each section one of the pictorial elements selected as typical and coded in PCM technology so that it itself represents its level. The difference between the typical picture element and each of the other picture elements of the same section is also PCM-coded.

Accordingly, although the number of quantization levels for the latter PCM code is smaller than that of the for the typical picture element. the transmission of the PCM codes a plurality of bits, which in total results in a remarkably large number of bits required.

909822/0858

-9- 285K81

An object of this invention is an encoder for Image signals that enable coding with fewer bits than the previous version.

Another object of the invention is an encoder in which the image is divided into sections and then encoded However, the boundary of each section in the .reconstructed Image is not recognizable.

Another object of the invention is an IL for picture signals in which the picture is set in sections whose size is changed depending on the local properties of the image, thereby reducing coding hit Bits is made possible.

The invention is characterized by the features specified in claim 1. "Advantageous further developments of the invention can be found in the subclaims.

According to this invention, an image composed of gray tones is divided into a plurality of sections. The picture elements of each section are dependent on the statistical distribution of their brightness values in gray values and Dissolving or distinguishing features classified and then coded. For example, within each section a comparison value that is dependent on the brightness distribution and the brightness value of each Image element then checked whether it is higher or lower than the comparison value. From the comparison results

909822/0858

iü

285H81

and the brightness values of the picture elements, two typical brightness levels are formed. These two typical Brightness level and the comparison results are transmitted in the form of codes for each section.

If a _n and a _{1 are} the typical brightness levels, x, the comparison value, x. the brightness level of each individual picture element and y. denote the brightness level of a picture again after the conversion by coding, the following relationship applies to the coding:

where it holds that φ. = 1 if x. ^ X. , φ. = 0 if x. <X, r> * i ' y £ = ti' it

and φ. the complement of φ. is. The typical brightness levels a _Q and a .. are gray values, φ. is a differentiator.

The typical brightness levels a_ and a .. and the comparison value x. are z. B. formed in the following way:

In a first method, a relationship χ = y is given, which means that the mean brightness values χ and y of the picture elements in each section before and after the Coding are unchanged. Furthermore, the average brightness

909822/0858

'"" 285ΊΑ81

value of all picture elements of a section is used as comparison value x, _ before coding, d. H. x. = x. The following mean square error in each section before and after the conversion

^N 0

C ² = ^ Σ (χ. -Y.) ^2
N oi ^xx

is minimized, with the typical brightness levels a _Q and a ₁ and the comparison value x. surrender. Denotes x. the brightness value of each picture element in each section, N _n the number of picture elements in the section, χ the average brightness of the picture elements in the section, x _T die

Yy

average brightness of that part of the picture elements whose brightness is smaller than the average brightness level x, N _{1 is} the number of these picture elements, χ the average brightness of the part of the picture elements whose brightness is greater than the average brightness level χ and N ~ the number of these picture elements, the mean brightness levels x, xT and ΈΖ result as follows:

Year 11

- i \

^X L ⁼

909822/0858

285U81

HN ₂

According to the condition that the middle square

Error £. at the typical brightness levels a "and a ..

Minimum values reach, d _o h. .— ^ - £ = O and -, <{- £ = O, the typical brightness levels a _Q and a "result

^a 0 " ^X L

^a 1 ^X H *

This method is called the two averages method.

In a second method, the typical brightness levels a _n and a ₁ as well as the comparison value χ from the ui ~ v.

above condition χ = y, the condition x_ = y, i.e. that

the exponential values of the brightnesses χ and y of the picture elements before and after the conversion in the sections remain unchanged, as well as the condition ^ - £ = 0, ie that the mean potentiated error is minimized depending on the comparison value x, derived. If χ is used to denote the mean brightness value of that group of picture elements whose brightness values are lower than the

909822/0858

Equal value, with N ₁ the number of these picture elements, with x "/ the mean brightness value of the group of picture elements whose brightness values are greater than the comparison value, with N _? the number of these picture elements, with χ the mean brightness value of the picture elements in each section before the coding and with S the deviation of the brightness level of the picture elements, the mean result

Brightness values x _T , x "and the deviation A as follows:

^N 1

Xjj,

^N 0

6 ² = ~ -Σ (X ₁ -

^N 0 i ^x

The comparison value x. arises to

^X t ~ 2

and . the typical brightness levels a _Q and a. to

F ²

^a 0 " ^X V (N ₁ ' ^a 1" ^X

909822/0858

This method is called the deviation method.

With this method, it would be in the interest of simplifying the procedure for forming the comparison value x. provided that χ. = χ, becomes the mean square Errors / not always minimized, the typical brightness levels however, a ~ and a .. are the same as above.

A third method consists in simultaneously fulfilling the conditions χ = y

ty 0 f? 0 / p * "*

3 - ^ - £ - ο; λτ-- £ = 0 and ^, - ζ * = 0. In the-

^3a o fc \ ^ t

in this case results for the comparison value x.

^x t ⁼ 1 ^(x i

and for the typical brightness levels a "and a.

Σ * i

^a 0 ^{= x} i ^ ^x t ^H 1

^a 1 - ^x i- ^x t ^N 2

This method is called the variable comparative value method.

According to what has been said so far, in this invention the typical brightness level Sq and a, - of each section

transferred as well as the distinguishing feature, which indicates

909822/0858

with which level a ₁ or a 1 each picture element is to be decoded. In other words, the information "1" or "0" is to be transmitted for each picture element, as well as the typical brightness values a "and a .. of this section for each section. Accordingly, the number of bits used is significantly reduced overall and, since the entire image is divided into sections, there is also an excellent reconstruction of the image,

According to this invention, the image is divided into individual sections and there is a risk that the boundaries these sections are visible. In order to avoid this, the brightness of a picture element outside the section becomes however, in the vicinity of its limit when determining the comparison value x, considered in the vicinity of the limit. Optionally, the comparison value, which indicates whether the brightness of the picture elements is close to the boundary of a section is above or below, depending on the comparison value im; subsequent section can be modified.

For example, in a case where the change in brightness is small in a particular part of the picture, even in a relatively large section, the degradation in image quality is relatively small. In such a case where the section is made large, the number of typical decreases

909822/0858

285H81

The brightness level and the number of bits can be reduced accordingly. On the other hand, in a part where the image changes greatly, the quality of the reconstructed image is markedly lowered even if the portion is made relatively small. Accordingly, the size of each portion is preferably changed in accordance with the property of the image. For example, a code conversion is carried out for a section of a predetermined size, and the absolute values of the differences between the brightness levels of the picture elements before and after the conversion £ ■ "Iy" XI r are added up for each section and depending on whether this sum is too small or too large, the block is made larger or smaller. It is also possible to use the squares of the absolute values £. add together for each section and influence the size of the section depending on the addition result. It is also possible to make the section small or large, depending on whether the difference D = a * - a _Q between the gray values a 1 and a ... is large or small.

