EP1779527A2 - Channel coding method with associated code and demodulation and decoding method and devices - Google Patents

Abstract

The information to be coded is coded by information bit or is recovered by decoding. In each coding step, the actual information bit is combined with a specific number of bits from the XOR code thus arising. During decoding, decisions in the form of a majority decision are taken with regard to the results between specific symbols from demodulation or directly between specific signal sections, preferably using fuzzy XOR operations.

Description

 description

Channel coding method with associated code as well

Demodulation and decoding methods and devices

This invention relates to channel coding and the associated decoding of signals related to one-dimensional signals (e.g., timing signals) as well as to two-dimensional signals (e.g., images), corresponding encoding and decoding devices.

Very popular and widespread are Reed-Solomon (RS) codes. The decoding of RS codes is purely algebraic and very difficult to understand and unintelligible to the user; there are no ready-made recipes for the integration of soft-decision methods. If one wishes to realize application-specific optimizations of an overall system (with coding, modulation, channel, demodulation and decoding), one must use the known (algebraic) coding and decoding as a finished black box in the overall system; a system optimization can then be realized only around this box, genuinely interlocked modifications, e.g. in connection with the modulation method used, and the application-specific integration of soft-decision methods are hardly possible in practice (with soft-decision, for example, the decodulator can inform the decoder with what certainty it is making its decisions This information is used by the decoder for decoding by optimizing a numerical security measure extending over several such decisions).

More recently, the so-called turbo codes have been introduced and successfully used. In turbo codes, two concatenated convolutional codes separated by terleaving, iteratively decoded by soft-in-soft-out Viterbi decoding. Turbo codes can in certain cases decode astonishingly strongly disturbed signals, but they also have disadvantages: the iterative method of operation leads to unpredictable computing time; Turbo codes require convolutional codes with infinite impulse response (HR). Such codes carry the risk of catastrophic error propagation. Good convolutional codes can only be designed using computer-aided design; there are no systematic design methods for "good" convolutional codes (even the definition of a quality criterion is difficult). The effects of Turbo Codes are very difficult to understand, so designing and optimizing is very difficult.

The prior art error correcting coding is described in R.H. Morelos-Zaragoza: "The Art of Error Correcting Coding"; Wiley 2002, ISBN 0471 49581 6 ".

The general object of the invention is to provide a robust, technically easily realizable coding and decoding with associated devices whose mode of operation is transparent in their individual steps, so that for the optimization of an overall system at various points, application-specific modifications possible are.

Special task is the integration of demodulation and decoding. In particular, the data should be able to be differentially evaluated in order to bring about greater security against low-frequency signal fluctuations.

According to the invention, this object is achieved with the objects according to the independent claims. Further embodiments according to the invention are defined in the subclaims.

Due to the use of comparisons, the decoding automatically adapts to fluctuations in signal properties. The mode of action is in theirs Single steps transparent. This greatly simplifies application-specific modifications for self-referencing and integration of the demodulation into the decoding process for the optimization of an overall system, with reliability-optimizing decoding directly at the signal level.

According to the invention, therefore, a method is provided for coding information that can be represented by partial information, wherein cells are to be assigned cell values by the coding, characterized by

a) partial information coding steps, b) partial information-specific classification of the cells in inversion cells and non-inversion cells, each partial information being assigned at least one inversion cell c) at least one partial information-specific cell tuple, each cell tuple comprising at least one

Inversion cell and at least one non-inversion cell, wherein the cells within each cell tuple are either only inversion cells or only non-inversion cells for each preceding coding step. d) for at least one coding step:

XOR linking of the relevant sub-information with the inversion cells assigned to this sub-information from b).

According to one development of the invention, it is provided that in the at least one coding step all inversion cells of the coding step belong to at least one cell tuple.

According to a further embodiment of the invention, the cell tuples are cell pairs with an inversion cell and a non-inversion cell. - A -

Signal values, in particular pixel values of images, can be assigned to the cell values.

In this case, the pixels can each assume a value of binary values.

A further embodiment of the invention, wherein a plurality of pixels are assigned to a cell, wherein the pixels are the pixels of an image or an image sequence or the signal values of another signal, is characterized by assignment of pixel patterns to the cell values.

A further embodiment of the invention is characterized by a differential coding of the part information processed first in advance.

A further embodiment of the invention comprises: XORing the result with a predetermined pattern, wherein the predetermined pattern in pictures is a two-dimensional checkerboard pattern and in the case of one-dimensional signals is a 101010... Pattern.

An alternative embodiment comprises: XORing the result with a random pattern.

In a further embodiment of the invention, for at least one step, a placement of the cells, in which at least 3 sets of cell pellets can be formed, which are separated from one another via at least one cell, can be provided.

Also provided according to the invention is an encoding device for information that can be represented by partial information, wherein cells are to be assigned cell values by the coding, characterized by registers with cells, where at least one cell is invertible in coding step by step as a function of partial information per register, preferably via XOR. Gate, wherein at least one invertible cell is interpretable as an inversion cell of one or more cell tubers with inversion cells and non-inversion cells, wherein the cells within each cell tuple for each preceding coding step are either only inversion cells or only non-inversion cells.

The registers can be arranged in a pipeline.

The invention further comprises a transmitting device which uses a coding method according to one of the embodiments with a phase modulator for at least one feature dimension.

Also included is a code generated by a coding method according to an embodiment of the invention, and a storage medium on which data is stored having a code formed by a coding method according to an embodiment, and a transmission protocol having a code formed by a coding method originated according to one embodiment.

In one embodiment of the invention, a fixed synchronization or referencing pattern is added to the transmission protocol after encoding, preferably in the form of a 010101 series of at least 6 cells.

Also provided according to the invention is a marking device with deforming marking, in particular needle embossing, with an encoding method according to the invention for the marking, the deformation strength being modifiable, preferably via the stop strength or the stroke of the needle, and a marking device with deforming marking, in particular scratch embossing, with a coding method according to the invention for the marking, wherein the direction of line elements which can be connected to one another is modifiable for signal modulation. Further provided is a labeling device with a Codierverfah ren for the lettering according to the invention, wherein for signal modulation, the width of lines or the size of dots or the color of dots or the size and color of dots or the direction of bar elements is changeable.

The invention also encompasses a code carrier for identification purposes with a code which has been created by a coding method according to the invention.

Furthermore, the invention also encompasses a method for decoding a code, such as can arise from a coding method according to the invention, wherein the information to be decoded can be represented by partial information and where the partial information is obtained from cells to be decoded, if it is referred to as an encoding method according to the invention emerged,

marked

A) by partial information decoding steps,

B) in that, for at least one decoding step, the partial information from cells to be decoded is obtained by cell-cell-wise comparison of inversion cells with non-inversion cells and comparison-result-dependent determination of the relevant partial information,

C) by comparing result-dependent linking of an inversion information with the inversion cells of the relevant decoding step, provided that at least one decoding step according to B) follows. The comparison of inversion cells and non-inversion cells can be carried out by comparison of cell values which are derived from the cells.

The comparison of inversion cells and non-inversion cells can also be done by comparing the contents of the cells in question.

A comparison result can be determined by evaluating several comparisons of cell values according to a quality function, preferably via the median of individual comparison measures.

- by the arithmetic mean of individual comparison measures.

A comparison result can be determined by evaluating cell values according to a quality function, by comparing mean values via inversion cells on the one hand with mean values via non-inversion cells on the other hand.

A comparison result can also be determined by evaluating a plurality of comparisons, preferably via the median of comparison functions to a plurality of pixels of the respective cells or

- About the arithmetic mean of comparison functions on each several pixels of the cells in question.

Also in the last decoding step, a link to an inversion information can be carried out and the cell values after the last decoding step can be interpreted as an error or error signal for a quality assessment.

