Image encoding that supports block partitioning and block merging

Abstract

Greater encoding efficiency can be achieved if, for a current block of an image whose bitstream indicates one of the supported partitioning patterns, partition inversion is avoided by merging blocks. In particular, if the indicated supported partitioning pattern specifies a subdivision of the block into two or more additional blocks, certain candidate encoding parameters are removed for all additional blocks except the first additional block in the encoding order. Specifically, those encoding parameters that match the encoding parameters associated with any of the additional blocks that, when merged with the corresponding additional block, would result in one of the supported partitioning patterns are removed from the set of candidate encoding parameters for the corresponding additional block.This avoids redundancy between partition coding and merge coding.

Description

Image encoding that supports block partitioning and block merging This application relates to image and / or video encoding and, in particular, codecs that support block partitioning and block merging. Many image and / or video codecs process images in blocks. For example, predictive codecs use block granularity to strike a balance between very precise prediction parameters set at high spatial resolution (which consumes too much related information) and overly approximate prediction parameters, which increases the number of bits needed to encode the prediction residue due to the lower spatial resolution. The optimal setting for prediction parameters lies somewhere between these two extremes. Several attempts have been made to find the optimal solution to the problem described above. For example, instead of using a regular subdivision of an image into blocks arranged in regular rows and columns, multitree partitioning seeks to increase the freedom to subdivide an image into blocks with a reasonable amount of subdivision information. However, even multitree partitioning requires signaling a significant amount of data, and the freedom to subdivide an image is still quite limited, even when using such a method. LI B. ET AL, "Redundancy reduction in Cbf and merge coding", JCTVC-C277, describes a proposal for the HEVC standard according to which merge coding is restricted to the extent that a partitioning of a coding unit into two prediction units is restricted to the extent that a first sub-partition in partition order may not be a merge candidate of the second prediction unit in partition order. DE FORNI R. ET AL, "On the Benefits of Leaf Merging in Quad-Tree Motion Models", IMAGE PROCESSING, 2005, describes a leaf merging procedure as an extension of quad-tree subdivision for the purpose of encoding video motion. To avoid redundancies between quad-tree subdivision and leaf merging, this article proposes allowing merging only with immediately adjacent leaves of a larger size, or leaves of the same size but with a different parent. To better balance the amount of related information needed to signal image subdivision with the freedom to subdivide the image, block merging can be used to increase the number of possible image subdivisions to a reasonable amount of additional data required to signal the merging information. For the blocks being merged, the encoding parameters only need to be transmitted once within the bitstream, similar to how the resulting merged group of blocks would be a directly subdivided portion of the image. However, there is still a need for improved encoding efficiency due to the redundancies introduced by the combination of block merging and block subdivision. Therefore, the object of the present invention is to provide a coding concept that has increased coding efficiency. This object is addressed by the pending independent claims. The underlying idea of ​​the present invention is that a further increase in encoding efficiency can be achieved if, for a current block of an image where the bitstream points to one of the supported partitioning patterns, partition inversion is avoided by merging blocks. In particular, if the pointing of the supported partitioning patterns specifies a subdivision of the block into two or more additional blocks, certain encoding parameter candidates are eliminated for all additional blocks except the first additional block in the encoding order.Specifically, these candidate encoding parameters are removed from the candidate encoding parameter set for the respective additional block, whose encoding parameters are the same as the encoding parameters associated with any of the additional blocks that, when merged with the respective additional block, would result in one of the supported partitioning pattern. This measure avoids redundancy between partitioning and merging encoding, and the signaling overhead for signaling merging information can be further reduced by exploiting the smaller size of the candidate encoding parameter set. Furthermore, the positive effects of combining block partitioning with block merging are maintained. That is, due to the combination of block partitioning with block merging, the variety of achievable partitioning patterns increases compared to the case without block merging.The increase in signaling overhead remains within reasonable limits. Finally, block merging allows additional blocks to be joined beyond the current block boundary, thus offering granularities that would not be possible without block merging. Applying a slightly different view of the set of candidates for merging, the idea explained above is manifested, according to a further aspect of the present invention, in a decoder configured to decode a bitstream signaling one of the supported partitioning patterns for a current block of an image, the decoder being configured to remove, if the signaled pattern of the supported partitioning patterns specifies a subdivision of the block into two or more additional blocks, for all additional blocks except a first additional block of the additional blocks in an encoding order, from a set of candidate blocks for the respective additional blocks, candidate blocks which, when merged with the respective additional blocks, would result in one of the supported partitioning patterns. Advantageous implementations of the present invention are the subject of the appended dependent claims. Preferred embodiments of the present application are described below in more detail with respect to the figures, including: Fig. 1 shows a block diagram of an encoder according to one embodiment; Fig. 2 shows a block diagram of a decoder according to one embodiment; Fig. 3 shows a block diagram of a possible internal structure of the encoder in Fig. 1; Fig. 4 shows a block diagram of a possible internal structure of the decoder in Fig. 2; Fig. 5a schematically shows a possible subdivision of an image into tree root blocks, encoding sets (blocks) and prediction sets (partitions); Fig. 5b shows a subdivision tree of the root tree block shown in Fig. 5a, down to the partition level, according to an illustrative example; Fig. 6 shows an implementation for a set of possible supported partitioning patterns according to an implementation; Fig. 7 shows possible partitioning patterns that effectively result from combining block merging and block partitioning when using block partitioning according to Fig. 6; Fig. 8 schematically shows candidate blocks for a JUMP / DIRECT mode according to one embodiment; Figure 9-11 shows portions of syntax from a syntax according to a realization; and Fig. 12 schematically shows the definition of neighboring partitions for a partition according to an implementation. With regard to the following description, it is noted that whenever the same reference symbol is used in relation to different figures, the explanations concerning the respective element presented in relation to one of these figures will apply equally to the other figures, provided that such transfer of explanations from one figure to another does not conflict with the remaining description of that other figure. Embodiments of the present invention are described, among others, in Figures 1 and 2 and in the corresponding passages of the description. It should be noted that, in general, in the description, the term "implementation" should be understood to mean "example" unless it is used to refer to matter within the scope of the appended claims. In other words, any teaching not within the scope of the appended claims is presented only to provide further examples and does not pertain to the invention. Figure 1 shows an encoder 10 according to an embodiment of the present invention. The encoder 10 is configured to encode an image 20 into a bit stream 30. Naturally, the image 20 could be part of a video, in which case the encoder would be a video encoder. Image 20 comprises a block 40 that is currently to be encoded by encoder 10. As shown in Fig. 1, image 20 can comprise more than one block 40. For example, image 20 can be subdivided into a regular arrangement of blocks 40 such that the blocks 40 are arranged in rows and columns, as exemplified in Fig. 1. However, any other subdivision of image 20 into blocks 40 is also possible. In particular, the subdivision of image 20 into blocks 40 can be fixed, i.e., known to the decoder by default, or it can be signaled within the bit stream 30 to the decoder. In particular, the blocks 40 of image 20 can vary in size.For example, a multi-tree subdivision, such as a quad-tree subdivision, can be applied to image 20 or a regular pre-subdivision of image 20 into regularly arranged tree root blocks to obtain blocks 40 which, in this case, form the leaf blocks of the multi-tree subdivision. In any case, encoder 10 is configured to signal within bitstream 30 one of the supported partitioning patterns for current block 40. That is, encoder 10 decides whether, in some cases, for example, the speed distortion optimization sense is better for further partitioning block 40, and as to which of the supported partitioning patterns should be used for current block 40 in order to tailor the granularity at which certain encoding parameters are set within current block 40 of image 20. As will be described in more detail below, encoding parameters can, for example, represent prediction parameters such as inter-prediction parameters. Such inter-prediction parameters can, for example, comprise a reference image index, a motion vector, and the like.The supported partitioning patterns may, for example, include a no-partition mode, i.e., an option whereby the current block 40 is not further partitioned; a horizontal partitioning mode, i.e., an option whereby the current block 40 is subdivided along a horizontal line into a higher and a lower portion; and a vertical partitioning mode, i.e., an option whereby the current block 40 is subdivided vertically along a vertical line into a left and a right portion. Furthermore, the supported partitioning patterns may also include an option whereby the current block 40 is further subdivided regularly into four additional blocks, each comprising one-quarter of the current block 40.Furthermore, the partition can belong to all the blocks 40 in image 20 or simply a suitable subset thereof, such as those with a certain encoding mode associated with them, such as inter-prediction mode. Additionally, the set of possible blocks for which merging will be applied to the block partition(s) can be further limited by bitstream signaling for each block 40 for which merging could be performed, indicating whether merging will be available for the block partition(s) or not. Naturally, such signaling could also be applied to each candidate partition for potential merging individually.In addition, different subsets of the supported partitioning modes may be available for block 40, depending, for example, on the block size, the subdivision level of block 40 if it is a multi-tree subdivision leaf block, in combination or individually. That is, while the subdivision of image 20 into blocks to obtain, among other things, block 40 can be fixed or signaled within the bitstream, the partitioning pattern to be used for the actual block 40 is signaled within bitstream 30 in the form of partitioning information. Consequently, the partitioning information can therefore be considered as a kind of extension of the subdivision of image 20 into blocks 40. On the other hand, some additional relevance of the original granularity of subdividing image 20 into blocks 40 may still remain.For example, encoder 10 can be configured to signal within bit stream 30 the encoding mode to be used for the respective portion or block 40 of image 20 at the granularity defined by block 40, while encoder 10 is configured to vary the encoding parameters of the respective encoding mode within the respective block 40 at an increased (finer) granularity defined by the respective partitioning pattern chosen for the respective block 40. For example, the encoding mode signaled at the granularity of blocks 40 can distinguish between intra-prediction mode, inter-prediction mode, and the like, such as inter-temporal prediction mode, inter-view prediction mode, etc.The type of encoding parameters associated with the one or more sub-blocks (partitions) resulting from the partitioning of the respective block 40 then depends on the encoding mode assigned to the respective block 40. For example, for an intra-coded block 40, the encoding parameters may comprise a spatial direction along which the image content of previously decoded portions of image 20 is used to fill the respective block 40. In the case of an inter-coded block 40, the encoding parameters may comprise, among other things, a motion vector for compensated motion prediction. Figure 1 illustrates how the current block 40 is subdivided into two additional (smaller) blocks, 50 and 60. Specifically, it illustrates a vertical partitioning mode. The smaller blocks 50 and 60 can also be referred to as sub-blocks 50 and 60, partitions 50 and 60, or prediction sets 50 and 60.In particular, encoder 10 is configured to remove, in those cases where the indicated supported partitioning patterns specify a subdivision of the current block 40 into two or more additional blocks 50 and 60, for all additional blocks except the first additional block of the additional blocks 50 and 60 in an encoding order, from a set of encoding parameter candidates for the respective additional block, the encoding parameter candidates that have encoding parameters that are the same as the encoding parameters associated with any of the additional blocks that, when merged with the respective additional blocks, would result in one of the supported partitioning patterns. More precisely, for each of the supported partitioning patterns, an encoding order is defined among the one or more resulting partitions 50 and 60. In the case of Fig.1, the coding order is illustrated in an exemplary way by an arrow 70, which defines that the left partition 50 is coded before the right partition 60. In the case of a horizontal partitioning mode, it could be defined that the upper partition is coded before the lower partition.In any case, the encoder 10 is configured to remove, for the second partition 60 in encoding order 70, from the set of candidate encoding parameters for the respective second partition 60, candidate