The invention is explained in more detail below with reference to drawings. Show it:

. Fig. 1 is a block diagram of an invention

according to the embodiment of a coding device for image signals;

909822/0858

Fig. 2 shows an example of a in sections below

split picture;

Fig. 3 is a block diagram showing an example

for an encoder 14 as used in Fig. 1;

Fig. 4, for example, a block diagram for

a comparison value calculator 17, like him is used in Figure 3;

Fig. 5 is a block diagram for an embodiment

for example a _n, as used in Figure 3 of the computer ~. 19;

Fig. 6 is a block diagram showing an example

for a decoder for decoding a coded signal from the invention Shows encoder;

Fig. 7 is a block diagram of an embodiment

game for a decoder 42, as used in Fig. 6; ·

Fig. 8 is a block diagram showing another off

management form of the coding according to the invention

909822/0858

device shows using a microprocessor;

9 is a flow chart showing the basic

Shows process steps for encoding;

10 shows a flow chart of a read-in routine

for a section to be coded;

11 is a flow chart of a routine for the

Calculating the comparison value;

Fig. 12 is a flow chart showing a routine for

shows the calculation of the distinguishing feature and a flow chart for the calculation the typical brightness level;

Fig. 13 is a flow chart showing the essential Ver

shows operational steps for decoding;

14A are drawings showing the relationship between 14C

a section to be coded and a

Show reference section;

Fig. 15 is a flow chart showing a routine for the

Figure 9 shows reading of a reference section;

909822/0858

285U81

16A shows the relationships of a picture element up to 16C

at the border of the section and one

modified comparison value for a corrected, typical brightness level;

17 is a flow chart for a routine for

Modification of the comparison value;

18 shows the relationship between first and zv. ^T o; ^: , -

th sections;

19 shows a routine for reading in a first

Section;

Fig. 20 is a block diagram showing an example of

the case of controlling the size of the section in a device according to this Invention shows;

Fig. 21 is a block diagram for an example of one

Coding device according to FIG. 20;

Fig. 22 is a block diagram showing an embodiment

example of the section control circuit 71 in Fig. 20;

909822/0858

23A shows the relationships between large, to 23C

medium and small sections;

Fig. 24 is a flow chart for an encoding process

gear in a case of controlling the size of the section; and

Fig. 25 is a flow chart for a decoder

operation in the case of section size control.

In Fig. 1, a picture taken by a television camera 11 is shown Image quantized by an A / D converter 12 pixel by pixel and stored in an image memory 13. The one from the television camera 11 captured image is, as shown in Fig. 2, divided into, for example, N χ M sections, and each section consists of η χ m picture elements. That in the image memory 13 stored image is read into the coding device 14 in sections.

If x. denotes the brightness value of a picture element in the section, this brightness value becomes x. by the encoder 14 according to the following procedure in y. implemented:

909822/0858

²¹ "285U81

where φ. = 1 and φ. - 0, v / if x.> X and 0. = 0 and ^ T = 1, if x. <Fx. . The values χ. , a ~ and a ₁ are obtained by one of the three aforementioned methods from the distribution of the brightness values of the picture elements of the section. The coding device 14 converts the image information of the section into gray values a ″ and a. and distinguishing features (0Q, φ *, φ) (q = m χ n). These codes

are stored in a buffer memory 15, from which they at a constant speed suitable for the transmission path to which they are sent via a modem 16 will.

The coding device is constructed as shown in FIG. 3, for example. The output of the image memory 13 is connected to a comparison value computer 17, a comparison stage 18, an a ^ computer 19 and an a * computer 21. In the comparison value computer 17, for example, as shown in FIG. 4, the picture element signals coming from the picture memory 13 via the terminal 22 are added up one after the other in an adder 23 which is initially reset to zero. After the summation of all picture elements of a section has ended, the output of the adder 23 is applied to the dividend input of a dividing stage 24 and divided there by a divisor which is stored in the divisor register 25. The value of this divisor is equal to the number of all picture elements of the section m χ η and is given by a sig-

909822/0858

285U81

nal set at input 26. This way becomes the middle one The brightness value χ of the picture elements in the section is determined and results in the comparison value x. at connection 27.

The comparison value formed in this way is fed to the other input terminal of the comparison stage 18 in FIG. The comparison stage 18 compares the brightness value x. of each individual picture element with the comparison value x, and supplies the buffer memory 15 with the comparison feature (φ.), φ. and the value “φΤ inverted by the inverter 28 are passed on to the a .. computer 21 or the a_ computer 19 in order to obtain the gray values a» and a ₁ .

The image segment signals x coming from the image memory 13. are fed to a gate 29 via the terminal 31, for example, as shown in FIG. The distinguishing feature ^ T from the inverter 28 is also supplied to the gate 29 as a control signal via terminal 32. Depending on whether the distinguishing feature φ. has the value 1 or 0, the gate 29 is opened or closed. Accordingly, the picture element signals in the case where wox. <x. / ie {57 = 1 is, collected via gate 29 in an adder 33. In this way, all picture elements for the x. <Cx. is added up in the adder 33.

909822/0858

285H81

The number of gate control signals φ. at terminal 32 is counted by the divisor counter 34 and results in the number N _{1 of} those picture elements in each section for which x. «<x. The dividing stage 35 divides the addition result of the adder 33 by that in the diviscr counter 34

stored number of pixels N ₁ and thus forms the gray value a "at the terminal 36. The a ₁ calculator 21 in Fig. 3 is identical to that a_ computer built 19, and supplies the gray value a ₁ at terminal 37 obtained The thus Gray values a ₁ and a 1 are stored in the buffer memory 15 at suitable times.

The image signal encoded as described above is decoded in the following manner. For example, as shown in Fig. 6, the coded image signal via a connection 38 from the transmission link taken over and demodulated in a modem 39. The demodulated signal is first stored in a buffer memory 41 stored and then passed to a decoder 42, in which each picture element signal is decoded and then is written into an image memory 43. The content of the image memory 43 is processed by a digital / analog converter read out and converted into an analog signal, which is converted back into an image by a monitor 45.