In a further embodiment, the invention comprises a decoding method with a multi-valued logic in the coded cell values, in particular with corresponding multivalued phase modulation at the signal level and with polyvalent fuzzy logic at the cell values to be decoded.

Also provided according to the invention is a decoding device for information which can be represented by partial information, from a code in which cells are assigned cell values in each case,

marked by

a number of registers each comprising register cells, of which at least one register is used for receiving the information to be decoded, and wherein at least one register cell can be reversibly decoded step by step depending on the decoded partial information, preferably via XOR Gate, wherein at least one invertible register cell as an inversion cell of one or more cell tubers with inversion cells and non-inversion cells is inter¬ pretierbar, wherein the cells within each cell tuple for each subsequent decoding step either only inversion cells or only non-inversion cells, the partial information is obtained from register cells can be determined by cell-cell-wise comparison of inversion cells with non-inversion cells and comparison-result-dependent determination of the relevant partial information.

The decoding device may comprise summation and threshold elements with which it is possible to decide on partial information to be decoded.

At least one comparison can be realized by a fuzzy XOR gate.

Also provided according to the invention is a decoding device with a decoding method of the invention for a code in which first coded partial information is coded differentially, characterized by a circuit the decoded partial information for the inversion of the differential coding can be temporarily stored.

Further provided is a receiving device with a decoding device according to the invention, as well as a code reading device for a code, which can arise according to one of the coding methods of the invention, characterized by an image recording device in which at least one image recording parameter, in particular Lighting parameter, can be influenced by a quality measure that is dependent on an error image.

The invention will be explained with reference to examples and the drawing. In the drawing show:

1 to 10 an introductory explanation of the invention by way of example,

11 an indication of inversion cells,

12 shows a hardware realization for an example,

13 is an explanation of FIG. 12,

14 shows the general scheme of the message transmission, FIGS. 15 to 19 show various implementation variants,

20 and 21 an explanation of the term "scattered placement",

Figs. 22 to 31 show an example of multi-dimensional auxiliary arrangements.

First of all, a preliminary remark on terms used here is required, since known terms for this invention are partly used in a modified sense, and new terms are introduced.

An image is referred to as a two-dimensional signal, a one-dimensional signal also as a time signal. In the case of one-dimensional signals in which the signal values are modulated simultaneously according to several dimensions (eg according to amplitude and phase at the same time or in the case of quadrature amplitude modulation), here becomes Distinction of several feature dimensions (although these signals are simply called "multi-dimensional signals" in the coding literature). A color image signal is accordingly addressed as a two-dimensional signal which is modulated, for example, in the three independent feature dimensions color angle, saturation and intensity.

An image may be preprocessed into a one-dimensional signal, e.g. by scanning a barcode (1-D code).

A matrix code is a two-dimensional array of cells, e.g. in the popular ECC-200 standard.

The general block diagram of the message transmission with channel coder, modulator, channel, demodulator and decoder is shown in FIG. 14: The information u to be transmitted is supplied to a channel coder COD with redundant coding x, which in turn is supplied to a modulator MOD which receives the signal S generated. The coding takes place with the aim that, after a faulty transmission via the channel CH (disturbed signal S *), the original information can be reconstructed again. The demodulator provides receive values y, which are converted by the decoder into an estimate u of the original information u.

The information is represented by an alphabet of information symbols. In the binary case, the information symbols are infobits with the symbols "0" and "1", another example would be a ternary alphabet with the symbols "1", "2", "3". One or more information symbols occurring (in particular: infobits) are summarized in the following partial information. The information thus consists of partial information.

The encoder provides code symbols (code bits in binary case) at the output. The modulator transforms the code symbols into a signal that can be decomposed into cells.

In the case of image signals, cells represent image areas of at least one pixel (pixels), for example the (possibly pre-segmented) modules of a two-dimensionally printed matrix code or the image content of small parcels which uniformly scans the overall image. The cells can also consist of only one pixel, then the term cell is identical to the term pixel. The image areas are La, but not necessarily, contiguous.

The same applies to time signals where the cells (i.a. short, coherent) signal ranges represent at least one signal value. To simplify matters, the term pixel is also used here for individual values of any one-dimensional or multidimensional signal.

The content of cells can be represented by cell values. In general, cell values are symbolic or numerical representations of the contents of cells. For example, when modulating a matrix code, a logical 0 cell (cell value 0) may be realized by a vertical bar and a logical 1 cell (cell value 1) by a horizontal bar, the areal extent of an otherwise unspecified spot ( Dot), or simply by dark or bright pixels, different colors, etc. Cases with more than 2 possible cell values are eg 1 for red, 2 for green, 3 for blue. For time signals, the cell values are i.a. Symbols for the usual points in the signal space, but they can also be symbols for very specific geometries in the time domain (e.g., long / short after Morse code, or distinctly different geometries, such as "saw tooth rising" vs. "saw tooth falling").

In the case of several feature dimensions, the cell values can exist in a dimensioned manner, but the cell values can also represent several dimensions in summary. The modulator transforms the code symbols into a signal which can be decomposed into cells, whereby ia, but not necessarily, a one-to-one correspondence exists between the code symbols at the output of the coder and the possible implementations of the cells at the output of the coder Cell values can be represented. In the examples below, we assume such one-to-one correspondence and look at cell values just at the output of the encoder, with the downstream modulator generating the cell contents from the cell values.

In general, therefore, the task of coding is the generation of cell values.

Conversely, at the demodulator output, cell values appear as reception values y, in which case i.a. the number of possible cell values is greater than the number of possible cell values in coding. Thus, for example, in the case of a binary input alpha at the demodulator output, ideal "1" signals with the cell value "9" and ideal "0" signals with the cell value "1" appear, with arbitrary, e.g. all intermediate values 2... 8, in order to express how well the cell content corresponds to the corresponding ideal signals. "5" lies exactly in the middle and expresses that the cell content can not be demodulated. On the basis of such cell values, soft-decision methods can be realized. Of course, the demodulator can also provide "0" - "l" values separately from reliability values.

In conventional systems, the modulation is performed after the channel coding and the decoding after the demodulation, each independently. According to this scheme, the demodulator converts the ia disturbed signal into output symbols of an output alphabet y. With a conventional hard decision, the number of output symbols y of the demodulator is identical to the number of symbols x at the input of the modulator; the decoding is purely algebraic. Soft-Decision has the number of output symbols y of the demodulator greater than the number of symbols x at the input of the modulator.

However, depending on the implementation variant, no strict conceptual separation between coding and modulation or between demodulation and decoding is carried out here. Similar to the known coded modulation (eg TCM: trellis coded modulation), these limits are hereby abolished: Generally speaking, the decoding comprises the stepwise extraction of the partial information either from cell values (demodulation and decoding separated) or from cell contents, ie directly from the Signal (demodulation and decoding integrated).

The latter is the case with a decoding method according to the invention which is based on a direct comparison of cell contents; This comparison thus takes place directly on the signal level. Thus, e.g. large dissimilarity of two cells represented by the value "9" and high similarity by the value "1", with any, e.g. whole intermediate values 2 ... 8, to express how similar the cell contents (signal sections) are. "5" lies exactly in the middle and expresses that the comparison of the cell contents neither speaks for "similar" nor for "dissimilar".

Side note: the examples "1" ... "9" listed in this patent application are chosen only for the sake of clarity and can of course be replaced by any other evaluation schemes.

Compressive source encoding or decoding or other preprocessing / postprocessing (e.g., differential precoding) is possible and will not be considered further here.

Cells and cell values are to be understood in the sense given above, not only for images, but also for one-dimensional signals. Starting point for the co-operation Dierung is an arrangement of cells with preferably identically occupied Zel¬ len. According to feature d) in claim 1, XOR operations of the partial information to be coded with the inversion cells of the coding step are performed for a coding step. As a result, after the last coding step, the cells assume the desired cell values. This last coding step may be followed by further post-processing steps, such as the XOR operation with a checkerboard pattern or a pseudorandom pattern, see below.