encoding parameters that have encoding parameters that are the same as the encoding parameters associated with the first partition 50 in order to avoid the result of this merging, specifically the fact that both partitions 50 and 60 would have the same encoding parameters associated with the same ones that, in fact, they could equally produce by choosing the non-partitioning mode for the current block 40 at a lower encoding rate. To be more precise, encoder 10 is configured to use block merging effectively in conjunction with block partitioning. Regarding block merging, encoder 10 determines a respective set of encoding parameter candidates for each partition 50 and 60. The encoder can be configured to determine the sets of encoding parameter candidates for each partition 50 and 60 based on the encoding parameters associated with previously decoded blocks. Specifically, at least some of the encoding parameter candidates within the sets of encoding parameter candidates can be the same; that is, they can be derived from the encoding parameters of previously decoded partitions.Additionally or alternatively, at least some of the coding parameter candidates may be derived from coding parameter candidates associated with more than one previously coded partition, through a suitable combination, such as a median, mean, or similar.However, since encoder 10 is configured to perform the determination of the reduced set of encoding parameter candidates, and if more than one such encoding parameter candidate remains after deletion, the choice among the remaining non-deleted encoding parameter candidates, for each of the non-first partitions 60, in order to establish encoding parameters associated with the respective partition depending on the non-deleted or chosen encoding parameter candidate, encoder 10 is configured to perform the deletion in such a way as to delete encoding parameter candidates that would effectively lead to a reunification of partitions 50 and 60.That is, syntax constellations are effectively avoided according to which an effective partitioning situation is encoded more complexly than in the case of directly pointing to this partitioning simply by using partitioning information only. Furthermore, as the sets of candidate encoding parameters are reduced, the amount of related information needed to encode the merge information in bitstream 30 can decrease due to the smaller number of elements in these candidate sets. In particular, because the decoder is able to determine and subsequently reduce the sets of candidate encoding parameters in the same way as the encoder in Fig. 1, the encoder 10 in Fig. 1 can exploit the reduced sets of candidate encoding parameters by, for example, using fewer bits to insert a syntax element into bitstream 30, specifying which of the non-eliminated candidate encoding parameters will be used for merge.Naturally, the introduction of the syntax element into bitstream 30 can be completely suppressed if the number of non-eliminated encoding parameter candidates for the respective partition is only one. In any case, due to merging—that is, setting the encoding parameters associated with the respective partition based on the remaining, or selected, non-eliminated encoding parameter candidates—encoder 10 is able to suppress the completely new insertion of encoding parameters for the respective partition into bitstream 30, thereby also reducing the related information. According to some embodiments of this application, encoder 10 can be configured to signal refinement information within bitstream 30 to refine the remaining, or selected, encoding parameter candidates for the respective partitions. According to the description in Fig. 1, as previously stated, encoder 10 is configured to determine which merge candidates to be eliminated by comparing their encoding parameters with the encoding parameters of the partition whose merge would produce another supported partition pattern. This method of handling encoding parameter candidates would effectively eliminate at least one encoding parameter candidate in the illustrative case of Fig. 1, for example, whenever the encoding parameters of the left partition 50 form an element of the set of encoding parameter candidates for the right partition 60. However, additional encoding parameter candidates may also be eliminated if they are equal to the encoding parameters of the left partition 50.However, according to another embodiment of the present invention, the encoder 10 could be configured to determine a set of candidate blocks for each second partition and subsequent partitions in coding order, with the removal of that candidate block or those blocks from this set of candidate blocks which, when merged with the respective partition, would result in one of the supported partition patterns. In a sense, this means the following: The encoder 10 can be configured to determine merge candidates for a respective partition 50 or 60 (i.e., the first and subsequent partitions in coding order) such that each element of the candidate set has exactly one partition of the current block 40 or any of the previously coded blocks 40 associated with it, in the sense that the candidate adopts the respective coding parameters of the associated partition.For example, each element in the candidate set could be equal to, i.e., adopted from, one of such encoding parameters of previously encoded partitions, or at least could be derived from the encoding parameters of just one of such previously encoded partitions, such as by further scaling or refinement using additional refinement information. However, the encoder 10 could also be configured to accompany such a candidate set with other elements or andidates, specifically encoding parameter candidates that have been derived from a combination of encoding parameters from more than one previously encoded partition, or that have been derived—by modification—from encoding parameters of a previously encoded partition, such as by simply taking the encoding parameters from a list of move parameters.For "combined" elements, there is no 1:1 association between the encoding parameters of the respective candidate element and a respective partition. According to the first alternative in the description of Fig. 1, encoder 10 could be configured to remove all candidates from the entire candidate set whose encoding parameters are equal to the encoding parameters of partition 50. According to the last alternative in the description of Fig. 1, encoder 10 could be configured to remove only the element from the candidate set that is associated with partition 50.By harmonizing both perspectives, encoder 10 could be configured to remove candidates from the portion of the candidate set that exhibits a 1:1 association with some previously encoded partitions (e.g., neighbors), without extending the removal (and the search for candidates with matching encoding parameters) to the remaining portion of the candidate set that has encoding parameters obtained through merging. However, if merging also resulted in a redundant representation, this could be resolved by removing the redundant encoding parameters from the list or by also performing the redundancy check for the merged deadlines. Having described an encoder according to one embodiment of the present invention, with reference to Fig. 2, a decoder 80 according to one embodiment is described. The decoder 80 of Fig. 2 is configured to decode the bit stream 30 which, as described above, signals one of the supported partitioning patterns for an actual block 40 of the image 20. The decoder 80 is configured to, if the signaling of the supported partitioning pattern specifies a subdivision of the actual block 40 into two or more partitions 50 and 60, remove for all partitions except the first partition 50 from the partitions in encoding order 70, i.e., for partition 60 in the example illustrated in Figs.1 and 2, from a set of candidate encoding parameters for the respective partition encoding parameter candidates that have encoding parameters that are the same as, or equal to, encoding parameters associated with any of the partitions that, when merged with the respective partition, would result in one of the supported partition patterns, specifically one that has not been signaled within bit stream 30 but is, nevertheless, one of the supported partition patterns. In other words, the functionality of the decoder largely coincides with that of the encoder described with respect to Fig. 1. For example, the decoder 80 can be configured so that, if a number of the non-eliminated encoding parameter candidates is non-zero, it sets encoding parameters associated with the respective partition 60 depending on one of the non-eliminated parameter candidates. For example, the decoder 80 sets the encoding parameters of partition 60 to be equal to one of the non-eliminated encoding parameter candidates, with or without further refinement and / or with or without scaling according to a time distance to which the encoding parameters refer, respectively.For example, the encoding parameter candidate to be merged with the non-eliminated candidates may have a different reference image index associated with it, distinct from a reference image index explicitly noted within bitstream 30 for partition 60. In that case, the encoding parameters of the encoding parameter candidates may define motion vectors, each related to a respective reference image index, and the decoder 80 may be configured to scale the motion vector of the ultimately chosen non-eliminated encoding parameter candidate according to the relationship between the two reference image indices. Therefore, according to the aforementioned alternative, the encoding parameters undergoing merging would encompass the motion parameters, while the reference image indices would remain separate from them.However, as stated above, according to alternative realizations, the reference image indices could also be a part of the encoding parameters that are subjected to fusion. The same applies to the encoder in Fig. 1 and the decoder in Fig. 2: the merging behavior can be restricted to the interpredicted blocks 40. Therefore, the decoder 80 and the encoder 10 can be configured to support intra- and inter-prediction modes for the current block 40 and perform merging and candidate removal simply if the current block 40 is encoded in inter-prediction mode. Consequently, the encoding / prediction parameters of such previously encoded interpredicted partitions can simply be used to determine / build the candidate list. As discussed earlier, the encoding parameters can be prediction parameters, and decoder 80 can be configured to use the prediction parameters of partitions 50 and 60 to derive a prediction signal for the respective partition. Naturally, encoder 10 also derives the prediction signal in the same way. However, encoder 10 additionally sets the prediction parameters along with all other syntax elements within bitstream 30 to achieve some optimization. Furthermore, as previously described, the encoder can be configured to insert an index into a non-eliminated encoding parameter candidate simply if the number of non-eliminated encoding parameter candidates for a given partition is greater than one. Therefore, decoder 80 can be configured to expect bitstream 30 to contain a syntax element specifying which non-eliminated encoding parameter candidate is used for merging, depending on the number of non-eliminated encoding parameter candidates for, say, partition 60.However, the possibility of the candidate set becoming smaller than two can generally be avoided by extending the candidate list / set, as described above, using combined encoding parameters—that is, parameters derived from combining the encoding parameters of more than one (or more than two) previously encoded partitions. This is achieved by restricting the performance of reducing the candidate set to those candidates obtained by adopting, or deriving from, the encoding parameters of exactly one previously encoded partition. The reverse is also possible: generally eliminating all encoding parameter candidates that have the same value as those in the partition, resulting in a different supported partitioning pattern. With regard to the determination, the decoder 80 acts as the encoder 10 does. That is, the decoder 80 can be configured to determine the set of candidate coding parameters for the partition or partitions after the first partition 50 in coding order 70 based on the coding parameters associated with previously decoded partitions. In other words, a coding order is defined not only between partitions 50 and 60 of a respective block 40, but also between the blocks 40 of the image 20 itself. All partitions that have been coded before partition 60 can therefore serve as the basis for determining the set of candidate coding parameters for any of the subsequent partitions, such as partition 60 in the case of Fig. 2.As also described earlier, the encoder and decoder can restrict the determination of the set of candidate encoding parameters to partitions in a specific spatial and / or temporal neighborhood. For example, the decoder 80 can be configured to determine the set of candidate encoding parameters for a non-first partition 60 based on the encoding parameters associated with previously decoded partitions neighboring the respective non-first partition, where such partitions may be outside and inside the current block 40. Naturally, the determination of merge candidates can also be performed for the first partition in encoding order. The extraction is simply not performed. Consistent with the description in Fig. 1, the decoder 80 can be configured to determine the set of candidate encoding parameters for the respective non-first partition 60 of an initial set of previously decoded partitions, excluding those encoded in an intra-prediction mode. Furthermore, in case the encoder introduces subdivision information into the bitstream in order to subdivide image 20 into blocks 40, the decoder 80 can be configured to retrieve the subdivision of image 20 into such encoding blocks 40 according to the subdivision information in bitstream 30. With regard to Figures 1 and 2, it should be noted that the residual signal for the current block 40 can be transmitted through the bitstream 30 at a granularity that may differ from the granularity defined by the partitions with respect to the encoding parameters. For example, the encoder 10 in Figure 1 can be configured to subdivide block 40 into one or more transformation blocks in parallel or independently of the partitioning into partitions 50 and 60. The encoder can signal the respective transformation block subdivision for block 40 through additional subdivision information. The decoder 80, in turn, can be configured to recover this additional subdivision of block 40 into one or more transformation blocks according to the additional subdivision information in the bitstream, and to derive a residual signal from the current block 40 from the bitstream in sets of these transformation blocks.The significance of the transformation block partitioning lies in the fact that the transformation, such as DCT, in the encoder and the corresponding inverse transformation, such as IDCT, in the decoder are performed within each individual transformation block of block 