909822/0858

The decoder is constructed according to FIG. A demultiplexer, which is controlled by a control signal via terminal 47 ', separates the signal from the buffer memory 41 into the gray values a "and a. as well as the distinguishing feature φ .. The components a _Q , a. and φ. are transferred to Putferrogict ^ r 48``, 49 and 51. The information φ. opens from the buffer register 51 or closes the gate 52 and 53 such that they either the information A _{is N} or A, sr in the pouf. ~ registers 48 'and 49' is stored, to pass through, thus providing at the output 54 the information for each picture element signal of section 2, ur available. So if on the Sandt? - side φ. = Or dhx <x ,, gate 52 is opened by the inverse output? 5T of an inverter 55 and lets the information a "in the buffer register through to terminal 54. On the other hand, φ is on the transmission side. = 1, dhx ^ Tx., The gate 53 is determined by the information φ. opened and forwards the information a... from the buffer register 49 to the connection 54. Since the position of each picture element in each section is known, the output at connection 54 is written in the picture memory 43 at an address corresponding to said position. When the picture element signals of all sections have been decoded in this way and the decoded information for an image is stored in the image memory 43, a still image of an image is reconstructed on the monitor 45.

909822 / Ώ858

The coding and decoding described above can also be achieved using a so-called microcomputer. For example, as shown in Fig. 8, the output of the television camera 11 is on for the purpose of converting it into a digital signal an analog / digital converter 12 and the output of this converter 12 is applied to the system bus 46. To the system bus 46 the image memory 13 and the digital / analog converter 44 are also connected. There is also a central unit (CPU) 47, a read only memory (ROM) 48 and a random access memory (RAM) 49 are connected to the system bus 4 6. The output of the digital / analog memory 44 is connected to the monitor 45.

ROM 48 stores an encoding ^c and decode program and the CPU 47 reads the program from step your ROM 48 made, it interprets, executes it and thus causes the encoding and decoding. The image memory 13 stores the output of the television camera 11 in digital form or the decoded image signal. The content of the image memory 13 is read out by the digital / analog converter 44 and converted into an analog signal for feeding the monitor 45. RAM 49 is used to temporarily store data required for coding and decoding.

Figure 9 shows a basic flow diagram for coding. At startup, the content becomes N of a line segment store

909822/0858

in a step S = 1, then the content M of a column section memory is set in a step S ₂ = 1. In one step S.-. The picture element signals of the section designated by the memory contents N = I and M = 1 are read into a temporary memory. On the basis of these picture element signals, the comparison value x is calculated _{in a step S 4.} In a step S ₁ . will the

L- Zj

Comparison value stored in a comparison value memory. Next, in a step 6, a distinguishing feature 0 _{TT is} formed from each picture element signal and the associated comparison value χ and the basic values, d, h. the typical brightness levels a _n and a ₁ are calculated. Each section is composed of picture elements in η-rows and m-columns, ie 1 = 1, 2, ... ' m and J = 1, 2, ..... n.

In step S ₇ the values a », a .. and Φ _{Ύ1 are} transmitted and then at step S ₀ the content M des

Column section memory increased by 1. In the next step Sg is checked whether the memory content of M has reached its maximum value Mm and in the case that the content of M is larger than the maximum value Mm of the content N is increased of the line-portion memory in a step S ₁₀ umd. 1 In step S ₁₁ it is checked whether the content N exceeds its maximum value Nm and if the content N is less than or equal to the maximum value Nm, the process jumps back to step S ₂ -

909822/0858

If in step S _q the content M of the column section memory is less than or equal to the maximum value Mm, the sequence jumps back to step S ^. If the content N _{exceeds the maximum value Nm in step S 11} , this means that the coding of the entire image has been completed and the process is stopped.

For example, FIG. 10 shows a read-in routine for a coding section for step S ". First, in a step S _12, the content M of the column section memory is set to the value M, and the content N of the row section memory is set to the value N of the section M, N to be read. The content J of a memory for the section line address is set to Λ _{in a step S 1 ^.} The content I of a memory for the section column address is also set to 1 in a step S ^. Next, in a step S ^. _{r is} the computation I + (M- 1) xm for an image line address

i and the computation J + (N-1) xn is performed for an image column address j. The brightness level of a picture element x .. with the address (i, j) is read in a step S ₁ ^ from the image memory 13 and then transferred to RAM 49th Subsequently, in a step S ₁₇ the content I of the section column address memory is increased by 1 and in a step S ₁₈ it is checked whether the memory content I is less than, equal to or greater than m. If I ^ m, the flow jumps back to step S .., - by changing the brightness of the next

_17/18 909822/0858

285H81

Image element is read out and transmitted. If Ι> ίπ in S ₁₈ , the content J of the section line address memory is increased by 1 in S ^ q and the result of this addition is checked _{in step S "o.} If J ^ n, the process jumps back to step S ₁₄ , while the read-in process is ended if J> -n. In this way, the addresses of the picture elements in each section are assigned stepwise and the picture element signals are read into a temporary memory.

A calculation routine for the comparison value x in step S ₄ in FIG. 9 is shown in FIG. 11. The routine begins with the contents of a summing memory SUM is set equal to zero in a step S _91st In the next step S ₃₃ , the content J of the section column address memory is made 1, and in a step S ₃₃ the content I of the section row address memory is set equal. In a step S _24, the direction indicated by the contents of I and J from the address memory pixel-signal x _T is added to the summing memory to the contents of SUM and stored in the addition result. Subsequently, in a step S ₉₁ . the content I of the section column address memory is increased by 1 and this increased result is checked in a step S_. If I ^ m, it jumps

Sequence back to step S ₃₄ , in which the brightness 18/19

903822/0858

levels of the picture elements in the section can be stored one after the other. If Ι> Ίη, the process in which the content of the J section line address memory is increased by 1 proceeds to a step S _97., The result of this addition is checked in a step S "o. If J ^ n, the flow jumps back to step _S33 , whereas if J> n, the flow advances to a step S _? q, in which the accumulated value SUM in the adding 'richer is divided by the number of picture elements in the section, mxn, T, whereby the comparison value x. is obtained, which is stored in the comparison value memory.

_{FIG. 12 shows the calculation routine} shown in step S in FIG. 9 for the variables Φ 1Ί ι ^ _n and a ... First, in a step S _3n, the contents S ₁ and S "of the first and second summing memories and the contents N ₁ and N" of the first and second counting memories are made zero. Then, in a step S _31, the content J of the section row address memory is made 1, and in a step S ₃₃ the content I of the section column address memory is also set to 1. In a step S ₃₃ , the brightness χ _χ of the picture element designated by the contents I and J of this address memory is compared with the comparison value x. compared. If x _TT <Tx., The sequence goes to a t υ t

Step S-. In which the content φ of a distinction 19/20

909822 / 08S8

iuerkmalsspeichers is set to zero at a corresponding address and the content N _{1 of} the first counting memory is increased by 1 _{in a step S 35.} The picture element signal χ

is added to the content S _{1 of} the first totalizing memory. If in step S -,., Χ> x, the flow advances to