In the invention reference is made to XOR connections or case-dependent inversions; These operations (or corresponding circuits and gates) are to be understood in the general sense given below and further explained:

- in conventional binary logic, or - in multivalued logic, or

- in binary or polyvalent fuzzy logic.

The invention is first explained by means of an instructive example, Figs. 1 to 10; Subsequently, based on the example, the procedure for generalization will be described, as well as special additives.

The example Fig. 1 to 10 comes from the field of image processing; the transfer of the following discussion to one-dimensional signals is entered into places where portability may not be obvious.

Example FIGS. 1 to 10: Information to be encoded with 10 bits is to be encoded, the information bits 1 to 10. In this example, the partial information is one bit each, and 10 encoding steps are required for each one bit ¬ derlich. Another subdivision into partial information would be, for example, with the possible cell values 1, 2, 3; then correspondingly fewer coding steps would be required. The same applies, for example, if you combine two bits each and work with four possible cell values.

Given is a two-dimensional image signal with 10 x 3 cells. The 30 cells are to be used redundantly for coding the 10 information bits. The cells consist in the example of individual binary pixels, possible cell values are initially 0 or 1.

The arrangement of the cells does not necessarily correspond to the final arrangement; the arrangement used for the data transmission or data storage (for example inscription with a dot code) will as a rule deviate from the arrangement described here because at the end a scattered placement, similar to interleaving, is performed (see below).

In the case of time signals, it is also helpful to arrange the cells mentally in a two-dimensional grid. The arrangement based on the following is therefore only helpful for the following considerations and is usually subsequently changed, in particular by a scattered placement.

Encoding:

As initialization, all cells of the code to be created are given the same basic value, e.g. 0, assigned.

Side note: the allocation of the same basic values is not technically necessary; one could assign differently and, when decoding in the case of the comparison operations between every two cells (see below), invert the interpretation if the initial assignment of the relevant cells is different. We define for the first coding step a set Ml of cells, which in our example specifically comprises all cells, designated by "i" in FIG. These cells in Ml are all now inverted if the first information bit is 1; if it is 0 they are all not inverted. In other words, the cells in M1 are XOR-linked to the first information bit, (assuming the base value 0, this corresponds to a copy of IO to the cells of M1). Cells of cell quantities treated in this way are called inversion cells in the following and are represented by "i" in FIGS. 1 to 10. By contrast, uninversion cells remain unchanged, irrespective of the value of the information bit, represented by "n" in FIGS. 1 to 10.

This first step in this example does not exactly correspond to claim 1; however, the following steps are in accordance with claims 1 and 2:

In general, for each following subinformation - i. for each encoding step according to the invention - there is a set of inversion cells (the remaining cells are non-inversion cells) and a set of cell tuples 1, wherein a cell tuple consists of at least one inversion cell and at least one non-inversion cell of this partial information. In principle, not every inversion cell needs to belong to a cell tuple. Inversion cells and non-inversion cells may belong to several cell tubers.

The partial information-specific - and thus dependent on the coding step - division into inversion cells and non-inversion cells and the definition of the partial information-specific cell tuples are referred to below as step configuration, the step configurations of all claimed steps taken together briefly called the configuration. The configuration is predetermined in advance, manually or with computational program, and is known to the encoder and decoder. In the example Figs. 2 ... 10, each inversion cell belongs to a cell tuple, and each cell tuple consists of exactly one inversion cell and one non-inversion cell, and one inversion cell belongs to only one cell tuple at a time.

Since it has to be coded a total of 10 information bits, we still define a total of 10 - 1 = 9 sets of cell tuples.

In our example, each cell tuple consists of exactly one inversion cell and one non-inversion cell. Some cell tuples 1 are indicated in the figures as double arrows between each one inversion cell and one non-inversion cell.

In Fig. 2, the amount M2 of cell tubers comprises six inversion cells and six non-inversion cells. There is no need to use any particular geometric system in choosing the distribution of inversion and non-inversion cells and in the choice of cell tuples, as indicated by the arrows. There are also basically overlaps allowed, i. A cell may belong to several cell tuples.

The inversion cells of the second step are XOR-linked to the second information bit. That if they have already been inverted due to the first information bit and are now inverted on the basis of the second information bit, they again assume the original basic value.

We come now to the third information bit and to the cell tuple quantity M3, see Fig. 3. The set M3 comprises in our example exactly the same cells as M2. In the choice of inversion cells and non-inversion cells, again, no special system needs to be used. However, when defining the cell pairs starting with M3, an additional rule must be observed (which is already fulfilled for Ml and M2 in each case): The cells of a cell tuple must have all been treated the same in XOR operations of previous steps, regardless of the current value of the preceding information bits. This is fulfilled if, for each preceding coding step, they are in each case either all inversion cells or all non-inversion cells.

A cell tuber forbidden in this sense is indicated in Fig. 3 under No. 2. It is forbidden because its non-inversion cell was inverted depending on the second information bit, but its inversion cell was not.

This rule did not need to be followed for the previous steps, because in the example their cells were always satisfied anyway (all cells of M2 are in Ml inversion cells).

In the next step with information bit 4 and cell tuple quantity M4, see Fig. 4, we again have free choice of cell pairs in the particular choice of M4, as before in M2.

In the next step with information bit 5 and cell tuple quantity M5, see FIG. 5, we again have the same situation for the special choice of M5 as for information bit 3 and cell quantity 3.

In the next step with information bit 6 and cell tuple quantity M6, see FIG. 6, we again have the same situation for the special choice of M6 as for information bit 2 and cell quantity 2.

We come now to the information bits 7 to 10, with the cell tuple sets M7 to MIO, see Fig. 7 to Fig. 10. Due to the special set selection and choice of inversion / non-inversion bit constellations is the first time for M3 formulated request always fulfilled if the cells of a cell pair are überein¬ other. The coding consists simply in the application of the XOR operations to the inversion cells according to the given configuration, partly infor- mation-wise, whereby the order in which the partial information is assigned to the quantities is basically indifferent.

The minimum distance of this code example is obvious, since a) at least 5 bits change with each information bit b) at crossovers of inversion cell quantities (resetting of the relevant bit) at least two information bits are involved, which themselves invert several further cells.

By adding a parity bit to the information bits, the minimum distance can be increased.

The example of Figs. 1 to 10 corresponds to claim 2, i. that all inversion cells listen to a partial information of at least one cell tuple of this partial information.

An arrangement according to claim 1 is shown in FIG. 11, similar to FIG. 6, where, however, compared to FIG. 6, the quantity M6 is reduced, but the remaining inversion cells "i" are retained, marked "I" in FIG , decoding:

The description is again based on the example Fig. 1 bislO.

The decoding is based on the comparison of the contents of inversion cells and non-inversion cells within cell tuples.

The cell tuple sets and thus the partial information are processed in the opposite order to coding, here MIO, M9, ... Ml. Since in the example we have binary-coded information symbols and binary coded cells, and since the cell tuples here are cell pairs with one inversion cell and one non-inversion cell, the procedure here is to describe as follows: If the cells of a cell pair considered are unequal, that speaks for itself that the associated current information bit is set; if they are the same, this indicates that the corresponding current information bit is not set.

The decision is preferably made by a majority decision of the participating cell pairs of the cell tuple of the step configuration.

If further decoding steps are pending, each time a decision has been made that the information bit is set, the inversion cells are inverted to the current step (as a matter of course, corresponding flags can be set). In this way, the information bits 10, 9,... 2 are successively decoded.