40. In order to reconstruct image 20 as block 40, the encoder 10 then combines, as it adds, the prediction signal derived by applying the encoding parameters in the respective partitions 50 and 60, and the residual signal. However, it is noted that the residual encoding may not involve any transformation and inverse transformation, and that the prediction residue is encoded in the spatial domain instead, for example. Before describing further details of possible embodiments below, a possible internal structure of the encoder and decoder of Figs. 1 and 2 will be described in relation to Figs. 3 and 4. Fig. 3 shows in an exemplary way how the encoder 10 can be constructed internally. As shown, the encoder 10 can comprise a subtractor 108, a transformer 100, and a bitstream generator 102, which, as indicated in Fig. 3, can perform entropy coding. Elements 108, 100, and 102 are connected in series between an input 112 that receives the image 20 and an output 114 that produces the bitstream 30 mentioned above. In particular, the subtractor 108 has its non-inverting input connected to input 112 and the transformer 100 is connected between an output of the subtractor 108 and a first input of the bitstream generator 102 which, in turn, has an output connected to output 114.The encoder 10 of Fig. 3 further comprises an inverting transformer 104 and an adder 110 connected in series, in the order mentioned, to the output of the transformer 100. The encoder 10 further comprises a predictor 106, which is connected between an output of the adder 110 and an additional input of the adder 110 and the inverting input of the subtractor 108. The elements in Fig. 3 interact as follows: Predictor 106 predicts portions of image 20, and the prediction result, i.e., the prediction signal, is applied to the inverting input of subtractor 108. The output of subtractor 108, in turn, represents the difference between the prediction signal and the respective portion of image 20, i.e., a residual signal. The residual signal is then subjected to transformation encoding in transformer 100. That is, transformer 100 can perform a transformation, such as a DCT or similar, and subsequent quantization of the transformed residual signal, i.e., the transformation coefficients, to obtain transformation coefficient levels.The inverting transformer 104 reconstructs the final residual signal output produced by transformer 100 to obtain a reconstructed residual signal that corresponds to the residual signal fed into transformer 100, except for the information loss due to quantization in transformer 100. The addition of the reconstructed residual signal and the prediction signal as output of predictor 106 results in a reconstruction of the respective portion of image 20, which is then fed from the output of adder 110 to the input of predictor 106. Predictor 106 operates in different modes, as described above, such as intra-prediction mode, inter-prediction mode, and the like.The prediction mode and the corresponding encoding or prediction parameters applied by predictor 106 in order to obtain the prediction signal are forwarded by predictor 106 to entropy encoder 102 for insertion into the bit stream. Figure 4 shows a possible implementation of the internal structure of the decoder 80 in Figure 2, which corresponds to the possibility shown in Figure 3 with respect to the encoder. As shown therein, the decoder 80 can comprise a bitstream extractor 150, which, as shown in Figure 4, can be implemented as an entropy decoder, an inverse transformer 152, and an adder 154, which are, in that order, connected between an input 158 ​​and an output 160 of the decoder. Furthermore, the decoder in Figure 4 comprises a predictor 156 connected between an output of the adder 154 and an additional input of the same. The entropy decoder 150 is connected to a parameter input of the predictor 156. Briefly describing the functionality of the decoder in Fig. 4, the entropy decoder 150 is used to extract all the information contained in the bit stream 30. The entropy coding scheme used can be either variable-length coding or arithmetic coding. The entropy decoder 150 recovers the transformation coefficient levels, representing the residual signal, from the bit stream and forwards them to the inverse transformer 152. Furthermore, the entropy decoder 150 recovers all the coding modes and associated coding parameters from the bit stream and forwards them to the predictor 156. Additionally, partitioning information and merging information are extracted from the bit stream by the extractor 150.The inversely transformed residual signal, i.e., reconstructed, and the prediction signal as derived by predictor 156 are combined, as they are added, by adder 154, which in turn produces as output the reconstructed signal thus recovered at output 160 and forwards it to predictor 156. As is clear from comparing Figs. 3 and 4, elements 152, 154 and 156 correspond functionally to elements 104, 110 and 106 of Fig. 3. In the preceding description of Figures 1 to 4, several different possibilities have been presented regarding possible subdivisions of image 20 and the corresponding granularity by varying some of the parameters involved in encoding image 20. One such possibility is described again with respect to Figures 5a and 5b. Figure 5a shows a portion of image 20. According to the implementation of Figure 5a, the encoder and decoder are configured to first subdivide image 20 into tree-root blocks 200. One such tree-root block is shown in Figure 5a. The subdivision of image 20 into tree-root blocks is performed regularly in rows and columns, as illustrated by dashed lines. The size of the tree-root blocks 200 can be selected by the encoder and signaled to the decoder by bit stream 30.Alternatively, the size of these 200-tree root blocks can be fixed by default. The 200-tree root blocks are subdivided using quad-tree partitioning to produce the previously identified 40 blocks, which can be referred to as encoding blocks or encoding sets. These encoding blocks or encoding sets are drawn with thin continuous lines in Fig. 5a. The encoder accompanies each 200-tree root block with subdivision information and inserts this information into the bitstream. This subdivision information indicates how the 200-tree root block should be subdivided into 40 blocks. At a granularity of, and in sets of, these 40 blocks, the prediction mode varies within the 20-block. As previously stated, each 4-block...Each block that has a certain prediction mode, such as inter-prediction mode, is accompanied by partitioning information regarding which supported partitioning pattern is used for the respective block 40. In the illustrative case of Fig. 5a, for many encoding blocks 40, the non-partitioning mode has been chosen so that the encoding block 40 spatially coincides with the corresponding partition. In other words, the encoding block 40 is simultaneously a partition that has a respective set of prediction parameters associated with it. The type of prediction parameters, in turn, depends on the mode associated with the respective encoding block 40. However, it is shown as an example that other encoding blocks are further partitioned.The encoding block 40 in the upper right corner of the tree-root block 200, for example, is shown to be partitioned into four partitions, while the encoding block in the lower right corner of the tree-root block 200 is shown to be vertically subdivided into two partitions. The subdivision for partitioning into partitions is illustrated by dashed lines. Fig. 5a also shows the encoding order between the partitions thus defined. As shown, a first transverse depth-in-order is used. Across the edges of the tree-root block, the encoding order can be continued in a scan order according to which the rows of tree-root blocks 200 are scanned in rows from top to bottom of Figure 20.This measure maximizes the probability that a given partition will have a pre-coded partition adjacent to its top and left edges. Each block 4—or each block with a specific prediction mode, such as inter-prediction mode—can have a merge switch indicator within the bitstream that shows whether merging is enabled for the corresponding partitions within it. It should be noted that partitioning blocks into prediction sets could be restricted to a maximum of two partitions, with the exception of the smallest possible block size for blocks 40.If quad-tree subdivision is used to obtain the 40 blocks, this could avoid redundancy between the subdivision information for subdividing image 20 into block 40 and the partitioning information for subdividing block 40 into partitions. Alternatively, partitioning into one or two partitions could be allowed, with or without asymmetric partitioning. Figure 5b shows a subdivision tree. Solid lines illustrate the subdivision of the root tree block 200, while dashed lines symbolize the partitioning of the leaf blocks of the quad tree subdivision, which are the coding blocks 40. That is, the partitioning of the coding blocks represents a kind of extension of the quad subdivision. As noted earlier, each encoding block 40 can be subdivided in parallel into transformation blocks so that the transformation blocks can represent a different subdivision of the respective encoding block 40. For each of these transformation blocks, which are not shown in Figs. 5a and 5b, a transformation to convert the residual signal from the encoding blocks can be performed separately. The following describes further embodiments of the present invention. While the preceding embodiments focused on the relationship between block merging and block partitioning, the following description also includes aspects of the present application related to other coding principles known in current codecs, such as JUMP / DIRECT modes. However, the following description should not be considered a mere description of separate embodiments, i.e., embodiments separate from those described above. Rather, the following description also discloses possible implementation details for the embodiments described above.Therefore, the description below uses reference symbols from the figures already described above, so that a possible respective implementation described below will also define possible variations of the embodiments described above. Most of these variations can be individually transferred to the embodiments above. In other words, the embodiments of this application describe methods for slowing down related information in image and video encoding applications by merging the syntax elements associated with particular sets of samples, i.e., blocks, in order to transmit associated encoding parameters. The embodiments of this application are particularly capable of combining the merging of syntax elements with partitioning parts of an image into various partitioning patterns and combining this with JUMP / DIRECT modes, in which the encoding parameters are inferred from the spatial and / or temporal neighborhood of a current block. To the extent that the embodiments described above can be modified to implement merging for sets of samples, i.e., blocks, in combination with different partitioning patterns and JUMP / DIRECT modes. Furthermore, before describing these variations and more details, an overview of image and video codecs is presented. In image and video encoding applications, the sample arrays associated with an image are typically partitioned into particular sets of samples (or sample sets), which can represent rectangular or square blocks, or any other collection of samples that includes arbitrarily shaped regions, triangles, or any other shape. The subdivision of the sample arrays can be fixed by the syntax, or the subdivision is signaled (at least partially) within the bitstream. To maintain the speed of connected information signaling small subdivision information, the syntax generally allows only a limited number of options that result in simple partitioning, such as subdividing blocks into smaller blocks. One partitioning scheme often used is the partitioning of square blocks into four smaller square blocks, or into two rectangular blocks of the same size.or in two rectangular blocks of different sizes, where the partitioning used is actually indicated within the bitstream. Sample sets are associated with particular encoding parameters that can specify prediction information or residual encoding modes, etc. In video encoding applications, partitioning is often performed for motion representation purposes. All samples in a block (within a partitioning pattern) are associated with the same set of motion parameters, which may include parameters specifying the prediction type (e.g., list 0, list 1, or biprediction; and / or translational or affine prediction, or a prediction with a different motion model), parameters specifying the reference images used, and parameters specifying the motion relative to the reference images (e.g., displacement vectors,Vectors of affine motion parameters (or vectors of motion parameters for any other motion model), which are generally passed as a difference to a predictor; parameters specifying the accuracy of the motion parameters (e.g., half-sample or quarter-sample accuracy); parameters specifying the weighting of the reference sample signal (e.g., for lighting compensation); or parameters specifying the interpolation filter used to derive the motion-compensated prediction signal from the current block. It is assumed that for each set of samples, individual encoding parameters are passed (e.g., to specify the prediction and / or residual encoding). In order to achieve improved encoding efficiency,The present invention presents a method and particular embodiments for merging two or more sample sets into so-called sample set groups. All sample sets in such a group share the same encoding parameters, which can be transmitted together with one of the sample sets in the group. By doing so, the encoding parameters do not need to be transmitted for each sample set in the sample set group individually; instead, the encoding parameters are transmitted only once for the entire group of sample sets. As a result, the rate of related information transmission of the encoding parameters is reduced, and the overall encoding efficiency is improved. As an alternative strategy,Additional refinement can be transmitted for one or more of the encoding parameters for one or more of the sample sets within a group of sample sets. The refinement can be applied to all sample sets in a group or only to the sample set for which it is transmitted. The embodiments of the present invention relate particularly to combining the fusion process with partitioning a block into several sub-blocks 50, 60 (as mentioned above). Typically, image or video encoding systems support various partitioning patterns for a block 40. For example, a square block may remain unpartitioned, or it may be partitioned into four square blocks of the same size, or into two rectangular blocks of the same size (where the square block may be divided vertically or horizontally), or into rectangular blocks of different sizes (horizontally or vertically). The exemplary partitioning patterns described are illustrated in Fig. 6. In addition to the above description, partitioning may involve even more than one level of partitioning. For example, square