JJ Xu - t

Step S ^ ₅₇ , in which the content Φ _{ΎΊ of} the distinguishing feature memory at the corresponding address is made 1. In a step S ™ the content N- of the second counting memory is increased by 1 and x _TT is then added to the content S “of the second totalizing memory. Steps S ₃₆ or S _{39 are} followed by a step S ^ _Q , in which the content I of the section column address memory is increased by one. The result of this addition is checked _{in a step S 41.} If Iism, the process jumps back to step S ₀₃ , in which the next picture element signal is compared with the comparison value. If I> m, the process proceeds to a step S _43, in which the contents of the J-portion row address memory is increased by the first The result of this addition is made in a step S _43, and when J ^ η, the flow returns to step S ^ "and when J> n, the content of S is ₂ of the second summing memory by the content of N" of the second count memory divided, with which the typical brightness level a _{1 is obtained} in a step S ₄₄ . Next, in a step S -, - the content N. of the first

45 1

Counting memory is checked, and if the content IvL is not zero, the content S _{1 of} the first summing memory is checked by the

9G9822 / 0858

Contents Ή _{Λ of} the first counting memory divided, which results in a step S ₄ , the typical brightness level a_. If in step S ₄₁ - N. = 0, the typical brightness level a "is set equal to the value a ...

Next, the decoding sequence will be described with reference to FIG. Assuming that the image memory 13 in FIG. 8 contains the components a _nt a. and φ. _{Γ of} an image

U I Xd

saved. First, in a step S0, the content N of the line _{segment memory} is made 1 and in a step S 4g the content M of a column segment memory is also set to 1. The components a _nr a. and 0 _{TT of} each picture element of the section in

U I J-ü

read into the RAM memory 49 in a step Sc _0. Then, in a step S ₁ -H, the components a »and a. written as the output value V _x - in the image memory 13 , namely at a corresponding address depending on whether the distinguishing _feature φ Ί 1 or 0 is. In the next step Sr, the content M of the column section memory is increased by 1, and this result is checked in a step S ₁ - *. If M ^ Mm, the flow jumps back to step S _Q in which a code of the next section is read into the temporary memory. If M • 'Mm, the content N of the line segment memory is increased by 1 _{in a step S 54.} Then the content N is checked and if N ^ Nm, the ab-

^21/22 909822/0858

³² "285H81

run back to step S. «and if N> Nm, this means the termination of the decoding of all picture elements and the decoding process is stopped.

The decoding sequence can also be achieved in the following way: first a picture element address (ij) is defined on the picture, the components a _n , a ₁ and φ. . of the section to which the address (ij) belongs are read out from the image memory 15, the decoding

y. . = φ. . a "+ φ. . a. is carried out and the decoded output y .. is written into the image memory 15 at the address (ij). A corresponding decoded output is obtained in connection with each of the other addresses in a similar manner.

With the encoder described above, it is possible to obtain a reconstructed picture with a small number of bits and excellent resolution. In particular, the comparison value x. and the typical brightness levels a _Q and a * are set so that the code conversion error between the picture elements of each section is minimized. In this way it is possible to obtain a reconstructed image which is of excellent quality when compared with the location of the original image corresponding to the respective section. However, in some cases there is a possibility that there will be translation errors at the boundary of the sections

²² 909822 / 08S8

pile up so that their contours in the reconstructed image visibly ground.

The boundaries of the sections in the reconstructed image can be made invisible by the following procedure. Regard we a section, the picture elements of which are to be coded (this section is referred to in the following as the coding section) and a reference section which includes and is larger than the coding section. For example, Figure 14A shows a coding section 51> which contains 8x8 picture elements, and a reference section larger than the coding section 51 by two picture elements on all four sides and consequently consists of 12x12 picture elements. Also, as shown in FIG. 14B, the reference portion 52 may be attached to two butting sides each have to be two picture elements larger than the coding section, 52 and therefore consist of 10x10 picture elements exist. Further, as shown in Fig. 14C, the reference portion 52 may be by two picture elements on one side thereof be larger than the coding section 52 and therefore consist of 8x10 picture elements.

When coding the picture elements of the coding section, the mean brightness level of the reference section 52 is used as the comparison value x. used. For the coding section 51, the distinguishing feature φ. and the typical bright

909822/0858

285H81

keitspegel a_ and a., in the same way as described above , won.

When the coding of the picture elements of the coding section 51 is carried out by a stored program, the reference section 52 is read into a temporary memory by means of the routine shown in FIG. In this example, the reference section 52 is related to the coding section 51 as shown in FIG. 14A and has the size rxr, where r = η + 2. In a step S ^ _6, the contents M and N of the column and Line segment memory is read out and the content J of the segment line address memory is set equal to 1 in a step S ^ _7. Subsequently, in a step S _110, the content I of the section column address memory is set equal to 1. In a step S ™, the calculations I + (M-1) χ m and J + (N-1) xn-1 of the row address j and the column address i are carried out. In a step S _fi0 , the brightness x... Of the picture element with the picture address ij is read from the picture memory 13. In a step S _fi1 the content is increased of the I section column address memory by 1, and this result is checked in a step S _62nd If I <r, the process jumps back to step S _1. ,, in which the next address is calculated. If I> r, the process proceeds to a step S, - ,, in which the content J of the output

DJ

cut line address memory is increased by 1. The result

909822/0858

285H81

This addition is checked in a step S, -, and if J ^ r, the process jumps back to step S, - _o , while if J> r, the coding process is stopped. In this way, the brightness of each picture element in the reference portion of the coding portion indicated by the contents N and M of the row and column portion memories is read into the temporary memory, and the comparison value x is calculated by the same method as shown in FIG. The components φ -, ^ ι a _n and a, become

υ υ ι

using the comparative value x obtained in this way. calculated in the same way as in Fig. 12 shown. By using the comparison value from the reference section as described above, it is possible to obtain a To avoid accumulation of coding errors at the boundary of the coding section.

With regard to the picture elements of a section adjacent to the coding section, a comparison value change between the two sections can be modified by referring to the comparison value of the neighboring section, so that the change in the picture is gradual. For example, let the comparison value of a coding section (MN) composed of 8x8 picture elements be χ (M, N) and the comparison value of a coding section (M + 1, N), which is adjacent to the section (M, N), be x _fc (M + 1, N), as in

S09822 / 0858

285U81

16A shown. Then the comparison value of a picture element column 54 of the section (M, N), which is in the third position next to the boundary line 53 between the sections, becomes equal to x. (M, N) is set and the comparison value of a picture element column 55 of the section (M + 1, N), which is in the third position next to the boundary line 53, equals χ (M + 1, N). These comparison values are shown in Fig. 16B, connected by a straight line 56 and the values χ, ~ and x _fc? / which lie on this line 56 at the positions which correspond to the picture element columns 57 and 58 of the section (M, N) in the second or first position next to the boundary line 53 are used as comparison values for these picture element columns. In a similar way, the comparison values for the picture elements of the picture element columns 59 and the section (M + 1, N) are given the values x + 1, N) are a _Q , and a. ,.