The information bit 1 again represents a special case: Here, the decision is not made via a comparison of different cells, but via a direct (preferably majority decision) comparison with the base value set in the coding, ie 0 or 1.

If the cells have each been directly inverted (work without corresponding markers, see above), a matrix completely occupied by the basic value remains at the end before decoding undisturbed cells. In the case of originally faulty cells, an error image (or 1D error signal) is produced at the end with correct decoding. This error image can be used to assess the quality of the decoding or the signal.

Apart from information bit 1, in this example, therefore, all comparisons are made directly between current cell contents, without reference to any reference dimensions. If, for example, all cells were the invert (for phase modulation, for example, by an unwanted global phase shift of 180 degrees), nevertheless all bits 2 ... 10 would be correctly decoded.

15 shows such a majority decision in a system with amplitude modulation, with an encoding of 2 infobits and 6 code bits; For each of the two coding steps, there are quantities (M1 or M2) with three cell tuples, each with one inversion cell and one non-inversion cell.

At the output of the demodulator are here binary reception values, which are compared via logi¬ XOR. Due to a majority decision on these XOR results, the infobits are estimated (maj = 1: "1" has majority).

In the following generalizations are described.

The first generalization is the transition from conventional logic to a fuzzy approach.

So far we have assumed a binary representation of the cells.

In practice, disturbances occur which the demodulator can numerically evaluate and which it is to evaluate for a reliability-optimizing decoding.

We first assume that a demodulator assigns a (fuzzy) value between 1 and 9 to each cell, where 1 means "sure black (horizontal ...)", 9 means "sure white (vertical ...) "5 means" unknown ", 7 means" rather white (vertical) ", 8 means" almost certainly white (vertical) ", etc. An odd number of evaluation levels is preferable, as you then have (as with the value 5) the possibility of a neutral rating.

Single numerical values were chosen because of the better view, the value 1 was chosen as the lowest value, since with an odd number of possible values one lies exactly in the middle and the 5 as average value is particularly illustrative.

According to the invention, the XOR operation is replaced by the following operation, with the numbers 1 and 9 to be taken as examples:

XOR (A ₅ B) = ABS (AB) + 1; (ABS = absolute amount)

For example, XOR (3,3) = 1 ('equal'), XOR (1,9) = 9 ('not equal'), XOR (3,7) = 5 ('indefinite').

The inversion is carried out by:

INV (A) = (9-A) +1.

For the sake of completeness, we still specify the AND and OR operations:

AND (A ₅ B) = Min (A, B), OR (A, B) = Max (A, B).

In this case, "+ 1" is defined for neutral evaluation because of the particular clarity of FIG.

The equality request on decoding is replaced by an XOR operation on the respective values and the query whether the result of this XOR operation is> 5 or <5. Advantageously and according to the invention, the equality query is not performed directly on the cells of a cell tuple, but instead the values of several identical cells (inversion cells or non-inversion cells) are added conventionally and thereafter, with appropriate normalization, the number of cells involved, the> 5 / <5 query performed. In between, it may still be advisable to set suitable non-linear characteristics to suppress single small errors. Advantageous as majority decision is also the application of the median operator over the results of several cell pairs, with subsequent> 5 / <5 comparison.

Thus, for each information bit, a majority decision on several "individual opinions" advantageously takes place, with insecure cells automatically holding back from the decision process due to their value "proximity 5" (the closer they are to the 5), quite the opposite to algebraic ones Decoding methods, in which even values close to the limit can inevitably lead to individual wrong decisions, which then have to be made costly or made good or lead to rejections.

If the value that leads to the decision is 5 or too close to the value 5, then either a rejection is carried out or the decoding process is continued for both alternatives and the final decision is made dependent on additional criteria, as determined by soft- Decision systems is known be¬.

Fig. 16 shows such a majority decision with fuzzy approach. When coding a 3 x 5 matrix code, the cells are signaled with "up bar" or "down bar". The demodulator calculates received values between 1 and 9 for the received cells, with the meaning: 1: safe downwards 2: almost safe downwards ... 5: insecure ... 8: almost safe upwards 9: safely upwards , The decoder works only on the basis of fuzzy XOR comparisons of such reception values. In order to bring about a majority decision with evaluation of the collateral, the obtained XOR values are simply averaged and the result of the mediation m compared with "5".

In the example of FIG. 16, the first decision is "equal" (ie, infobit = 0), so there is no need to invert the inverse cells of the first step for the second step, otherwise fuzzy inversion would be required for the cells concerned realize, formula see above.

FIG. 17 shows another example with a majority decision based on the fuzzy approach. When transmitting a binary code as a one-dimensional signal in 0/180-degree phase modulation, the demodulator calculates reception values between 1 and 9 for the received cells, with the meaning: 1: 0 degree 2: almost certainly 0 degrees ... 5: uncertain ... 8: almost certainly 180 degrees 9: sure 180 degrees.

The decoder works only on the basis of fuzzy XOR comparisons of such reception values. In order to bring about a majority decision with evaluation of the collateral, the obtained XOR values are simply averaged and the concatenation result m compared with "5".

In the example of FIG. 17, the first decision is "unequal" (ie, infobit = 1), therefore inversion of the inversion cells of the first decoding step must take place for the second step, as indicated by the circle symbols (9-> l, 5-> 5).

The second generalization is the integration of demodulation by direct comparison at the signal level.

So far we have made a comparison between every two cells via an "analog" XOR function. However, the comparison can preferably and according to the invention be realized directly on the signal plane (pixel plane), in the case of time signals, for example by correlation, in the case of images, for example by comparison of the relevant image sections. This has obvious advantages in particular if the properties of the cell contents (locally or temporally) change at low frequency.

For example, if one has to distinguish cells with a small cross from those with a small ring, then the question is, "Are cells A and B rather the same or rather unequal?" Technically much more reliable answer than the calculation of good individual evaluation measures ("similar cross?" ...), because the latter requires the availability of a reference cross and a reference ring: imagine that the diameters of the circles due to any disturbing influences itself can change slowly; then once a given reference circle does not help anymore.

It is therefore not worked by comparison with given references, son¬ countries by comparison between different signal parts; Disturbances which are common to these different signal parts (this is almost always the case with low-frequency or locally adjacent signal parts) are automatically calculated. In image processing applications, signal comparisons can be realized, for example, by normalized correlation.

So no reference patterns for cell demodulation are needed; The system automatically carries out self-referencing in the course of the decoding.

Fig. 18 shows a procedure with direct signal comparison. In the coding of a 3 × 5 matrix code, as in FIG. 16, the cells are signaled with "up-bars" or "down-bars". Demodulation and decoding take place in one based on a comparison of cell contents. Here, a majority decision on such comparison results. As an illustration, we shall again work with values 1 ... 9, here with the meaning:

1: cell contents certainly equal to 2: cell contents almost certainly the same ...

5: Comparison not possible (not sensible, because there is too little signal energy) ... 8: Cell contents almost certainly not equal to 9: Cell contents certainly not equal

The majority decision can in turn be made by averaging such results and subsequent comparison of the mean m with the value "5".

If, in a decoding step, the decision was made "unequal" and subsequent decoding steps follow, then the inversion cells of this step are to be inverted. Since the cell contents, that is, the signals, can not always be inverted (this only works in special cases, for example in the case of binary amplitude modulation), it is proposed here to substitute an auxiliary cell field with binary values assigned to the cells. If now an inversion cell is to be inverted, technically the respective binary auxiliary cell is inverted. In cell comparisons in the following steps, it should now be noted how many of the cells to be compared have auxiliary cells with log. "1" zuge¬ are assigned. If the number is odd, the result of the comparison is inverted. Of course, the auxiliary field is initialized with log. Filled "0".