sub-blocks may also be additionally partitioned optionally using the same partitioning patterns.The problem that arises when such a partitioning procedure is combined with a merging procedure that allows the merging of a block (square or rectangular) with, for example, one of its neighboring blocks, is that the resulting partition can be achieved through different combinations of partitioning patterns and merging signals. Therefore, the same information can be transmitted in the bitstream using different codewords, which is clearly suboptimal with respect to encoding efficiency. As a simple example, consider a square block that is not further partitioned (as illustrated in the upper left corner of Fig. 6). This partitioning can be directly signaled by sending a syntax element indicating that this block is not subdivided.But the same pattern can also be signaled by sending a syntax element that specifies that this block is, for example, subdivided into two vertically (or horizontally) aligned rectangular blocks 50, 60. We can then transmit merge information specifying that the second of these rectangular blocks is merged into the first rectangular block, resulting in exactly the same partitioning as when we signal that the block is not further divided. The same thing can also be achieved by first specifying that the block is subdivided into four square sub-blocks and then transmitting merge information that effectively merges all four of these blocks. This concept is clearly suboptimal (since we have different codewords to signal the same thing). The embodiments of the present invention describe a concept and possibilities for reducing the rate of related information and thus increasing coding efficiency by combining the concept of merging with the concept of providing different partitioning patterns for a block. If we observe the example partitioning patterns in Fig. 6, the "simulation" of the block not further divided by any of the partitioning patterns with two rectangular blocks can be avoided by prohibiting (i.e., excluding from the bitstream syntax specification) the case of a rectangular block being merged with a first rectangular block.When the problem is analyzed more deeply, it is also possible to "simulate" the unsubdivided pattern by merging the second rectangular block with any other neighbor (i.e., not the first rectangular block) that is associated with the same parameters (e.g., information to specify the prediction) as the first rectangular block. The embodiments of the present invention condition the transmission of merging information such that the transmission of particular merging parameters is excluded from the bitstream syntax when these merging parameters result in a pattern that can also be achieved by pointing to one of the supported partitioning patterns. For example, if the current partitioning pattern specifies subdivision into two rectangular blocks, as shown in Fig. 1 and 2, for example, before sending the merging information for the second block, i.e., 60 in the case of Fig.In Figures 1 and 2, it can be verified which of the potential merge candidates has the same parameters (e.g., parameters to specify the prediction signal) as the first rectangular block, i.e., 50 in the case of Figures 1 and 2. All candidates with the same movement parameters (including the first rectangular block itself) are then removed from the merge candidate set. The codewords or flags transmitted to signal merge information are adapted to the resulting candidate set. If the candidate set is empty due to parameter verification, no merge information is transmitted. If the candidate set consists of a single entry, it is only signaled whether the block is merged or not, but it is not necessary to signal the candidate, as this can be derived at the decoder side, etc.For the previous example, the same concept is also used for the partitioning pattern that divides a square block into four smaller square blocks. Here, the merge flag sending is adapted so that neither the partitioning pattern that doesn't specify any subdivision nor either of the two partitioning patterns that specify a subdivision into two rectangular blocks of the same size can be achieved by a combination of merge flags. Although we described the concept more clearly in the previous example with specific partitioning patterns, it should be evident that the same concept—avoiding the specification of a particular partitioning pattern by using a combination of another partitioning pattern and the corresponding merge information—can be used for any other set of partitioning patterns. The advantage of the described invention over a concept that only allows partitioning is that it provides much greater freedom to designate the partitioning of an image into parts that are associated with the same parameters (e.g., to specify the prediction signal). As an example, the additional partitioning patterns resulting from the fusion of square blocks from a larger subdivided block are shown in Fig. 7. However, it should be noted that many more resulting patterns can be achieved by fusing with additional neighboring blocks (outside the previously subdivided block).With only a few codewords to signal partitioning and merging information, a variety of partitioning possibilities are provided, and an encoder can select the best option (for a given encoder complexity) in terms of speed distortion (e.g., minimizing a particular speed distortion measure). The advantage over a strategy where only one partitioning pattern (e.g., subdividing into four equal-sized blocks) is provided in combination with the merging strategy is that frequently used patterns (such as rectangular shapes of different sizes) can be signaled by a short codeword instead of multiple subdivision and merge flags. Another aspect to consider is that the concept of fusion is, in a sense, similar to the JUMP or DIRECT modes found in video encoding designs. In JUMP / DIRECT modes, motion parameters for a current block are not transmitted but are inferred from a spatial and / or temporal neighborhood. In a particular efficient concept of JUMP / DIRECT modes, a list of candidate motion parameters (reference frame indices, displacement vectors, etc.) is created from a spatial and / or temporal neighborhood, and an index is transmitted to this list specifying which of the candidate parameters is chosen. For bipredicted blocks (or multi-hypothesis frames), a separate candidate can be signaled for each reference list.Possible candidates may include the block at the top of the current block, the block to the left of the current block, the block at the top left of the current block, the block at the top right of the current block, the median predictor of several of these candidates, the block co-located in one or more previous reference frames (or any other already encoded block, or a combination obtained from already encoded blocks). When combining the concept of merging with the JUMP / DIRECT mode, it should be ensured that both the JUMP / DIRECT mode and the merge mode do not include the same candidates. This can be achieved through different configurations. It is possible to enable the JUMP / DIRECT mode (for example, with more candidates than the merge mode) only for particular blocks (for example, larger than a specified size, or only for square blocks, etc.) and not support the merge mode for these blocks.Alternatively, the JUMP / DIRECT mode can be removed, and all candidates (including parameters representing a combination of parameters for spatial / temporal neighbor blocks) are added to the merge mode as candidates. This option was also mentioned earlier with respect to Figs. 1-5. The augmented candidate set could only be used for particular blocks (larger than a given minimum size, or square blocks, etc.), while a reduced candidate set is used for other blocks. As a further variant, the merge mode can be used with a reduced candidate set (e.g., only the top and left neighbors), and additional candidates (e.g., the top-left mode, the co-located block, etc.) are used for the JUMP / DIRECT mode.Also in such configurations, JUMP / DIRECT modes can only be allowed for particular blocks (larger than a given minimum size, or square blocks, etc.), while merge mode is allowed for a larger set of blocks. The advantage of such combinations is that multiple options are provided for signaling the reuse of already transmitted parameters (e.g., to specify the prediction) for different block sizes. For example, more options are provided for larger square blocks, since the additional bit rate spent here provides an increase in rate distortion efficiency. For smaller blocks, a smaller set of options is provided. Increasing the set of candidates here would not bring any gain in rate distortion efficiency due to the small ratio of samples per bit required to signal the selected andidate. As mentioned previously, embodiments of the present invention also provide an encoder with greater freedom in creating a bitstream, since the fusion strategy significantly increases the number of possible partitions for the sample arrays of an image. Because the encoder can choose from more options, for example, to minimize a particular velocity distortion measure, encoding efficiency can be improved. For example, some of the additional patterns that can be represented by a combination of sub-partitioning and fusion (e.g., the patterns in Fig. 7) can be further tested (using the corresponding block sizes for motion estimation and mode decision), and the best of the patterns provided by pure partitioning (Fig. 6) and by partitioning and fusion (Fig. 7) can be further tested.7) Selection can be based on a particular speed distortion measure. Furthermore, for each block, it can be tested whether merging with any of the already encoded candidate sets reduces a particular speed distortion measure, and then the corresponding merge flags are set during the encoding procedure. In summary, there are several ways to operate an encoder. In a simple strategy, the encoder could first determine the best subdivision of the sample arrays (as in state-of-the-art encoding schemes). Then, for each sample set, it could check whether merging with another sample set or another group of sample sets reduces a particular speed distortion cost measure.In this case, the prediction parameters associated with the merged group of sample sets can be re-estimated (for example, by performing a new motion search), or the prediction parameters already determined for the current sample set and the candidate sample set (or group of sample sets) for the merger could be evaluated for the group of sample sets under consideration. In a broader strategy, a particular measure of velocity distortion cost could be evaluated for additional groups of candidate sample sets. As a specific example, when testing the various possible partitioning patterns (see Fig. 6, for example), some or all of the pattern that can be represented by a combination of partitioning and merging (see Fig. 7, for example) can be further tested.In other words, a specific motion estimation and mode decision procedure is performed for all patterns, and the pattern that produces the smallest velocity distortion measure is selected. This procedure can also be combined with the low-complexity procedure described above, so that for the resulting blocks, it is further tested whether merging them with already coded blocks (e.g., outside the patterns in Fig. 6 and Fig. 7) produces a decrease in a velocity distortion measure. The following section describes some possible detailed implementations for the embodiments described above, such as the encoders in Figures 1 and 3 and the decoders in Figures 2 and 4. As previously mentioned, these are usable in image and video encoding. As described above, images or particular sets of sample arrays for images can be decomposed into blocks, which are associated with specific encoding parameters. Images generally consist of multiple sample arrays. In addition, an image can also be associated with additional auxiliary sample arrays, which can, for example, specify transparency information or depth maps. The sample arrays of an image (including auxiliary sample arrays) can be grouped into one or more so-called plane groups, where each plane group consists of one or more sample arrays.The plane groups of an image can be encoded independently or, if the image is associated with more than one plane group, with predictions for other plane groups within the same image. Each plane group is typically decomposed into blocks. The blocks (or the corresponding blocks in the sample arrays) are predicted using inter-image or intra-image prediction. The blocks can be of different sizes and can be square or rectangular. The partitioning of an image into blocks can be fixed by the syntax or can be (at least partially) signaled within the bitstream. Syntax elements that signal subdivision into blocks of predefined sizes are often transmitted. Such syntax elements can specify whether and how a block is subdivided into smaller blocks and are associated with encoding parameters, for example, for prediction purposes. An example of possible partitioning patterns is shown in Fig. 6.For all samples within a block (or corresponding blocks within sample sets), the decoding of the associated encoding parameters is specified in a certain way. In the example, all samples within a block are predicted using the same set of prediction parameters, such as reference indices (identifying a reference image within the already encoded image set), motion parameters (specifying a measure for the movement of a block between a reference image and the current image), parameters for specifying the interpolation filter, intra-prediction modes, and so on. The motion parameters can be represented by displacement vectors with a horizontal and a vertical component, or by higher-order motion parameters, such as affine motion parameters consisting of six components.It is also possible for more than one set of particular prediction parameters (such as reference indices and motion parameters) to be associated with a single block. In that case, for each set of these particular prediction parameters, a single intermediate prediction signal is generated for the block (or the corresponding blocks of sample sets), and the final prediction signal is constructed by combining and overlaying the intermediate prediction signals. The corresponding weighting parameters, and potentially also a constant offset (which is added to the weighted sum), can be set for an image, a reference image, or a set of reference images, or they can be included in the prediction parameter set for the corresponding block.The difference between the original blocks (or the corresponding sample array blocks) and their prediction signals, also called the residual signal, is typically transformed and quantized. Often, a two-dimensional transformation is applied to the residual signal (or the corresponding sample arrays for the residual block). For transformation encoding, the blocks (or the corresponding sample array blocks) for which a particular set of prediction parameters has been used can be further subdivided before applying the transformation. The transformation blocks can be the same size as, or smaller than, the blocks used for prediction. It is also possible for a transformation block to include more than one of the blocks used for prediction. Different transformation blocks can have different sizes, and the transformation blocks can represent quadratic or rectangular blocks.In the previous example for Figs. 1–5, it was observed that the leaf nodes of the first subdivision—that is, the coding blocks 40—can be further partitioned in parallel into the partition that defines the granularity of the coding parameters, on the one hand, and the transformation blocks to which the two-dimensional transformation is applied individually, on the other hand. After the transformation, the resulting transformation coefficients are quantized, and the so-called transformation coefficient levels are obtained. The transformation coefficient levels, as well as the prediction parameters and, if present, the subdivision information, are encoded by entropy. In the latest generation of image and video encoding standards, the possibilities for subdividing an image (or a group of shots) into blocks provided by the syntax are very limited. Typically, you can only specify if and (potentially how) a block of a predefined size can be subdivided into smaller blocks. As an example, the largest block size in H.264 is 16x16. 