A routine for such a comparison value change is shown in FIG. 17. In a step S _fiI - the contents of the column and line section memories M and N are made, corresponding to the coding section (M, N). In a step S _eiZ , the content J of the section row address memory is made equal to 1 and in a step S _ß7 the content I of the section column address memory is also made equal to 1. In one

909822/0858

In step S _fi o the contents of this section address memory are checked and if 1 ^ I ^ 6 and 1 ^ J 02, i. h _c that is to be encoded pixel in a region 61 of Figure "16A, the procedure proceeds to a step Sg _9, while in the other cases, the procedure proceeds to step S _170th In step Sg ₉ , the content J of the section line address memory, the comparison value χ. . of the section (M, N) and the comparison value ^x -i- / μ μ 11 of the subsequent section (M, N-1) _{are stored in an address register P or a comparison value register T Q} or a comparison value register T_. In step S ₇₀ it is checked whether 7 <£ I ^ 8 and 1i £ j ^ L6, ie whether the picture element to be coded lies in an area 62. If this picture element is in the area 6 2, the flow advances to a step S ₇₁ , in which the memory content 9-1, the comparison value x. ,. and the comparison value χ. _1st of the adjacent section (M + 1.-N) _{are stored in the address register P or the comparison value register T n} or the comparison value register T ". If, in step S _70, the picture element to be coded is outside the area 62, the process proceeds to a step S ₇₂ , in which it is checked whether 3 ^ I <£ 8 and 7 ^ J- ^ 8, ie whether that is the case coding picture element is within the area 63 of Fig. 16A. If the picture element is in the area 63, 9-J is stored in the address register and the comparison values x. ... _{N) of} the section (M, N) and x _fc (M, N + 1)

^26/27 909822/0858

285H81

of the neighboring B3.ocks (M, N + 1) are stored in the comparison value registers T "and T_, respectively. If it is determined in a step S_-j that the picture element to be coded lies outside the area 63, the flow _{advances to a step S 74} , in which it is checked whether 1 <I <2 and 3j ^ J ^ S, ie whether the picture element is present in an area 64 of Fig. 16A. If this is the case, the content I of the section column address memory is stored in the address register P and the comparison values

^X t (MH) ^{and x} t iM-1 N) of ^{the sections} (MfN) and (M-1, M) are entered in the comparison value registers T 1 and T _n . If it is determined in step S ₇₄ , - that the picture element to be coded is not in the area 64, the flow advances to a step S - ,, -, in which over-

/ b

It is checked whether 3 <^ I ^ 6 and 3 ^ J- ^ 6, that is to say whether the picture element to be coded lies in a region that is not to be modified in the central area of the section (M, N) according to FIG. 16A. If this is the case, the procedure goes to a step S ₇₇ , in which the comparison value x. . . is entered in the comparison value register _{T Q.} Otherwise the address IJ is outside the block (M, N) and the routine is terminated.

In a step S ₇₈ , the calculation of the change in the comparison value is carried out on the basis of the content P des

909822 / 0S58

285H81

Address register P and the contents x. and χ, the comparison value registers Τ ″ and T _Q ″, which were stored _{in steps S 6} ″, S ₇₁ , S _{7 ″} and S _71. For example, this calculation looks like this:

(3 - P). X ₊ . ¹ + (2 + P)

This changed comparison value or the comparison value, de ν did not have to be changed in step Snn, is written into the comparison value memory in a ™ step Soq at an address IVy, which corresponds to the address of the respective picture element. Next, in a step Sp _0, the content I of the section column address memory is increased by 1 and in a step S _0n this content I is transferred.

o I

checks. If 1 ^ 8, the flow jumps back to the step Sg ₈ / in order to carry out the comparison value correction for the next picture element. If I>"8, the content J of the flow is increased of the section-line address memory by 1, and this content J is checked in a step S _83rd when J ^ .8, returns to step S _FI7 and if J> 8, the comparison value correction is completed for each picture element of the section (M, N).

The following method can also be used to prevent the coding errors from hitting the boundaries of the section.

²⁸ 909822/0858 bad or "g" nal

⁴⁰ "285U81

collect. For example, as shown in Fig. 18 shows the image frame divided into first portions 66, each consisting of 8x8 picture elements of which is _t, as indicated by the solid lines and, in second sections 67, each of which also consists of 8x8 image learning ducks, however, ', as indicated by the broken lines, is shifted by half a division in the row and column direction. One of adjacent picture elements belongs to a first section 66, the other to a second section 67. In FIG. 18, those pictures which are in odd-numbered rows and columns or in even-numbered rows and columns are identified and belonged to by black circles to a first portion 66, currency rend the remaining picture elements are indicated by white circles and belonging to the second portion 67. for each first portion 66 of the differentiator to be φ by the same type coding. and generates the gray values of a 'and a _1. Similarly, for every second section 67 the distinguishing feature φ 1 and the gray values a _{1 and a 1 are} determined in the same way.

For such coding, for example, the first sections 66 are sequentially stored in a temporary memory read in and the picture elements of each section are copied by a procedure as previously described.

28/29 " - ·

909822/0858

285U81

dated. Then, the second sections 67 are sequentially written into a temporary memory and the picture elements of each section are encoded. In connection with the example of FIG. 18, FIG. 19 shows a routine for reading in such first sections. In a step S ₀₄ the content J of the section row address memory and in a step S _fir the content I of the section column address memory are set equal to 1. In a step S ₀ , - the row and column addresses exi i = 21-1 + (M-1) x4 and j = 2J-1 + (N-1) χ 4 are calculated. In a step S " ₇ an address k = I + J-1 is calculated in the buffer memory and in a step S ₀₀

OO

the picture element x .. with the address ij is read out and written to the address k of the buffer memory. In a step S _O g, the content I of the section column address memory is increased by 1 and this result is checked _{in step S qf.} If II = L2, the process jumps back to

Step S ₀ c-, and if I> 2, the flow advances to step ob

Sg ₁ . In the step S _gi the content J of the section line address memory is increased by 1 and the result is checked in a step Sg. If Y ^ 2, the flow jumps back to step Sg ₅ . and in the case that J> 2, the flow advances to a step Sg., at which the content J of the section line address memory is made equal to 1, in a step Sg- the content becomes I of the section Column address memory set to 1. In a step Sg ₅ , image addresses i = 21 + (M - 1) χ 4 and j = 2J + (N - 1) χ 4, are

_29/30 909822/0858

285H81

calculates. In a step S _q ^ an address k = I + J + 3 is calculated in the buffer memory. In a step S " ₇ , the picture element signal ij is stored at the address k of the buffer memory. In a step Sg, the content I of the section column address memory is increased by 1 and in a step _Sqg the result is checked. If I <2, the flow goes back to step S _qc . and if I ? - 2 _r , the flow advances to a step S ₁ " _A , in which the content J of the section line address is increased by one. The result of this addition is _{checked in step S 101} , and if J ^ 2, the process goes back to step S _Q , and if J> 2, the process is ended.