In the example in FIG. 18, the first decision is "equal" (thus estimating the infobit = 0), therefore there is no need to invert the inversion cells of the first step for the second step, otherwise one would set corresponding logical flags corresponding to the inversion cells the first decoding step zu¬ are ordered. Fig. 19 shows another example with direct signal comparison. As in Fig. 17, it is the transmission of a binary code as a one-dimensional signal in O / 180-degree phase modulation. Demodulation and decoding take place in an on the basis of a comparison of the cell contents. Here a majority decision on such comparison results is realized. As an illustration, again work with values 1 ... 9, here with the meaning:

1: cell contents certainly equal to 2: cell contents almost certainly the same ... 5: comparison not possible (not sensible, because there is too little signal energy) ... 8: cell contents almost certainly not equal 9: cell contents certainly not equal

The majority decision can again be made by averaging such results and then comparing the mean with the value "5".

In the example of FIG. 19, the first decision is "unequal" (ie, estimation of the infobit = 1), therefore an inversion of the inversion cells of the first step must take place for the second step. The use of log. Marking, as described above for FIG. 18, which are "1" (for "inverting") for the next step, is symbolized here by the small circle symbols. In the second decoding step, the first comparison (output value "5") has no markers at "1", at the second comparison (output values "3") both affected cells are at "1", ie an even number. The result is NOT to invert.

A third generalization is to move from a logic to base 2 to a multi-valued logic to base 3, etc. The transition from two-valued to trivalent etc. logic can be motivated by unusual Datenforma¬ te, eg coding of 65. ..81 possible values (81 = 3 * 3 * 3 * 3), by the need to optimally use existing transmission channels, or by several feature dimensions of the realization of cells. The case-by-case inversions (XOR operations) are to be replaced for logic B according to the invention by modulo B additions (in coding) and - subtractions (during decoding) by a multiple of a fixed increment D. The increment D is at Fuzzy procedure greater than 1, preferably straight, otherwise it is 1. If the value of the partial information is j, the cell values for the inversion cells for this partial information are coded by D * j, modulo B * D, elevated.

During decoding, as in the past, cell values are compared; for comparison, the cell values of the relevant non-inversion cells are subtracted from those of the inversion cells modulo B * D. The result can be averaged over several subtractions as before (take the result modulo B * D). The result is divided by D and rounded off and then output as searched partial information t.

If there are further decoding steps, the inversion cells of the partial information are inverted as before, in which case the inversion is done by subtracting t * D, modulo B * D.

So far, decoding was based on comparing cells with the query "more or less equal?". Now, the comparison with the query "how much do the cell values differ?" In a non-fuzzy procedure (D = 1), the difference of the cell values corresponds to the decoding result of this step, which can be averaged as before. With fuzzy procedure (D> 1) one counts in D-stage quantization, which - if necessary after averaging - by the division by D is canceled again.

Polyvalent fuzzy realization is explained using an example with trivalent logic (value store of the partial information is 0, 1, 2), with B = 3, D = 4: Wer¬ temenge is 0, 1, 2, 3, 4, 5, 6 , 7, 8, 9, 10, 11 with the safe values 0, 4, 8 and The uncertain intermediate values 2, 6, 10. The fuzzy comparison of an inversion cell with value 10 and a noninversion cell with value 1 leads to the difference 9, divided by 4 and rounded gives 2. Thus, 2 is the sought Teilinforma¬ tion. If further decoding steps are pending, the affected inversion cells are inverted, ie 2 * 4 is subtracted, modulo 12.

This third generalization is particularly advantageous in polyvalent phase modulation.

Fig. 12 shows a hardware realization for a simple example. Fig. 12 is a schematic showing details such as e.g. Control signals are omitted. The coding is shown above, the decoding below.

In the example of Fig. 12, we have the following configuration, see Fig. 13 (cells zl ... z4):

Partial information 10 is a special case as in FIG. 1 and consists only of inversion cells.

Partial information Il has the cell tuples Al with inversion cell zl and noninversion cell z3, and cell tuple A2 with inversion cell z2 and noninversion cell z4. Partial information 12 has the cell tuples Bl with inversion cell zl and noninversion cell z2, and cell tuber B2 with inversion cell z3 and noninversion cell z4.

Coding (Example Fig. 12):

Again, we start with binary cells. It is a signal with 3 pieces of information to encode each one bit, 10, II, 12. A total of 4 bits are available for encoding. The coding path is a pipeline consisting of three registers 11 each having 4 binary cells 10. The cells are numbered 1 to 4 for each register from left to right for the following description, as well as from left to right; these numbers are not shown in FIG. 11 for the sake of clarity.

From coding step to coding step, the cells are transferred, from left to right, from register to register (line parallel with 4 shift registers of length 3 or line serial with a shift register of length 12). In the first step, the information bit IO is transferred to the first register (formally this can be considered as an XOR combination with a basic value of logic "0".) If these XOR operations are actually implemented, the basic logic value can just as well "1" are used).

In the second coding step, the information bit II is XOR-linked to the content of the first two cells of the second register, the result is written back to the cells in question. In the third coding step, the information bit 12 is XOR-linked to the contents of the first and third cell of the third register, the result is written back into the relevant cells.

Parallel to the processing of Il in the second register, the processing of the next 10 value in the first register can be done, etc. (pipeline).

The generated code is as follows (cl..c4 corresponds to the cell numbers 1 ... 4 of the registers):

12 Il IO cl c2 c3 c4

0 0 0 0 0 0 0

0 0 1 1 1 1 1

0 1 0 1 1 0 0

0 1 1 0 0 1 1

1 0 0 1 0 1 0

1 0 1 0 1 0 1

1 1 0 0 1 1 0

1 1 1 1 0 0 1

It is clear that "1" is only an even number, so the minimum distance is 2.

Decoding (Example Fig. 12):

For decoding, registers are processed in reverse order. In the first decoding step, the information bit 12 is obtained by XORing the contents of the first and second cells of the third register and XORing the contents of the third and fourth cells of the third register. If the comparison results are both 1, 12 = 1 is decoded, if they are both 0, 12 = 0 is decoded.

Since there are two comparison results in this deliberately simple example, a majority decision as in FIG. 15 is not possible. Advantageously, fuzzy cell values are used, and for this decision both comparison results are considered together by simple summation of the fuzzy XOR values, followed by a threshold value decision, as described for the examples FIG. 16-19, possibly supplemented by a non-linear one Curve. Such summation and threshold elements are indicated at 14 in FIG. Depending on 12, the first and the third cell of the third register are inver¬ by 12 with the relevant cell values (fuzzy) XOR- is linked and the result is written back to the cells.

Accordingly, proceed with the other registers:

In the second decoding step, the contents of the first and third cells of the second register (fuzzy) are XOR-linked, as is the content of the second and fourth cells of the second register, and II is preferably obtained by summing the fuzzy XOR values followed by a threshold decision. Dependent on Il, the first and the second cell of the second register are inverted by XORing Il with the respective cell values (fuzzy) and writing the result back into the cells.

In the third decoding step, the information bit IO is linked to the basic value by (fuzzy) XOR linking of the contents of the four cells. 10 is preferably obtained by summing the fuzzy XOR values followed by a threshold decision.

If an error signal ("error picture", see above) is to be obtained, the four cells of the first register are inverted here, depending on 10, by 10 being linked to the respective cell values (fuzzy) XOR and the result written back into the cells. This return branch is not necessary in the last step for the decoding itself.

Partial information 10 is a special case as in FIG. 1 and consists only of inversion cells. Here, according to FIG. 11, the decision is not made on the basis of a cell comparison. This can be remedied in a simple manner by a differential coding of (preferably consecutive) 10 values: in coding, the I0 values are replaced by XORing successive 10 values; in decoding, accordingly, before the summation tion and threshold element 14 to 10, a (fuzzy) XOR comparison built with cached from the last IO decision cells.