16x16 blocks are also known as macroblocks, and each image is partitioned into macroblocks at a first stage. For each 16x16 macroblock, you can specify whether it is encoded as a single 16x16 block, as two 16x8 blocks, as two 8x16 blocks, or as four 8x8 blocks. If a 16x16 block is subdivided into four 8x8 blocks, each of these 8x8 blocks can be encoded as one 8x8 block, or as two 8x4 blocks, or as two 4x8 blocks, or as four 4x4 blocks.The small set of possibilities for specifying block partitioning in next-generation image and video coding standards has the advantage that the rate of connected information used to signal subdivision information can be kept small, but it has the disadvantage that the bit rate required to transmit prediction parameters for blocks can become significant, as explained below. The rate of connected information used to signal prediction information generally represents a significant portion of the overall bit rate for a block. Encoding efficiency could be increased when this connected information is reduced, which, for example, could be achieved by using larger block sizes. It is also possible to increase the set of supported partitioning patterns compared to H.264. For example, the partitioning patterns represented in Fig.Six can be provided for square blocks of all sizes (or selected sizes). The actual images or images in a video sequence consist of arbitrarily shaped objects with specific properties. For example, such objects or parts of objects are characterized by a unique texture or a unique motion. And generally, the same set of prediction parameters can be applied to such an object or part of an object. But the object boundaries generally do not coincide with the possible block boundaries for large prediction blocks (e.g., 16x16 macroblocks in H.264). An encoder typically determines the subdivision (from the limited set of possibilities) that results in the minimum of a particular speed distortion cost measure. For arbitrarily shaped objects, this can result in a large number of small blocks.This statement also remains true when more partitioning patterns are provided (as mentioned earlier). It should be noted that the number of partitioning patterns should not be too large, as this then requires a lot of connected information and / or encoder / decoder complexity to signal and process these patterns. Therefore, objects with arbitrary shapes often result in a large number of small blocks due to partitioning. And since each of these small blocks is associated with a set of prediction parameters that must be transmitted, the connected information rate can become a significant part of the overall bit rate. But since several of the small blocks still represent areas of the same object or part of an object, the prediction parameters for a number of the resulting blocks are the same or very similar.Intuitively, coding efficiency could be increased by extending the syntax to allow not only subdividing a block but also merging two or more of the resulting blocks. This would produce a group of blocks encoded with the same prediction parameters. The prediction parameters for such a group of blocks only need to be encoded once. In the previous examples in Figures 1-5, for instance, the coding parameters for current clock 40 are not transmitted whenever merging occurs—that is, whenever the reduced set of candidates is not eliminated. In other words, the encoder does not transmit the coding parameters associated with the current block, and the decoder does not expect bitstream 30 to contain coding parameters for current block 40.Rather, according to their specific realizations, refinement information can simply be transmitted for the current merged block 40. A determination of a set of candidates and its reduction, as well as merging and so on, is also performed for the other 40 encoding blocks of image 20. The encoding blocks somehow form groups of encoding blocks along an encoding chain, where the encoding parameters for these groups are transmitted within the bitstream in their entirety only once. If the bit rate saved by reducing the number of encoded prediction parameters is greater than the additional bit rate spent encoding the merging information, the described merging results in greater coding efficiency. It should also be noted that the described syntax extension (for merging) provides the encoder with additional freedom when selecting the partitioning of an image or group of planes into blocks. The encoder is not restricted to performing the subdivision first and then checking whether any of the resulting blocks have the same set of prediction parameters. As a simpler alternative, the encoder could determine the subdivision first, as in state-of-the-art coding techniques.Then, for each block, you could check whether merging it with one of its neighboring blocks (or the associated predetermined block group) reduces a velocity distortion cost measure. In this case, the prediction parameters associated with the new block group can be re-estimated (for example, by performing a new motion search), or the prediction parameters already determined for the current block and its neighboring block or block group could be evaluated for the new block group. An encoder can also directly check (a subset of) the patterns provided by a combination of splitting and merging; that is, motion estimation and mode decision can be made with the resulting shapes, as mentioned earlier. Merge information can be signaled at the block level.Indeed, the merge could also be interpreted as inferring the prediction parameters for a current block, where the inferred prediction parameters are set equal to the prediction parameters of one of the neighboring blocks. In this case, it should be noted that combining different partitioning patterns and merge information can result in the same shapes (associated with the same parameters). This is clearly suboptimal, since the same message can be transmitted using different combinations of codewords. To avoid (or reduce) this drawback, the embodiments of the present invention describe a concept that prohibits the same shape (associated with a particular set of parameters) from being signaled by different partitioning and merge syntax elements. Therefore, for all blocks within a previously subdivided block, except the first in the encoding order—as verified in encoders and decoders such as 10 and 50—it is checked whether a merge would result in a pattern that could be signaled by a partition without merge information.All candidate blocks for which this is true are removed from the merge candidate set, and the transmitted merge information is tailored to the resulting candidate set. If no candidates remain, no merge information is transmitted; if one candidate remains, only a flag specifying whether the block is merged or not is transmitted, and so on. A preferred embodiment is described below for further illustration of this concept. The advantage of the described embodiments over a concept in which only partitioning is permitted is that they provide much greater freedom to signal the partitioning of an image into parts that are associated with the same parameters (e.g., to specify the prediction signal).The advantage compared to a strategy where only one partitioning pattern (e.g., subdividing into four equal-sized blocks) is provided in combination with the merging strategy is that often used patterns (such as rectangular shapes of different sizes) can be signaled by a short codeword instead of multiple subdivision and merge flags. Next-generation video coding standards like H.264 also contain specific intercode modes called JUMP and DIRECT modes, in which the parameters specifying the prediction are inferred entirely from spatially and / or temporally neighboring blocks. The difference between JUMP and DIRECT is that JUMP mode further indicates that no residual signal is transmitted. In several proposed enhancements to JUMP / DIRECT mode, instead of a single candidate (as in H.264), a list of possible candidates is inferred from a spatial and / or temporal neighborhood of the current block.Possible candidates may include the block at the top of the current block, the block to the left of the current block, the block at the top left of the current block, the block at the top right of the current block, the median predictor of several of these candidates, the block co-located in one or more previous reference frames (or any other already encoded block, or a combination derived from already encoded blocks). For a combination with the merge mode, it should be ensured that both the JUMP / STRAIGHT mode and the merge mode do not include the same candidates. This can be achieved through different configurations as mentioned previously. The advantage of the described combinations is that they provide multiple options for signaling the reuse of already transmitted parameters (e.g., to specify the prediction) for different block sizes. An advantage of the embodiments of the present invention is reducing the bit rate required to transmit prediction parameters by merging neighboring blocks into a block group, where each block group is associated with a unique set of encoding parameters, e.g., prediction parameters or residual encoding parameters. The merge information is signaled within the bit stream (in addition to the subdivision information, if present). In combination with different subdivision patterns and JUMP / DIRECT modes, it can be ensured that the JUMP / DIRECT mode and none of the provided patterns are "simulated" by sending the corresponding merge information. The advantage of the embodiments of the present invention is increased encoding efficiency resulting from a reduced rate of related information for the encoding parameters.The embodiments of the present invention are applicable to image and video coding applications, in which sets of samples are associated with particular coding or prediction parameters. The merging procedure described herein also extends to a third dimension or more dimensions. For example, a group of blocks in several video images could be fused into a single group of blocks. It could also be applied to 4D compression in light field coding. On the other hand, it can also be used for compression of 1D signals, where the 1D signal is partitioned and the resulting partitions are merged. The embodiments of the present invention also relate to a method for slowing down related information in image and video coding applications. In image and video coding applications, particular sets of samples (which may represent rectangular or square blocks, arbitrarily shaped regions, or any other collection of samples) are generally associated with a particular set of coding parameters. For each of these sample sets, the coding parameters are included in the bitstream. The coding parameters may represent prediction parameters, which specify how the corresponding set of samples is predicted using already coded samples. The partitioning of the sample arrays of an image into sample sets may be fixed by syntax or signaled by corresponding subdivision information within the bitstream.Multiple partitioning patterns may be permitted for a block. The encoding parameters for the sample sets are transmitted in a predefined order, which is given by the syntax. The embodiments of the present invention also represent a method by which it can be indicated that a current sample set is merged (for example, for prediction purposes) with one or more additional sample sets into a group of sample sets. Therefore, the possible set of values ​​for the corresponding merge information is tailored to the partitioning pattern employed, such that particular partitioning patterns cannot be represented by a combination of other partitioning patterns and the corresponding merge data. The encoding parameters for a group of sample sets need to be transmitted only once.In one particular embodiment, the encoding parameters for a current sample set are not transmitted if the current sample set is merged with a sample set (or group of sample sets) for which the encoding parameters have already been transmitted; instead, the encoding parameters for the current sample set are set equal to the encoding parameters of the sample set (or group of sample sets) with which the current sample set is merged. As an alternative strategy, additional refinement can be transmitted for one or more of the encoding parameters for a current sample set; the refinement can be applied to all sample sets in a group or only to the sample set for which it is transmitted. In a preferred embodiment, for each sample set, the set of all previously encoded sample sets is called the "set of causal sample sets." The sample sets that can be used for merging with a current sample set are called the "set of candidate sample sets" and are always a subset of the "set of causal sample sets." The way in which this subset is formed may be known to the decoder or may be specified within the bitstream. In either case, the encoder 10 and the decoder 80 determine that the set of candidates is reduced.If a particular current sample set is coded and its set of candidate sample sets is not empty, it is indicated (or derived) whether the current sample set merges with a sample set from this set of candidate sample sets and, if so, with which one (if more than one candidate exists). Otherwise, merging cannot be used for this block. Candidate blocks for which merging would result in a shape that could also be specified directly by a partitioning pattern are excluded from the candidate set in order to prevent the same shape from being represented by different combinations of partitioning information and merge data. That is, the candidate set is reduced by removing respective candidates as described above with respect to Figs. 1–5. In a preferred embodiment, a number of the set of candidate sample sets is zero or more sample sets containing at least one particular non-zero number of samples (which may be one, two, or even more) that represent direct spatial neighbors of any sample within the