The routine for reading the second portion 67 into a temporary memory is the same as the routine of Fig. 19 except that the addresses i = 21-1 + (M-1) x 4 χ 2 and j = 2J + (N - 1) χ 4 + 2 can be calculated _{in steps S 04} - and S _nc. The first

whether S) D

and second sections 66 and 67 do not always have to be shifted from one another by exactly half a division, as Fig. 18 shows, and as long as they are shifted from each other at all, it is possible to prevent the Concentrate coding errors at the boundaries of the sections.

A concentration of coding errors at the section boundaries 30/31

909822/0858

285U81

can also be avoided by correcting the typical brightness levels a 1 and a ... for picture elements which are close to the boundary of a section during the coding. This change can be achieved in the same way as is described in FIG. 16 for the change in the comparison value. That is, if the typical brightness levels of one section (M, N) are a and a ^ and the typical brightness levels of the adjacent section (M + 1, N) are a and a .., levels that are between the typical brightness levels a "and a", and levels that _{lie between the typical Helligskext levels a 1} and a ₁ , as indicated by black circles in Pig. 16C are used for the picture elements of the sections (M, IJ) and (M + 1, N) in the vicinity of the boundary line 53 therebetween. The calculation for the correction of these brightness levels can be achieved using the same procedure as previously described with respect to FIG. 17 with respect to the correction of the comparison value.

Following the above, the entire picture is divided into sections of the same size. But even if part of the Image in which the image changes gradually over a larger area covered by a larger portion is, "the quality of the reconstructed image is not deteriorated. Conversely, the quality of the reconstructed Image can be enhanced when using for an area

31/32.

909822/0858

285H81

large image changes a small section is selected. Accordingly, the size of each portion is preferably influenced depending on the property of the image in the particular area. An example of such control will be described with reference to FIG. The picture elements of a section are read from the image memory 13 into the coding device 14 'and coded in the manner previously described in FIG. 3, from which the typical brightness levels a ₁ and a 1 and the distinguishing feature φ. result. These components are not sent immediately, but instead decoded by the coding device 14 'and deliver the decoded signal y .. This decoded signal y. goes to a section control circuit 71, which generates the error £. between the original picture element signals x. and the decoded signal £. returns as well as the sum total of the squares £. ' the error 6. for all image elements in this section. Now, if this total sum is greater than a predetermined value, the section is reduced and its picture element signals are read into the coding device 14 ', encoded in a similar manner and the encoded output is again to the. Errors get decoded. This process is repeated until the total sum of the squares £. smaller

2 than the predetermined value. When the total sum of S · has become smaller than the predetermined value, the coded output is sent to the buffer memory at the same time

909822/0858

285H81

passed on. The coding device 14 'is designed such that a coding device 14 according to FIG. 3 also has the decoding function as shown in FIG. The distinguishing feature φ is controlled by a timing controller 75. from the comparison stage 18 and the gray values a ~ and a ₁ from the a ~ and a computers 19 and 21 are read into the registers 72, 73 and 74, respectively. The outputs a _Q and a. off the register! ¹ 73 and 74 are fed to gates 76 and 77, respectively, which in turn are identified by the distinguishing feature φ. can be controlled from register 72. If the distinguishing feature φ. is equal to "1", the gate 76 is opened while the gate 77 is controlled via an inverter 78. If the component φ. has the value "0", the gate 77 is opened. The outputs of these gates are combined and result in the decoded output signal y .. at connection 79.

In the section control circuit shown in Fig. 22, the original picture element signal x. and the decoded signal y., which both occupy the same position in the section under consideration before and after the code conversion, are fed via the image memory 13 or the connection 79 to a subtracter 81, which determines their difference, S- - Y- - x. forms. This difference £. is squared in a squaring circuit. The squared output £ .. is fed to an adder 83 for the purpose of adding up. As soon as the

2 2

total sum £ .. of squares £. of all picture elements of a

909822/0858

cut is formed, the output of the adder 83 is fed to a comparison stage 84, which it with a predetermined value To (in the case that the portion is composed of "8x8 picture elements) from one preset default circuit 85 compares. The comparison stage 84 generates an output signal D. = 0

2 2

or D ₁ = 1, depending on whether ^ ₁ ^ ^T o ^or £ · - ^ ^T ft ' ^If D ₁ = O ^ the coding control circuit 86 outputs a control signal to the timing controller 75 from FIG. and thus causes the contents of the registers 72, 73 and 74 in FIG. 21 to be transmitted as coded outputs to the buffer memory 15. Upon receipt of the control signal, the timing control 75 actuates a multiplexer 88 which uses the distinguishing feature φ. and the gray values a ~ and a .. are sequentially transferred from the registers 72, 73 and 74 to the buffer memory 15.

In the event that D ₁ = 1, the coding control circuit 86 gives the timing controller 75 a command to disassemble the block and at the same time controls the specification circuit 85 so that it changes its content to a default value T. which is the size of the then corresponds to the section to be subdivided. Up to this point in time, picture element signals were read in from sections each composed of 8 × 8 picture elements as shown in Fig. 23A. Now, as shown in Fig. 23B , the section is divided again

909822/0858

into four sections 91, each of which is made up of 4x4 picture elements exists and for each of these blocks the picture element signals are read into a temporary memory, on the like Way encoded and then decoded to get the decoded output

y. to obtain. This output will be the same as the original

Picture element x. compared to form the total sum ^ ₁ , which ultimately decides whether the comparison stage output D. receives the value "0" or "1". Since the number of picture elements in each section has changed to 4x4, the time control 75 issues a command to change the divisor to the comparison value computer 17 via a connection point 27 and thus changes the set value of the divisor register 25 in FIG. 4 from 8x8 4x4. If the output of the comparison stage, as a result of the coding of the section 91, which consists of 4x4 picture elements, has the value 0, the coded output is sent. However, if the output of the comparison stage has the value 1, the section 91 is further divided into four sections 92, each of which, as shown in FIG. 23C, consists of 2 × 2 picture elements. For each of these sections 92 a similar coding process for sending out the coded output takes place. In the section control circuit 71 of Fig. 22, encoded section size information indicating the size of the section is set in a status register 93 by the encoding control circuit 86, and when the encoded output is to be sent to the buffer memory 15, if the section size has been changed,

909822/0858

285H81

information about the new size is transferred to the buffer memory 15 via a connection 94.