The simple example shown here has a code rate of 3/4.

The decoding can also be carried out in a pipelined manner.

The error image can be evaluated cell by cell as a deviation from the logical basic value "0" or "1" (see above).

Placement:

Geometrically, the place where cells are ultimately in the signal does not matter; it depends only on the partial information-specific division into in- version and non-inversion cells and the assignment to cell tuples. Thus, we could exchange locally any two cells in the figures shown in FIGS. 1 to 10, as long as the meaning "inversion cell" - "non-inversion cell" in each of FIGS. 1.10 is reversed and also the affiliation with the cell tuples ,

Not even the geometric 10 x 3 arrangement of the cells needs to be maintained. It is quite possible to place the cells in an other format at the signal output, in particular one-dimensionally. The placement of cells shown in FIGS. 1 to 10 can therefore be used merely as an auxiliary arrangement, that is to say as a purely conceptual arrangement, for a better illustration of the configuration. The final placement, ie the temporal order or local arrangement of the cell contents in the channel, may be identical to the auxiliary arrangement or even deviate therefrom. The auxiliary arrangement can certainly also be realized with more than two dimensions, even for one-dimensional signals. The final placement of the cells is now preferably carried out according to the invention in scattered placement.

In a scattered placement, cell tuples may be scattered to a partial information.

Technically, we speak here of a scattering placement in an encoding step if at least three sets of cell tuples can be formed for this coding step in the final placement, called placement quantities, which are spatially or temporally separated from one another by at least one cell in the case of time signals ,

The aim of the scattered placement is the same as in the known interleaving, by which not only single errors but also burst errors can be corrected.

When decoding, majority decisions are preferably made for a plurality of cell tuples for each decoding step. If these cell tuples are widely separated, a bundle error affects only one of the cell tuples; With at least three widely spaced amounts of cell tuples, the placement quantities, the right decisions can be made by majority decision on placement quantities.

Of course, the placement pattern must be known to the encoder as well as the decoder.

In the sense of local cell comparisons, however, the inversion and non-inversion cells within a cell tuple should, on the other hand, be close to one another. Cell comparisons to be performed locally are always advantageous if signal disturbances have a similar or similar effect on neighboring cells. This is the case, for example, with pictures where the light conditions within the picture can hardly be local, but very different globally; A good example is needle-embossed matrix codes, where the imprints in the image may appear in very different forms due to different illumination and viewing angles, but hardly differ from each other in closely related imprints due to locally similar illumination conditions. In the case of one-dimensional signals, there are such phenomena: With amplitude modulation, the reception level is generally subject to low-frequency fluctuations; in phase modulation, the fundamental phase can change, but it usually remains approximately constant over relatively long periods of time.

Fig. 20 shows an example of scattered placement for an example similar to Figs. 2 to 10: coding steps 2 and 3 correspond to the step configuration of Figs. 2 and 3, respectively; coding step i corresponds to the step configuration of Fig. 7; However, the cell field is larger with 14 x 3 cells, see Fig. 20 above. For the scattered placement, the second line (possibly cyclically) is shifted to the right by 5 cells, the third line is shifted to the right by 10 cells, see FIG. 20 below. The assignments to the cell tuples are also shifted. In coding steps 2 and 3, one thus obtains three quantities 31 of cell tuplets, which fulfill the condition of placement quantities, ie are separated from one another spatially by at least one cell.

FIG. 21 shows a one-dimensional final cell arrangement for coding steps 1 to 8 with three or four cell tuples per step. In step 1 to 5 there are three placement quantities each consisting of one cell tuple, each with one inversion cell and one noninversion cell, which are separated from each other by at least two cells each.

Note: Not all inversion cells must belong to cell tuples, as already described above with reference to FIG. 11. In the example of FIG. 21, too, such inversion cells exist. Those that belong to cell tuples are marked with "i" indicates that those that do not belong to cell tuples are marked with "I". The latter cells play no role in the consideration of the placement quantities.

The previous considerations have disregarded a scattered placement to simplify the description.

An extension according to the invention is preferably to link the data with a grid XOR-, after completion of the encoding, e.g. with a chessboard or in one-dimensional with a 010101 .. sequence, or even a random sequence. This is called in the following grid operation. As a result, a certain structuring of the data can be achieved, such that, independently of the coded information, (locally) a certain minimum number of cells is always assigned a logical 1 and a minimum number is logically 0 (in the example cited above, too after scattered placement - achieved by using a simple checkerboard grid, that in the 3 x 10 field always at least 7 logical 1 cells and at least 7 logical 0 cells befin¬ the). The code gets thereby a certain pattern. This property can advantageously be used for the demodulation, in particular also for the synchronization, with a retroactive effect from the decoding to the demodulation. The most primitive example would be to go through a binarization threshold until the minimum numbers are met. Also for a rejection criterion, this property is useful; primitive example: a completely black image is automatically rejected without additional special queries. Before the actual decoding, a grid operation is performed to undo it.

Alternatively or in addition, the coding result can also be supplemented by a fixed synchronization or referencing pattern, for example a sequence of alternating symbols such as 0101010 ... 012012012012 ... etc.

An important issue is the optimal design of the configuration. Two systematic approaches to the configuration, both of which can be recognized starting from the example of FIGS. 1 to 10, can be combined with one another:

1) Disassemble successively into small regions of different (X) coordinates, which are associated with inversion and non-inversion cells, as can be seen from FIGS. 2 to 6.

2) From our example Figures 1 to 10 it can be seen that a dimensional organization of the cells may be helpful: the quantities M1 to M6 are oriented vertically, i. Vertically standing cells are all similar, while the quantities M7 to M10 are oriented horizontally. Also auxiliary arrangements with more than 2 dimensions are possible in principle.

Multi-dimensional auxiliary arrangements are explained below with reference to the example according to FIGS. 22 to 31 with a three-dimensional auxiliary arrangement:

To encode is originally an 8-bit information. In a preprocessing this is supplemented by a parity bit, so that in the following we assume that a 9-bit information is to be encoded bit by bit in 9 encoding steps S1..S9. The coding is binary, for a total of 64 cells with binary cell values are available.

These 64 cells are now alternatively arranged in a three-dimensional 4x4x4 arrangement, cubes for short.

FIGS. 22 to 30 show step configurations for steps S1 to S9. The three dimensions are referred to with reference to the figures with left / right, front / rear and top / bottom. In the figures, the inversion cells of the respective step are shaded. In steps S1, S2, S3 the division into inversion cells and non-inversion cells takes place left / right: in S1 (FIG. 22), in S2 (FIG. 23), and S3 (FIG. 24), one each arises from left to right 4x4x1 disc of inversion cells. The cell tubers 1 each consist of an inversion cell and a (not necessarily exactly one) non-inversion cell farther to the right (not necessarily directly adjoined, as indicated by the hatched arrows in FIGS. 22 and 23).

In steps S4, S5, S6, the division into inversion cells and non-inversion cells takes place front / rear: in S4 (FIG. 25), in S5 (FIG. 26), and S6 (FIG. 27), one each arises from front to back 4x4xl slice of inversion cells. The cell tubers 1 each consist of an inversion cell and a (not necessarily exactly one) non-inversion cell further back (not necessarily directly adjacent).

In steps S7, S8, S9, the division into inversion cells and non-inversion cells takes place up / down: in S7 (FIG. 28), in S8 (FIG. 29), and S9 (FIG. 30), one each arises from top to bottom 4x4xl slice of inversion cells. The cell tubers 1 each consist of an inversion cell and a (not necessarily exactly one) non-inversion cell further down (not necessarily directly adjacent).

Of course, mutatis mutandis permutations and other changes may be made, e.g. In FIG. 30, the inversion cells and noninversion cells of the cell tuples 1 can be reversed.