current sample set. In another preferred embodiment of the invention, the set of candidate sample sets may additionally (or exclusively) include sample sets containing one particular non-zero number of samples (which may be one, two, or even more) that have the same spatial location, i.e., are comprised of both the candidate sample sets and the current sample set currently undergoing merging, but are contained in a different image.In another preferred embodiment of the invention, the set of candidate sample sets can be derived from previously processed data within the current image or in other images. The derivation procedure may include spatial directional information, such as transformation coefficients, associated with a particular direction and image gradients of the current image, or it may include temporal directional information, such as neighboring motion representations. From such data available at the receiver and other related data and information (if present), the set of candidate sample sets can be derived.The elimination of candidates (from the original candidate set) that would result in the same shape that could be represented by a particular partitioning pattern is derived in the same way in the encoder and decoder, so that the encoder and decoder derive the final candidate set for merging in exactly the same way. In a preferred embodiment, the sample sets considered are rectangular or quadratic blocks. The merged sample sets then represent a collection of rectangular and / or quadratic blocks. In another preferred embodiment of the invention, the sample sets considered are image regions of arbitrary shape, and the merged sample sets represent a collection of image regions of arbitrary shape. In a preferred embodiment, one or more syntax elements are passed for each sample set, specifying whether the sample set is merged with another sample set (which may be part of an already merged group of sample sets) and which of the candidate sample sets is used for the merge. However, the syntax element is not passed if the candidate set is empty (for example, due to a deletion of candidates that would produce a partition that could be signaled by different non-merge partitioning patterns). In a preferred embodiment, one or two syntax elements are passed to specify the merge information. The first syntax element specifies whether the current sample set is merged with another sample set. The second syntax element, which is passed only if the first syntax element specifies that the current sample set is merged with another sample set, specifies which of the sets of candidate sample sets is used for the merge. In a preferred embodiment, the first syntax element is passed only if a derived set of candidate sample sets is not empty (after the potential removal of candidates that would produce a partition that could be signaled by a different partitioning pattern without merging).In another preferred embodiment, the second syntax element is transmitted only if a derived set of candidate sample sets contains more than one sample set. In another preferred embodiment of the invention, the second syntax element is transmitted only if at least two sample sets from a derived set of candidate sample sets are associated with different encoding parameters. In a preferred embodiment of the invention, the merge information for a sample set is encoded before the prediction parameters (or, more generally, the particular encoding parameters that are associated with the sample sets). The prediction or encoding parameters are transmitted only if the merge information indicates that the current sample set is not being merged with another sample set. In another preferred embodiment, the fusion information for a sample set is encoded after a subset of the prediction parameters (or, more generally, the particular encoding parameters associated with the sample sets) has been transmitted. The subset of prediction parameters may consist of one or more reference image indices, one or more components of a motion parameter vector, a reference index and one or more components of a motion parameter vector, and so on. The already transmitted subset of prediction or encoding parameters can be used to derive a (reduced) set of candidate sample sets. As an example, a difference measure can be calculated between the already encoded prediction or encoding parameters and the corresponding prediction or encoding parameters of an original set of candidate sample sets.Only those sample sets for which the calculated difference measure is less than or equal to a predefined or derived threshold are included in the final (reduced) set of candidate sample sets. The threshold can be derived based on the calculated difference measures. Alternatively, only those sample sets for which the difference measure is minimized can be selected. Or, a sample set can be selected based solely on the difference measure. In this latter case, the merge information can be reduced to specify only whether the current sample set is merged with the single candidate sample set. The following preferred realizations are described for sample sets representing rectangular and quadratic blocks, but can be extended to arbitrarily shaped regions or other sample collections in a simple manner. 1. Derivation of the initial set of candidate blocks The derivation of the initial sample set described in these sections refers to the derivation of an initial set of candidates. Some of the candidate blocks can be subsequently eliminated by analyzing their associated parameters (e.g., prediction information) and removing those candidate blocks for which a merge would result in a final partition that could also be obtained using other partitioning patterns. This procedure is described in the following subsection. In a preferred embodiment, the initial candidate block set is formed as follows. Starting from the upper-left sample position of the current block, its left neighbor sample position and its upper neighbor sample position are derived. The initial candidate block set can only have up to two elements, specifically those blocks from the causal block set that contain one of the two sample positions. Therefore, the initial candidate block set can only have the two blocks directly adjacent to the upper-left sample position of the current block as its elements. In another preferred embodiment of the invention, the initial candidate block set is given by all blocks that have been coded prior to the current block and contain one or more samples representing direct spatial neighbors (direct spatial neighbors may be restricted to direct left neighbors and / or direct upper neighbors and / or direct right neighbors and / or direct lower neighbors) of any sample in the current block. In another preferred embodiment of the invention, the initial candidate block set additionally (or exclusively) includes blocks containing one or more samples that are located in the same position as any of the samples in the current block but are contained in a different (already coded) image. In another preferred embodiment of the invention, the initial candidate block set represents a subset of the block (neighbor) sets described above.The subset of candidate blocks can be fixed, pointed, or derived. Deriving the subset of candidate blocks can take into account decisions made for other blocks in the image or in other images. For example, blocks that are associated with the same (or very similar) encoding parameters as other candidate blocks might not be included in the initial set of candidate blocks. In a preferred embodiment of the invention, the initial set of candidate blocks is derived as for one of the embodiments described above, but with the following restriction: Only blocks that use compensated motion prediction (inter-prediction) can be elements of the candidate block set. That is, intra-coded blocks are not included in the (initial) candidate set. As previously stated, it is possible to expand the list of candidates with additional candidates for block merging, such as by combining combined bipredictive merge candidates, unscaled bipredictive merge candidates, and a zero-move vector. The derivation of the initial set of candidate blocks is performed by both the encoder and the decoder in the same way. 2. Derivation of the final set of candidate blocks. After deriving the initial set of candidates, the associated parameters of the candidate blocks within the initial set are analyzed, and candidates for merging are eliminated if a merge would result in a partition that could be represented using a different partitioning pattern. If the sample arrays to be merged are of different shapes and / or sizes, identical partitionings may exist that can be described by at least two different codewords. For example, if the coder decides to split a sample array into two sample arrays, this split would be reversed by merging the two sample arrays. To avoid such redundant descriptions, the set of candidate blocks for merging is restricted based on the particular block shapes and splits that are allowed.On the other hand, the permissible shapes of sample arrays can be restricted depending on the specific candidate lists used for merging. The two splitting and merging facilities must be designed together so that, in combination, redundant descriptions are avoided. In a preferred embodiment of the invention, the set of division modes (or partitioning modes) depicted in Fig. 6 are supported for square blocks. If a square block of a particular size is divided into four smaller square blocks of the same size (lower left pattern in Fig. 6), the same set of partitioning patterns can be applied to the resulting four square blocks so that a hierarchical partitioning can be specified. After deriving the initial set of candidate blocks, the reduction of the candidate lists is performed as follows. If the current block is not partitioned further (top left pattern in Fig. 6), the list of initial candidates is not reduced. That is, all initial candidates represent the final candidates for the merge. If the current block is partitioned into exactly two blocks of arbitrary size, one of these two blocks is coded before the other, as determined by the syntax. For the first coded block, the initial candidate set is not reduced. However, for the second coded block, all candidate blocks with the same associated parameters as the first block are removed from the candidate set (this includes the first coded block). If a block is partitioned into four square blocks of the same size, the initial candidate list for the first three blocks (in coding order) is not reduced. All blocks in the initial candidate list are also present in the final candidate list. However, for the fourth (last) block in coding order, the following applies: If blocks that are in a different row (in the partitioning scheme as illustrated in the lower left of Fig. 6) than the current block have the same associated parameters (e.g., movement parameters), all candidates that have the same movement parameters as the block already coded in the same row as the current block are removed from the set of candidates (this includes the block in the same row). If blocks in a different column (in the partitioning scheme as illustrated in the lower left of Fig. 6) than the current block have the same associated parameters (e.g., movement parameters), all candidates with the same movement parameters as the block already coded in the same column as the current block are removed from the candidate set. (This includes the block in the same column.) In a low-complexity variation of the implementation (using the partitioning patterns in Fig. 6), the reduction of the candidate lists is performed as follows. If the current block is not partitioned further (top left pattern in Fig. 6), the list of initial candidates is not reduced. That is, all initial candidates represent the final candidates for the merge. If the current block is partitioned into exactly two blocks of arbitrary size, one of these two blocks is coded before the other, as determined by the syntax. For the first coded block, the initial candidate set is not reduced. However, for the second coded block, the first coded block in the partitioning pattern is removed from the candidate set. If a block is partitioned into four square blocks of the same size, the initial candidate list for the first three blocks (in coding order) is not reduced. All blocks in the initial candidate list are also present in the final candidate list. However, for the fourth (last) block in coding order, the following applies: If for the block in the other row (other than the current block) that is coded later, the merge information indicates that it merges with the first coded block in that row, the block in the same row as the current block is removed from the candidate set. If for the block in the other column (other than the current block) that is coded later, the merge information indicates that it merges with the first coded block in that column, the block in the same column as the current block is removed from the candidate set. In another preferred embodiment, the same partitioning patterns depicted in Fig. 6 are supported, but without the patterns that partition the square block into two rectangular blocks of equal size. Reduction of the candidate list proceeds as described in any of the embodiments described above, with the exception of the pattern that divides the block into four square blocks. Here, all initial candidates are allowed for all sub-blocks, or only the candidate list of the last coded sub-block is restricted as follows: If the three previously coded blocks are associated with the same parameters, all candidates associated with these parameters are removed from the candidate list. In a low-complexity version, the last coded sub-block cannot be merged with any of the three previously coded sub-blocks if these three sub-blocks have already been merged. In another preferred embodiment, a different set of partitioning patterns is supported for a block (or any other form of sample array set). For sample array sets that are not partitioned, all candidates from the initial candidate lists can be used for merging. If a sample array is partitioned into exactly two sample arrays, for the sample arrays that are first in coding order, all candidates from the initial candidate set are inserted into the final candidate set. For the second sample array in coding order, all candidates with the same associated parameters as the first sample array are removed. Or, in a low-complexity variation, only the first sample array is removed from the candidate set.For partitioning patterns that divide a sample array into more than two sample arrays, candidate elimination depends on whether another partitioning pattern can be simulated using the current partitioning pattern and the corresponding merge information. The candidate elimination procedure follows the concept explicitly described above, but considers only the actually supported candidate patterns. In another preferred embodiment, if the JUMP / DIRECT mode is supported for a particular block, merge candidates that are also present candidates for the JUMP / DIRECT modes are removed from the candidate list. This removal can replace the candidate block removals described above or be used in conjunction with the candidate block removals described above. 