To decode the coded signal that controls the section size, that for section size representative information from the received information taken out and based thereon an address is generated to use the same decoding method as shown in Fig. 7. to effect.

Fig. 24 shows a flow for the programmed execution of the picture element coding while controlling the section size. In a step S _{1 Q} ~ the content of the section line memory is set equal to 1 and in a step S _1n ^ the content M of the section column memory is set equal to 1. In a step S _1f ., The section consisting of 8 × 8 picture elements is read into a temporary memory and in a step S '"the components a_, a ₁ and φ. . for this section

calculates. The total sum ζ * of the squares of the error is calculated in a step S _1rtC from the components a _n , a., And φ. .

Ί Ub U Ί X]

calculated and compared in a step S ₁₀₇ with the predetermined value .Tg. Provided £ ,. ^ Tg become the components a _n , a ₁ and φ. . is sent in a step S ^ ₁₀₆ and in a step S'107 the content M of the section _{column memory}

909822/0858

increased by 1. Then in a step S ^ _nQ the-

Ί Uo

this content M checked. If M ^ -Mm, the flow jumps back to step S _{1 n4} , while if M> Mm, the content N of the section line memory is increased by one. The memory content N is checked in a step S ₁₁₀ and if N ^ Wm ,. If the sequence jumps back to a step S _1n -, but if N> Nm, the sequence is ended.

If, in the step S ₁₀₇ ,. £ ...> Tn, the content L of a line memory for middle sections is set equal to 1 _{in a step S 111 , and in a step S 11?} the content K of a column memory for middle sections is also set equal to 1. In a step S ₁₁ -. the picture element signals of a medium-sized section, which is denoted by L = 1 and K = 1 within the large section (M, N), are read in. In a step S ₁₁₄ , the components a _Q , a .. and φ. . calculated for the medium-sized section and in a step S ₁₁₁ - the pre-

mentioned total sum f. from the components a., a ₁ and φ. .

2 calculated. The total amount obtained in this way £ ..

is compared with the predetermined value T ₄ in a step S _11ß and if ^ ₁ ^ T ₄ , the components a _n , a ₁ and φ are in a step S _117. . sent. In addition, in a step S _118, the content K of the column memory for medium-sized sections is increased by 1, and this content K is shown in

909822/0858

a step S _1iq checked. If K ^ 2, the process jumps back to step S ₁₁ ^, while if K> 2, the content L of the line memory for medium-sized sections is increased by 1 _{in a step S 120.} The content L is then _{checked in a step S 1} _ .. and if L ^ 2, the process jumps back to step S ₁₁₇ , but if L> 2, the process jumps back to step S ^ " ₇ .

2 If it is found in step S ₁₁ that ^ ₁ > T., Does the procedure proceed to step S _1? ", In which the content Q of the line memory for small sections is set equal to 1, and in a step S ₁ " ~ the content P of a column memory for small sections is also set to 1. In a step S ₁ - ,. the image signals of a small portion indicated by P = 1 and Q = 1 in the central portion (K, L) and located within the large portion (M, N) are read into a temporary memory. _{The components a, f} a ₁ and 0 are derived from these picture elements. in a step S _{1? I} - calculated and sent in a step S ₁ ~>. In a step S ₁₉₇ , the content P of the column memory for small sections is increased by 1 and then this content P is checked _{in a step S 128.} If now P ^ 2, the process jumps back to step S ₁₂₄ and if P> 2, step S ₁₂ g follows, in which the content Q of the line memory for small memories

909822/0858

285H81

cuts is increased by 1. The content Q is _{checked in a step S 1} ^ _n and if Q ^. 2, the process goes back to step S ₁₅₃ , but if Q> 2, the next step is S ₁₁₈ .

Fig. 25 shows the flow chart for decoding a coded signal whose section size is controlled. The decoding process begins with a step S ₁₃₁ , in which the components a_, a * and φ ... of a section of the size corresponding to the section size information can be read into a temporary memory, and in a next step S ₁₃₂ it is checked whether the section size information is 8x8 or not. If YES, the read-in section is _{decoded in a step SI 33} and a subsequent step S ₁₃₄ checks whether the decoding of a complete image has been completed or not. If not, in a step S ₁₃₁ . the. Components a _n , a .. and φ. . of the next portion composed of 8 × 8 picture elements is read into the temporary memory, and then the flow goes back to step _S133 . When the frame size does not result in this step S ₁₃₃ to 8x8, the process proceeds to step S _136, where it is checked whether the section size information is 4x4 or not. If YES, the section of size 4x4 is decoded in a step S- ^ _n and then in a step S ₁₀₀

13/100

checks whether the decoding of the four (4x4) picture elements

909822/0858

285U81

sections is finished or not. If not, the components a ", a. and φ. . the next (4x4) -Bildelementabschnittes in the temporary memory is read, and after completion of successful decoding of four (4x4) -BiIdelementabschnitte in step S ₁ Oo, the flow returns to step S _134th If the result of step S _{136 is} NO, step S ₁₄₀ follows, which effects the decoding of the picture element sections with (2x2) picture elements. In step S ₁₄₁ , it is checked whether four (2x2) picture element sections have been successfully decoded or not. If not, the flow jumps back to step S _14n , but if the decoding is complete, the procedure proceeds

to step S- _OD .
I jo.

As described, in the image signal encoder according to this invention the picture frame is divided into a plurality of sections and for each of these sections at least two gray values and distinguishing features are obtained per picture element, so that the accuracy and the tonal value reproduction of the image is not negatively affected. In addition, is the number of bits used overall small, since not one for each picture element

909822/0858

A large number of bits can be coded and a high coding efficiency can be achieved. As the coding for each picture element by addition and which can be done for each section by division, the Coding can be easily carried out by a stored program using a microprocessor. the Structure for this can be made relatively simple compared with that formed from wired logic will. Since the distinguishing feature is a binary signal and corresponds to each individual picture element, damage Coding errors do not affect the picture quality significantly. The decoding consists in that a gate circuit through the Distinguishing feature is controlled and the structure therefor can be easily formed.

Furthermore, it is possible to avoid the accumulation of implementation errors at the section boundaries in that a comparison value of a reference section is used for each coding section which is greater than the coding section and overlaps it or by the fact that the comparison value in the vicinity of the section limits by the Comparison value of the adjacent section is corrected and decomposing the whole image into first and second section groups which overlap around adjacent picture elements to be assigned to the various section groups. It is also possible, although in the previous

909822/0858

Writing of the invention is not mentioned to make the sections in the reconstructed image invisible - that the boundary line between adjacent sections is made zigzag.

The efficiency of the coding can be further increased as a result are that the size of the sections changed depending on the image properties in the area under consideration will.