Minimum Distance: With each set bit of the 9-bit information, the corresponding inversion cells change at the associated step; these are arranged in the auxiliary arrangement as a 4 × 4 × 1 plate, see hatched cells in FIGS. 22 through 30, wherein the discs can be differentiated into disc types by dimension, ie in left / right discs (S1 ), front / rear discs (S4 ... S6) and top / bottom discs (S7 ... S9). Since the source information um a parity bit has been added, two arbitrary words of the 9-bit information differ by at least two bits, the finished coded cells to two such words thus differ by at least two slices. If these two disks belong to the same type, the distance is 2 x 4 x 4 = 32, ie half of the cube. The situation that two disks belong to different types is shown in FIG. 31. The cells which belong to both disks are inverted twice in this case, thus contributing nothing to the differentiation. Such cells are marked with 21 in FIG. 31 and drawn in a checkered manner. It remains 3 x 4 + 3 x 4, so 24 inversion cells, in which the two words concerned differ. The minimum distance is thus 24.

Although the code rate of the system is relatively low at 8/64 = 1/8, this is compensated for by the relatively high minimum distance and, in particular, by the general advantages of this invention: decoding solely by comparisons of cells taken from the cells , without reference cells or reference values, majority decisions over many comparisons, soft decision, possibility of correction of more cells than corresponds to half the Mindest¬ distance, direct evaluation of the signal without separate demodulation.

Of course, the invention may also be used as a chained coding in conjunction with a known coding, e.g. as inner coding, in combination with an algebraic coding (for example Reed Solomon) as outer coding and interleaving / deinterleaving between the coding types.

The invention has the following advantages:

The mode of operation of the method is transparent and vivid with its individual steps, in the system design very specific application-specific adjustments can be realized, eg modifications of the comparison functions, installation of characteristics, placement of cells, definition of the configuration, so that optimally adapted to the given application leaving optimizing decoding can be realized, with considerations directly on Signal¬ level in the course of decoding. Demodulation and decoding are coupled and cooperative.

The information bits can be assigned a specific degree of decoding security by means of suitable configuration (for example, in the coding of numerical measurement data, data is important, where errors in the significant bits have a strong impact, in the least significant bits are tolerable, among others). This can be achieved, in particular, by the following measure: the more significant partial information is coded AFTER the less significant and thus decoded BEFOR to the less significant partial information, namely with more inversion cells than in the less significant partial information. This means that the more significant partial information is decoded more reliably by majority vote. It also makes sense to assign steps to the significant partial information in which the cell tuples are as far apart as possible when scattered and the cells within the tuple are as close together as possible. In this regard, the signifiers would have partial information, e.g. in Fig. 21, the above-drawn step configurations cheaper than those further down.

As a simplest example of reliability-optimizing decoding, the example of FIG. 12 serves: the minimum distance of the code is 2, so with algebraic decoding methods only error detections, but no error corrections are possible. Even the described procedure with fuzzy logic (without signal comparisons over several pixels within the cells) permits error corrections, in some cases even for several cells, by automatically retaining in the threshold elements uncertain cells from the decision process and be "overruled" by secure cells.

Decisions are made in decoding solely on the basis of comparisons; it is therefore always relative considerations within the signal to be interpreted; the system references itself based on the current rush signal, even a signal inversion pays off automatically. The latter is particularly important in the case of transmission paths with phase modulation and codes to be read optically, in which the cells can undergo a contrast reversal, depending on the illumination and viewing direction.

A further increase of the decoding quality results from the fact that the comparison is realized directly on the signal level (pixel level). This has obvious advantages, in particular, when the properties of the cell contents (locally or temporally) change at low frequency.

Furthermore, the invention encompasses a method for coding information that can be represented by partial information, wherein cell values are assigned to the cells by the coding, characterized by a) partial information coding steps, b) partial information specific cell tuple sets whose cell tuples consist of two types of cells, inversion cells and noninversion cells , wherein each cell tuple consists of at least one inversion cell and at least one non-inversion cell, wherein the cells within each cell tuple for each preceding coding step either all

Inversion cells or all non-inversion cells are c) for at least one coding step:

XOR Linking of the relevant partial information with the inversion cells of the relevant cell tuple quantity.

Furthermore, the invention comprises the following embodiments:

The cell tuples can be cell pairs with an inversion cell and a non-inversion cell. Signal values, in particular also pixel values of images, can be assigned to the cell values.

The pixels can assume one of the binary values 0 or 1.

A cell may be associated with a plurality of pixels, wherein the pixels are the pixels of an image or a sequence of images or the signal values of other signals, with assignment of pixel patterns to the cell values.

For coding, a differential coding of the partial information processed first can be carried out in advance.

The result can be XORed with a pattern, for images with a two-dimensional checkerboard pattern, for one-dimensional signals with a 101010 ... pattern.

The result can also be XORed with a pseudo-random pattern.

The invention also comprises an encoding device for a code, which can arise according to one of the aforementioned methods according to the invention, wherein registers are used with cells which can be inverted in dependence on information bits, preferably via XOR gates.

Furthermore, the invention comprises a code, which can be characterized by such a coding method, when interpreted as being produced according to one of said coding methods, by a placement of the cells in which 2 different cell tuples of a cell tuple quantity are placed so far apart from each other they are separated by at least one other cell, which does not belong to these cell tuples. Furthermore, the invention comprises a code which can be formed after such a coding method, if it is interpreted as being produced according to such a coding method, by a placement of the cells in which pairs of inversion and non-inversion cells of a cell tuple adjoin at most one intermediate cell are, preferably directly adjacent.

Also included is in each case also a transmitting device which operates with such a coding method, characterized by a phase modulator for at least one feature dimension, as well as a storage medium on which data are stored working with one of the aforementioned codes, and a transmission protocol Code with a code, which can arise according to one of the aforementioned methods, characterized by the addition of a fixed synchronization or referencing pattern, preferably in the form of a 010101 series of at least 6 cells.

A further embodiment of the invention comprises a marking device with deforming marking, in particular needle embossing, with an aforementioned coding method for the marking, characterized in that the deformation strength is variable for signal modulation, preferably via the stop strength or the stroke of the needle.

A further embodiment of the invention comprises a marking device with deforming marking, in particular scratch embossing, with such a coding method for the marking, characterized in that the direction of bar elements, which can also be connected to one another, can be changed for signal modulation.

A further embodiment of the invention comprises a labeling device with such a coding method for the lettering, characterized in that the width of lines or the size of dots or the direction of line elements is variable for signal modulation. A further embodiment of the invention comprises a code carrier for Identifi¬ cation purposes, eg RFID, with an aforementioned code.

A further embodiment of the invention comprises a decoding method for a code as it can arise from an aforementioned coding method, wherein the partial information is obtained from cells to be decoded, wherein a cell consists of one or more pixels, the pixels representing the image are points of a picture or a picture sequence or the signal values of a signal, characterized in that, when interpreted as a coding method according to any one of claims 1 to 8, the partial information is obtained from the cells to be decoded by zelltupelweisen Ver equal to Contents of inversion cells with the content of non-inversion cells, further characterized by comparing result-dependent determining the relevant sub-information, comparing result-dependent inversion of the respective inversion cells, if further decoding followed.

In this case, the comparison result can be determined by a majority decision via cell comparisons, preferably via the comparison of summation values via inversion cells on the one hand with summation values via non-inversion cells on the other hand, or via the median of individual comparison masses.

In this case, the comparison can be realized by a comparison function on a plurality of pixels of the respective cells, preferably a normalized correlation value.

Another embodiment of the invention is characterized by evaluating the resulting defect image for a rejection criterion and / or for a quality assessment. The invention also comprises a decoding device for an aforementioned code, characterized by summation and threshold value elements with which it is possible to decide on partial information to be decoded.