3. Combination with SKIP / DIRECT modes The JUMP / DIRECT modes can be supported for all blocks or only for a particular block size and / or block shape. A set of candidate blocks is used for the JUMP / DIRECT modes. The difference between JUMP and DIRECT is whether or not residual information is sent. It is inferred that the parameters (for example, for prediction) of JUMP and DIRECT are the same as any of the corresponding candidates. The candidate is chosen by passing an index to the candidate list. In a preferred embodiment, the candidate list for JUMP / DIRECT may contain different candidates. An example is illustrated in Fig. 8. The candidate list may include the following candidates (the current block is denoted by Xi): Median (between Left, Up, Corner) Left block (Li) Top block (Ai) Corner blocks (In order: Top right (Ci1) , Bottom left (Ci2) , Top left (Ci3) ) Block co-located in a different image, but already encoded In a preferred embodiment, the candidates for merge include Li (left block) and Ai (top block). Choosing these candidates for merge requires a small amount of related information to indicate which block the current block will merge with. The following notation is used to describe the following realizations: set_mvp_ori is a set of candidates used for JUMP / STRAIGHT mode. This set is composed of {Median, Left, Top, Corner, Colocated}, where Median is the median (middle value in an ordered set of Left, Top, and Corner), and colocated is given by the nearest reference frame and scales according to time distance. set_mvp_comb is a set of candidates used for JUMP / DIRECT mode in combination with the block merge procedure. In the preferred implementation, the combination of JUMP / DIRECT mode and block merge mode can be processed with the original set of candidates. This means that JUMP / DIRECT mode has the same set of candidates as when activated alone. The advantage of combining these two modes stems from their complementarity in signaling interframe inter-frame related information. Although both modes use neighbor information to enhance the signaling of the current block, block merge processes only left and above neighbors, while JUMP / DIRECT mode processes up to five candidates. The primary complementarity lies in their different neighbor information processing strategies. The block merge procedure maintains the full set of neighbor information for all reference lists.This means that block merging maintains the complete contiguous information of these neighbors, not just their movement vectors per reference list, whereas JUMP / STRAIGHT mode processes prediction parameters separately for each reference list and transmits an index to a candidate list for each reference list. That is, for bipredicted images, two indices are transmitted to point to a candidate for reference list 0 and a candidate for reference list 1. In another preferred embodiment, a combined set of candidates, called set_mvp_comb, can be found for the JUMP / DIRECT mode in combination with the block merge mode. This combined set is part of the original set (set_mvp_ori) and allows for a reduction in signaling for the JUMP / DIRECT mode, due to the reduction of the candidate list: set_mvp_comb. The candidates that should be removed from the original list (set_mvp_ori) are those that might be redundant with the block merge procedure or that are not frequently used. In another preferred embodiment, the combination of the JUMP / DIRECT mode and the block merge procedure can be processed with the combined set of candidates (set_mvp_comb), which is the original set (set_mvp_ori) without the median. Due to the low efficiency observed for the median in the JUMP / DIRECT mode, reducing it from the original list improves coding efficiency. In another preferred embodiment, the combination of JUMP / DIRECT mode and block merging can be processed with the combined set of candidates (set_mvp_comb), which is the original set (set_mvp_ori) with only the Corner and / or Colocated as candidates. In another preferred embodiment, the combination of the JUMP / STRAIGHT mode and the block merging procedure can be processed with the combined set of candidates, set_mvp_ori, with only Corner and Co-located as candidates. Despite the complementarity between the JUMP / STRAIGHT mode and block merging, as already mentioned, the candidates that should be removed from the list are those that could be redundant with the candidates of the block merging procedure. These candidates are Left and Top. The combined set of candidates (set_mvp_comb) has been reduced to only two candidates: Corner and Co-located. The JUMP / STRAIGHT mode using this set of candidates, set_mvp_comb, combined with the block merging process, provides a significant increase in the efficiency of secondary information signaling between frames. In this embodiment, the JUMP / STRAIGHT mode and the merging mode do not share any candidate blocks. In further implementations, a slightly different combination of JUMP / DIRECT and merge modes can be used. It is possible to enable JUMP / DIRECT mode (for example, with more candidates than the merge mode) only for particular blocks (for example, blocks larger than a specified size, or only for square blocks, etc.) and not support the merge mode for these blocks. Alternatively, JUMP / DIRECT mode can be removed, and all candidates (including parameters representing a combination of parameters for spatially / temporally neighboring blocks) are added to the merge mode as candidates. This option is described in Figs. 1 to 5. The augmented candidate set could only be used for particular blocks (larger than a given minimum size, or square blocks, etc.), whereas a reduced candidate set is used for other blocks. Or, as a further variant, merge mode is used with a reduced candidate set (e.g., only the top and left neighbors), and additional candidates (e.g., the top-left neighbor, the co-located block, etc.) are used for JUMP / DIRECT mode. Also in such configurations, JUMP / DIRECT modes can only be allowed for particular blocks (larger than a given minimum size, or square blocks, etc.), while merge mode is allowed for a larger set of blocks. 4. Transmission of merger information For the preferred embodiment, and in particular for the embodiments in Figures 1 to 5, the following can be applied. Imagine that only the two blocks containing the left and upper neighbor samples of the upper-left sample of the current blocks are considered candidates. If the final candidate block set (after candidate elimination as described above) is not empty, a flag called the merge flag is set, specifying whether the current block is merged with any of the candidate blocks. If merge_flag is 0 (for "false"), this block is not merged with one of its candidate blocks, and all encoding parameters are transmitted normally. If merge_flag is 1 (for "true"), the following applies. If the candidate block set contains one and only one block, this candidate block is used for merging. Otherwise, the candidate block set contains exactly two blocks.If the prediction parameters of these two blocks are identical, these prediction parameters are used for the current block. Otherwise (the two blocks have different prediction parameters), a flag called merge_left_flag is set. If merge_left_flag is 1 (for "true"), the block containing the left neighbor sample position of the current block's upper-left sample position is selected from the set of candidate blocks. If merge_left_flag is 0 (for "false"), the other block (i.e., the upper neighbor) is selected from the set of candidate blocks. The prediction parameters of the selected block are used for the current block. In another implementation, a merged syntax element signaling the merge procedure is passed. In yet another implementation, merge_left_flag is passed regardless of whether the two candidate blocks have the same prediction parameters. It should be noted that the syntax element merge_ left_flag could also be called merge_index since its function is to index the chosen flag among the non-eliminated candidates. In another preferred embodiment, more than two blocks can be included in the candidate block set. Merge information (i.e., whether a block is merged and, if so, with which candidate block it is merged) is signaled by one or more syntax elements. In this case, the codeword set depends on the number of candidates in the final candidate set and is selected in the same way in the encoder and decoder. In one embodiment, merge information is transmitted using a syntax element. In another embodiment, a syntax element specifies whether the block is merged with any of the candidate blocks (see the merge flag described above). This flag is transmitted only if the candidate block set is not empty.The second syntax element indicates which of the candidate blocks is used for merging; it is only transmitted if the first syntax element indicates that the current block is being merged with one of the candidate blocks. In a preferred embodiment of the invention, the second syntax element is only transmitted if the set of candidate blocks contains more than one candidate block and / or if any of the candidate blocks has prediction parameters different from any other of the candidate blocks. The syntax may depend on how many candidate blocks are given and / or how different prediction parameters are associated with the candidate blocks. It is possible to add a set of candidates for block merging as was done for DIRECT mode. As described in other preferred implementations, the second syntax element, merge index, can only be transmitted if the candidate list contains more than one candidate. This requires deriving the list before parsing merge index, preventing these two procedures from running in parallel. To allow for higher parsing performance and make the parsing procedure more robust against transmission errors, this dependency can be eliminated by using a fixed codeword for each index value and a fixed number of candidates. If this number cannot be reached by a selection of candidates, auxiliary candidates can be derived to complete the list. These additional candidates can include so-called combined candidates, which are constructed from move parameters of possibly different candidates already in the list, and zero-move vectors. In another preferred embodiment, the syntax indicates which blocks in the candidate set can be fitted simultaneously in the encoder and decoder. If, for example, three block options are given for merging, only those three options are present in the syntax and are considered for entropy coding. The probabilities for all other options are considered to be zero, and the entropy codec is fitted simultaneously in the encoder and decoder. The prediction parameters inferred as a result of the merging procedure can represent the entire set of prediction parameters that are associated with a block, or they can represent a subset of these prediction parameters (e.g., the prediction parameters for a hypothesis of a block for which multi-hypothesis prediction is used). In a preferred embodiment, the syntax elements related to merge information are entropy-encoded using context modeling. These syntax elements may consist of the `merge_flag` and `merge_left_flag` elements described above. In a preferred embodiment, one out of every three context models is used to encode the merge_flag. The context model used, merge_flag_ctx, is derived as follows. If the set of candidate blocks contains two elements, the value of merge_flag_ctx is equal to the sum of the merge_flag values ​​of the two candidate blocks. If the set of candidate blocks contains one element, the value of merge_flag_ctx is equal to twice the merge_flag value of that one candidate block. In a preferred embodiment, merge_left_flag is encoded using a single probability model. Different context model encoding can be used for merge_idx (merge_left_flag). In other embodiments, different context models could be used. Non-binary syntax elements can be mapped to a sequence of binary symbols (bins). The context models for some syntax elements or bins of syntax elements can be derived based on syntax elements already passed from neighboring blocks, the number of candidate blocks, or other measures, while other syntax elements or bins of syntax elements can be encoded with a fixed context model. 5. Encoder Operation The inclusion of the fusion concept provides an encoder with greater freedom in creating a bitstream, as the fusion strategy significantly increases the number of possible partitions for the sample arrays of an image, albeit at a higher signaling overhead. Some or all of the additional patterns that can be represented by a combination of sub-partitioning and fusion (e.g., the patterns in Fig. 7, when the partitioning pattern in Fig. 6 is supported) can be further tested (using the corresponding block sizes for motion estimation and mode selection), and the best of the patterns provided by pure partitioning (Fig. 6) and by partitioning and fusion (Fig. 7) can be selected based on a particular velocity distortion measure.Furthermore, for each block, it can be tested whether a merge with any of the already coded candidate sets produces a decrease in a particular speed distortion measure, and then the corresponding merge flags are set during the coding procedure. In another preferred embodiment, the encoder could first determine the best subdivision of the sample arrays (as in state-of-the-art coding schemes). It could then check, for each sample set, whether merging it with another sample set or group of sample sets reduces a particular velocity distortion cost measure. In this case, the prediction parameters associated with the merged group of sample sets could be re-estimated (for example, by performing a new motion search), or the prediction parameters already determined for the current sample set and the candidate sample set (or group of sample sets) for merging could be evaluated for the group of sample sets under consideration. In another preferred embodiment, a particular velocity distortion cost measure could be evaluated for additional candidate groups of sample sets. As a particular example, when testing the various possible partitioning patterns (see Fig. 6, for example), part or all of the pattern that can be represented by a combination of partitioning and merging (see Fig. 7, for example) can be further tested. That is, for all patterns, a specific motion estimation and mode decision procedure is carried out, and the pattern that produces the smallest velocity distortion measure is selected. This procedure can also be combined with the low-complexity procedure described above, so that for the resulting blocks, it is further tested whether a merging with already encoded blocks (e.g., outside the patterns of Fig. 6 and Fig. 7) is feasible.7) produces a decrease in a measure of speed distortion. In another preferred embodiment, the encoder tests the different patterns that can be represented by partitioning and merging in a priority order and tests as many patterns as possible within a given real-time requirement. The priority order can also be modified based on the blocks already encoded and the partitioning patterns chosen. One way to transfer the realizations described above