In the foregoing, the sections are described as rectangular, but they can also range from triangular to rhombic or a similar shape. In addition, the previously used

embodiment described only one comparison value and two typical brightness levels a _n and a ,, but a plurality of comparison values can also be used. In such a case, in order to form the distinguishing feature φ \, the picture element signals of a section are assigned to any of a plurality of level ranges defined by the plurality of comparison values. Accordingly , this distinguishing feature becomes φ. represented by a multitude of bits. The typical brightness levels in the corresponding level ranges are calculated from this distinguishing feature φ., Each pixel signal and the comparison values. In the previous embodiment, the mean value κ of the picture element signals of each Ab-

909822/0858

Average as a comparison value χ, used and the typical Brightness level a »and a .. v / ground so that the

Total sum E * of the squares of the translation errors x. - y.

of a section can be minimized with respect to the typical brightness levels ag and a ₁ and that the mean values χ and y of the picture element signals before and after the conversion can be equal to one another. However, as previously described, it is also possible to form the comparison value x and the typical brightness levels a _n and a ₁ according to the deviation method or according to the method of the variable comparison value.

In the embodiment described, in the case of the

2 controlled section sizes checks whether the grand total £ _{Λ of} the squares of the errors £. of each picture element before and after the conversion exceeds a predetermined value or not. But it is also based on the sum total of the absolute values of the error £ . possible to decide whether to change the section size. Moreover, in the embodiment described, a large portion is changed to a smaller one, but it is also possible to change a small portion to a larger one. In the previous, the number of section sizes is three, but it is not limited to this. The default values Tg and T ₄ , which decide whether the section size is changed, can also be set independently of the section size. In the embodiment described, if the section size is changed, the larger and smaller dimensions are

909822/0858

Sections of a similar shape to one another are defined, but the smaller section can also look so that its size is only reduced in the horizontal or vertical direction. In any case, the size of the larger section should preferably be an odd multiple of the smaller section. In addition, in the case of a section with a slight change in brightness, it is possible to use only the mean values of the typical brightness levels a _Q and a. to be used as an output, ie only a typical gray value instead of the two typical brightness levels ·. At a point where there is a large change in brightness, the eyes do not distinguish sharply between absolute brightnesses. In view of this, the number of bits that quantize the typical gray level can be reduced for a portion where the smoothness level changes greatly. This further reduces the number of bits used. Although this invention is applicable to still picture coding, it is also applicable to moving picture coding as long as the coding is done at a sufficiently high speed compared to the moving picture speed.

909822/0858

Claims (11)

BLUMBAOH · WESER. BERB ^ N · KRAWER ZWIFJNEFe · DEER · EaREHM PATENT LAWYERS IN MUNICH AND WIESBADEN Patentconsult RsdeckestrsSe 43 800Q Munich 60 Telephone (089) £ 33603 / ÖS3604 Telex 05-212313 Telegraimie P.3! «Rr.consult Pdtenlconsul! Sonnenberger Straße -13 6200 Wiesbaden Telephone (06121) 5629-Ί5 / 561998 Telex C4-1S6 237 Telegrams Putenicon NIPPON TELEGRAPH AND TELEPHONE PUBLIC CORPORATION 78/8771 1,6, 1-chome, Uchisaiwai-cho, Chiyoda-ku, Tokyo, Japan ENCODER TOTG FOR D J LD S.I Gl-; ALE Patent claims • Ί J Encoding device for image signals in which one of gray tones existing image is divided into a plurality of coding sections and in each of these sections Picture elements are coded, characterized by means for a comparison value formation, for each coding section at least one based on the brightness distribution in this Form comparison value, means the brightness level of each picture element of each coding section Munich: K. Kramer Dipl.-Ing. · W. We * er D.pi.-Phys. Or. Rer. nat. · P .: Ui &. Dipj.-ing. · W. p. Bre .- '. I D: pL-CI st.i. C · '. f: 'ii not. Wiesbaden: PG BlumbuCi; Dipl.-iPCi. ■ P. Bergen Dp! .- lr; g Cr jur. · G Zv / .rner Dipl .- 'ng. Oip '.- W-ing. 909822/0850 ORIGINAL INSPECTED 285H81 to form a distinguishing feature of one of the brightness levels defined by the comparison assign and means for calculating typical brightness levels for the formation of lBiijdGStens two typical brightness levels from the assignment result and the brightness level of the images. "! renew each coding section. 2. Coding device for image signals according to claim 1, characterized in that said means for forming the comparison value use this comparison value form the basis of the brightness distribution of a reference section, the said coding section contains and is larger than this. 3. Coding device for image signals according to claim 1, characterized in that it also has comparison value correction means contains which the comparison value ■ for picture elements near the boundary of the mentioned Select the coding section so that it is between the comparison value and that of the neighboring section lies. 909822 / 03S8 4. Coding device for image signals according to claim 1, characterized in that the image frame v; pus in other coding sections is divided, which are shifted from the original coding sections υπ thereby one of neighboring picture elements one of the Assigning coding sections and the other to another coding section, both coding sections independently encoded from each other. 5. coding device for image signals according to claim 1, characterized by the presence of the following additional Characteristics: a decoding device for decoding the "picture elements - te of each coding section based on the distinction sine characteristic and the comparison value, Error calculating means for calculating the error in coding the coding section from each decoded Picture element and the associated original picture element, Means for changing the section size comparing the calculated error to a given value and changes the size of each coding block depending on the compared output values, Means which, if the error remains within a given range, the distinguishing feature and the typical 909822/0858 see brightness levels as real coded output quantities provide. 6. coding device for image signals according to claim 5, characterized in that said means for changing the section size increases the size of the coding section decrease if the error exceeds the specified value. 7. coding device for image signals according to claim 5, characterized in that two coding sections were chosen before and after the conversion so that the size one of the two sections is an integral multiple of the size of the other. 8. coding device for image signals according to claim 6, characterized in that the distinguishing feature consists of one bit per picture element and that the number of typical brightness levels is two and that the mentioned decoder with one of the logical values of the distinguishing feature the one of the two emits typical brightness level as output value and 909822/0858 with the other logical value the other typical Brightness level. 9. coding device for image signals according to claim 1, characterized in that the means mentioned for forming the comparison value form a comparison value, that the distinguishing feature is information which consists of one bit per picture element and that two typical brightness levels are used. 10. Coding device for image signals according to claim 9, characterized in that the device for forming the comparison value receives an average value of the brightness levels of the picture elements of a section and that the aforementioned device for forming the typical brightness level mean values a ″ and a-j of the brightness level calculate those picture elements which are below and above the comparison value. 11. coding device for image signals according to claim 1, characterized in that said means for forming the comparison value, said means for Establishing the distinguishing feature and the mentioned 909822/0858 Means for calculating the typical brightness levels using a common microprocessor formed v / earth. 909822/0858