The invention also comprises a decoding device for an aforementioned code, characterized by registers with cells which can be inverted in dependence on already obtained information bits, preferably via fuzzy XOR gates.

The invention further comprises a decoding device for an aforementioned code in which the first coded partial information (10) is coded differentially, characterized by a circuit with which decoded partial information (10) for the reversal of the differential coding can be temporarily stored.

The invention also includes a receiving device with an aforementioned decoding device.

Finally, the invention also encompasses a code reading device for a previously mentioned code, characterized by an image recording device, in which at least one image recording parameter, in particular also lighting parameters, can be influenced via a quality measure which depends on an error image.

Claims

Claims

1. A method for coding information that can be represented by partial information, wherein cells are to be assigned cell values by the coding,

marked by

a) partial information coding steps, b) partial information-specific classification of the cells in inversion cells and non-inversion cells, wherein each partial information is assigned at least one inversion cell c) at least one partial information-specific cell tuple, each cell tup consists of at least one inversion cell and at least one non-inversion cell, wherein the Cells within each cell tuple for each preceding coding step are either only inversion cells or only non-inversion cells. d) for at least one coding step:

XOR linking of the relevant partial information with the inversion cells assigned to this partial information from b).

2. Method for coding according to claim 1, characterized in that in the at least one coding step all inversion cells of the coding step belong to at least one cell tuple.

The method of coding of claim 1 or 2, wherein the cell tuples are cell pairs having an inversion cell and a non-inversion cell.

4. Method for coding according to one of the preceding claims, wherein signal values, in particular pixel values of images, are assigned to the cell values.

The method of coding of claim 4, wherein each of the pixels can take on a value of binary values.

6. The coding method according to claim 1, wherein a plurality of pixels are assigned to a cell, wherein the pixels are the pixels of an image or an image sequence or the signal values of another signal are characterized by the assignment of pixel patterns to the cell values.

7. A method of coding according to one of the preceding claims, gekenn¬ characterized by a previously made differential coding of the first be¬ worked part information.

8. A method of encoding according to any one of the preceding claims, characterized by XORing the result to a predetermined pattern, wherein the predetermined pattern in images is a two-dimensional checkerboard pattern and in one-dimensional signals is a 101010 ... pattern.

9. The method of coding according to one of claims 1 to 7, gekennzeich¬ net by XOR operation of the result with a random pattern.

10. A method of coding according to one of the preceding claims, gekenn¬ characterized by: for at least one step, a placement of the cells, in which at least 3 sets of cell tuples are formed, which are separated by at least one cell.

11. Coding device for information that can be represented by partial information, wherein cell values are to be assigned to the cells by the coding, characterized by registers (11) with cells (10), coding step by step depending on Partial information per register at least one cell is invertible, preferably via XOR gate (13), wherein at least one invertible cell as an inversion cell of one or more cell tuples with inversion cells and non-inversion cells is interpretable, wherein the cells within each cell tuple for each preceding coding step either only inversion cells or only non-inversion cells.

12. Encoder according to claim 11, characterized in that the registers are arranged in a pipeline.

13. Transmitting device which operates with a coding method according to one of Ansprü¬ che 1 to 10, characterized by a phase modulator for at least one feature dimension.

14. Code generated by a coding method according to one of claims 1 to 10.

15. Storage medium on which data is stored with a code which has arisen by a coding method according to one of claims 1 to 10.

16. Transmission protocol with a code, which has arisen by a coding method according to one of claims 1 to 10.

17. Transmission protocol according to claim 16, wherein after coding a fixed synchronization or referencing pattern is added, preferably in the form of a 010101 sequence of at least 6 cells.

18. Marking device with deforming marking, in particular needle embossing, with a coding method for marking according to claims 1 to 10, characterized in that, for signal modulation, the deformation mungsstärke, is variable, preferably on the stop strength or the stroke of the needle.

19. Marking device with deforming marking, in particular scratch embossing, with a coding method for marking according to claim 1 to 10, characterized in that for the signal modulation, the direction of bar elements, which may be interconnected, is variable.

20. Labeling device with a coding method for the labeling according to claim 1 to 10, characterized in that for signal modulation the

Width of strokes or the size of dots or the color of dots or the size and color of dots or the direction of stroke elements is changeable.

21. Code carrier for identification purposes with a code which has arisen by a coding method according to one of claims 1 to 10.

22. A method for decoding a code, such as may arise from a coding method according to one of claims 1 to 10, wherein the information to be decoded can be represented by partial information and wherein the partial information is obtained from cells to be decoded,

when interpreted as being produced by a coding method according to any one of claims 1 to 10,

marked

A) by partial information decoding steps,

B) in that, for at least one decoding step, the partial information from cells to be decoded is obtained by cell-cell-wise comparison of inversion cells with non-inversion cells and comparison-result-dependent determination of the relevant sub-information,

C) by comparing result-dependent linking of an inversion information with the inversion cells of the decoding step in question, provided that at least one decoding step according to B) follows.

23. A decoding method according to claim 22, characterized in that the comparison of inversion cells and non-inversion cells is carried out by comparison of cell values derived from the cells.

24. A decoding method according to claim 22, characterized in that the comparison of inversion cells and non-inversion cells is carried out by comparing the contents of the cells in question.

25. A decoding method according to claim 23, characterized in that a comparison result is determined by evaluating a plurality of comparisons of cell values according to a quality function, preferably - over the median of individual comparison measures.

- by the arithmetic mean of individual comparison measures.

26. A decoding method according to claim 23, characterized in that a comparison result is determined by evaluating cell values according to a quality function, via the comparison of mean values via inversion cells on the one hand with mean values over non-inversion cells on the other hand.

27. A decoding method according to claim 24, characterized in that a comparison result is determined by evaluating a plurality of comparisons, preferably over the median of comparison functions on several pixels of the respective cells or

- About the arithmetic mean of comparison functions on each several pixels of the cells in question.

28. A decoding method according to claim 22, characterized in that a link to an inversion information is also carried out in the last decoding step, that the cell values after the last decoding step are interpreted as an error image or error signal for a quality assessment ,

29. A decoding method according to any one of claims 22 to 28, with a multi-valued logic in the coded cell values, in particular with correspondingly multi-valued phase modulation on the signal level, characterized by multivalued fuzzy logic in the cell values to be decoded.

30. decoding device for information, which can be represented by partial information, from a code in which cells are assigned cell values in each case,

marked by

a number of registers (11) each comprising register cells (10), of which at least one register (11) serves to receive the information to be decoded, and wherein each register (11) at least one register cell decoding step by step in dependence on the decoded Partial information is invertible, preferably via XOR gates (13), wherein at least one invertible register cell as an inversion cell of one or more cell tuples with inversion cells and non-inversion cells is interpretable, wherein the cells within each cell tuple are either only inversion cells or only non-inversion cells for each subsequent decoding step,

wherein the partial information can be obtained from register cells by cell-cell-wise comparison of inversion cells with non-inversion cells and comparison-result-dependent determination of the relevant partial information.

31. Decoding device according to claim 30, characterized by summation and threshold value elements (14) with which it is possible to decide on partial information to be decoded.

32. The decoding device according to claim 30 or 31, wherein at least one comparison is realized by a fuzzy XOR gate (13).

33. Decoding device with a decoding method according to one of claims An¬ claims 22 to 28, for a code in which the first encoded part information (10) are differentially encoded, characterized by a circuit with the decoded part information (10) for the reversal of the differential encoding can be stored zwi¬.

34. Receiving device with a decoding device according to one of claims An¬ claims 30 to 33rd

35. Code reading device for a code, as may arise according to one of the methods 1 to 10, characterized by an image recording device, in which at least one image recording parameter, in particular Beleuchtungsparame¬ ter, can be influenced via a quality measure that is dependent on an error image.