to a specific syntax is explained below with reference to the following figures. In particular, Figs. 9-11 show different parts of a syntax that takes advantage of the realizations described above. Specifically, according to the implementation described below, image 20 is first split into coding tree blocks whose image content is encoded using the coding_tree syntax shown in Fig. 9. As shown herein, for entropy_coding_mode_flag=1, which refers, for example, to context-adaptive binary arithmetic coding or another specific entropy coding mode, the quad-tree subdivision of the current coding tree block is signaled within the coding_tree syntax portion by the flags called split_coding_unit_flag at mark 400. As shown in Fig.9. According to the implementation described below, the root tree block is subdivided as indicated by `split_coding_unit_flag` in a first depth-through order, as shown in Fig. 9a. Each time a leaf node is reached, it represents a coding set that is immediately encoded using the syntax function `coding_unit`. This can be seen in Fig. 9 when observing the `if` clause in 402, which checks whether the current `split_coding_unit_flag` is set. If so, the `coding_tree` function is recursively called, leading to an additional `split_coding_unit_flag` being passed / pulled in the encoder and decoder, respectively. If this is not the case, i.e., if split_coding_unit_flag=0, the current sub-block of the tree root block 200 in Fig. 5a is a leaf block, and in order to encode this coding set, the coding_unit function in Fig. 10 is called on 404. In the currently described embodiment, the previously mentioned option is used, whereby fusion is only usable for images for which the inter-image prediction mode is available. That is, intra-image sectors / images do not use fusion anyway. This is visible in Fig. 10, where the merge_flag is transmitted on 406 simply in case a sector type is not the same as the intra-image sector type. According to the present embodiment, fusion refers only to the prediction parameters related to inter-image prediction. According to the present embodiment, the merge_flag is signaled for the entire coding set 40 and also signals the decoder a specific partitioning mode for the current coding set, namely, non-partitioning mode.Therefore, the prediction_unit function is called 408, indicating that the current encoding set is a prediction set. However, this is not the only way to activate the merging option. Rather, if the merge_flag related to the entire encoding set is not set to 406, the prediction type of the non-intraimage sector's encoding set is indicated in 410 by the pred_type syntax element. Depending on this, the prediction_unit function is called for any partition of the current encoding set in, for example, 412 if the current encoding set is not further partitioned. Figure 10 simply shows four different partitioning options, but the other partitioning options shown in Figure 6 may also be available. Another possibility is that the PART_NxN partitioning option is unavailable, but the others are.The association between the names for the partitioning modes used in Fig. 10 and the partitioning options shown in Fig. 6 is indicated in Fig. 6 by the respective subscripts below the individual partitioning options. The prediction_unit function is called for each partition, such as partitions 50 and 60 in the coding order mentioned above. The prediction_unit function begins by checking the merge_flag at 414. If the merge_flag is set, a merge_index inevitably follows at 416. The check at stage 414 is to verify whether or not the merge_flag related to the entire coding set, as signaled at 406, has been set. If not, a merge_flag is signaled again at 418, and if this is set, a merge_index follows at 420, indicating the merge candidate for the current partition.Again, merge_flag is signaled for the current partition in 418 simply in case the current prediction mode of the current encoding set is an inter prediction mode (see 422). As is visible in Fig. 11, the transmission of the prediction parameters in use for the current prediction set in 424 is carried out, according to the present embodiment, simply in case fusion is not used for the present prediction set. Although the above description of the realization of Figs. 9-11 already describes most of the functionality and semantics, some additional information is presented below. `merge_flag[x0][y0]` specifies whether the inter-prediction parameters for the current prediction set (see 50 and 60 in the figures) are inferred from a neighboring interpredicted partition. The array indices x0, y0 specify the location (x0, y0) of the upper-left luma sample of the prediction block under consideration (see 50 and 60 in the figures) with respect to the upper-left luma sample of the image (see 20 in the figures). `merge_idx[x0][y0]` specifies the merge candidate index from the merge candidate list where x0, y0 specify the location (x0, y0) of the upper-left luma sample of the prediction block under consideration with respect to the upper-left luma sample of the image. Although not specifically stated in the preceding description of Figs. 9–11, the merge candidates or the merge candidate list are determined in this embodiment not only by using encoding parameters or spatially neighboring prediction set / partition parameters, but also by forming a candidate list using prediction parameters from temporally neighboring partitions of temporally neighboring and previously encoded images. Furthermore, combinations of prediction parameters from spatially and / or temporally neighboring prediction set / partition parameters are used and included in the merge candidate list. Naturally, a subset of these parameters can also be used. In particular, Fig. 12 shows one way to determine spatial neighbors, i.e., spatially neighboring partitions or prediction sets.Figure 12 exemplifies a prediction set or partition 60 and the pixels B0 to B2 and A0 and A1 that are located directly adjacent to the 500 edge of partition 60, specifically B2 which is diagonally adjacent to the upper left pixel of partition 60, B1 which is located vertically above and adjacent to the upper right pixel of partition 60, B0 which is located diagonally to the upper right pixel of partition 60, A1 which is located horizontally to the left of and adjacent to the lower left pixel of partition 60, and A0 which is located diagonally to the lower left pixel of partition 60. A partition that includes at least one of the pixels B0 to B2 and A0 and A1 forms a spatial neighbor and the prediction parameters of the same form a merge candidate. In order to perform the aforementioned removal of candidates that would lead to another partitioning mode that would also have been available, the following functions could be used: In particular, candidate N, i.e., the encoding / prediction parameters derived from the prediction / partition set covering pixel N= (B0, B1, B2, A0, A1), i.e., position (xN, yN), is removed from the candidate list if any of the following conditions are true (see Fig. 6 for the PartMode partitioning mode and the corresponding PartIdx partitioning index that indexes the respective partition within the encoding set): The PartMode of the current prediction set is PART_2NxN and PartIdx is equal to 1, and the prediction sets covering luma location (xP, yP - 1) (Partldx = 0) and luma location (xN, yN) (Cand. N) have identical motion parameters: The PartMode of the current prediction set is PART_Nx2N and PartIdx is equal to 1, and the prediction sets covering luma location (xP - 1, yP) (Par- tIdx = 0) and luma location (xN, yN) (Cand. N) have identical motion parameters: The PartMode of the current prediction set is PART_NxN and PartIdx is equal to 3, and the prediction sets covering luma location (xP - 1, yP) (PartIdx = 2) and luma location (xP - 1, yP - 1) (PartIdx = 0) have identical motion parameters: and the prediction sets covering the luma location (xP, yP - 1) (PartIdx = 1) and the luma location (xN, yN) (Cand. N) have identical motion parameters: .- The PartMode of the current prediction set is PART_NxN and PartIdx is equal to 3 and the prediction sets covering the luma location (xP, yP - 1) (PartIdx = 1) and the luma location (xP - 1, yP - 1) (PartIdx = 0) have identical movement parameters: and the prediction sets covering the luma location (xP - 1, yP) (PartIdx = 2) and the luma location (xN, yN) (Cand. N) have identical motion parameters: In this regard, note that the position or location (xP, yP) denotes the highest pixel in the current prediction set / partition. That is, according to the first element, all the coding parameter candidates derived by directly adopting the respective coding parameters from neighboring prediction sets, specifically prediction set N, are verified. However, the other additional coding parameter candidates can be verified in the same way to see if they are equal to the coding parameters of the respective prediction set that arises from obtaining another partitioning pattern also supported by the syntax.According to the realizations just described, the equality of the encoding parameters comprises a check of the equality of the movement vector, i.e., mvLX, the reference index, i.e., reflxLX, and the prediction flag predFlagLX indicating that the parameters, i.e., the movement vector and the reference index, associated with the reference list X, where X is 0 or 1, are used in the inter prediction. Note that the method just described for eliminating candidate coding parameters from neighboring prediction sets / partitions would also apply if the asymmetric partitioning modes shown in the right half of Fig. 6 were supported. In that case, the PART_2NxN mode could represent all horizontal subdivision modes, and PART_Nx2N could correspond to all vertical subdivision modes. Furthermore, the PART_2NxN mode could be excluded from the supported partitioning modes or patterns, in which case only the first two elimination checks would need to be performed. With regard to the realization of Figs. 9-12, it should also be noted that it is possible to exclude intrapredicted partitions from the candidate list, i.e., their encoding parameters are naturally not included in the candidate list. Furthermore, it is observed that three contexts could be used for the merge_flag and the merge_index. While some aspects have been described in the context of a device, it is clear that these aspects also represent a description of the corresponding method, in which a block or device corresponds to a step of the method or a feature of a step of the method. Similarly, the aspects described in the context of a step of the method also represent a description of a corresponding block or element or a feature of a corresponding device. Some or all of the steps of the method may be executed by means of (or using) a hardware device, such as a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, one or more of the most important methodological steps may be executed by such a device. Depending on certain implementation requirements, the embodiments of the invention can be implemented in hardware or software. Implementation can be carried out using a digital storage medium, such as a floppy disk, DVD, Blu-ray disc, CD, ROM, PROM, EPROM, EEPROM, or flash memory, which has electronically readable control signals stored therein. These signals cooperate (or are capable of cooperating) with a programmable computer system to execute the respective method. Therefore, the digital storage medium is computer-readable. Some embodiments according to the invention comprise a non-transient data carrier comprising electronically readable control signals, capable of cooperating with a programmable computer system in such a way as to execute one of the methods described herein. In general, the embodiments of the present invention can be implemented in the form of a computer program product with program code, where the program code fulfills the function of executing one of the methods when the computer program is run on a computer. The program code can be stored, for example, on a machine-readable medium. Other embodiments include the computer program for executing one of the methods described herein, stored on a machine-readable carrier. In other words, an embodiment of the inventive method is, therefore, a computer program that has program code to perform one of the methods described herein, when the computer program is executed on a computer. A further embodiment of the inventive methods is, therefore, a data carrier (or a digital storage medium, or a computer-readable medium) comprising, recorded thereon, the computer program for carrying out one of the methods described herein. The data carrier, the digital storage medium, or the recorded medium is typically tangible and / or non-transient. Another embodiment of the inventive method is, therefore, a data stream or a sequence of signals representing the computer program for carrying out one of the methods described herein. The data stream or the sequence of signals may be configured, for example, to be transmitted over a data communication connection, such as the Internet. Another embodiment comprises a processing means, for example a computer or a programmable logic device, configured or adapted to perform one of the methods described herein. Another embodiment comprises a computer that has installed on it the software program to perform one of the methods described in this document. Another embodiment according to the invention comprises an apparatus or system configured to transfer (e.g., electronically or optically) a computer program for performing one of the methods described herein to a receiver. The receiver may be, for example, a computer, a mobile device, a memory device, or the like. The apparatus or system may, for example, comprise a file server for transferring the computer program to the receiver. In some embodiments, a programmable logic device (e.g., a field-programmable gate array) can be used to perform some or all of the functionalities of the methods described herein. In some embodiments, a field-programmable gate array can cooperate with a microprocessor to perform one of the methods described herein. Generally, the methods are preferably executed by any hardware device. The embodiments described above are merely illustrative of the principles of the present invention. It is understood that modifications and variations of the arrangements and details described herein will be obvious to those skilled in the art. The scope of protection is defined by the appended claims. Therefore, the intention is to be limited only by the scope of the imminent patent claims and not by the specific details presented herein for the purpose of describing and explaining the embodiments.

Claims

1. Decoder (80) configured to decode a bitstream (30) pointing to one of the supported partitioning patterns for a current block (40) of an image (20), the decoder being configured to determine, if the pointing of the supported partitioning patterns specifies a subdivision of the current block (40) into two rectangular block partitions (50, 60), for each of the block partitions except a first block partition of the block partitions of the current block in an encoding order (70), determine, for the respective block partition (60) of the current block, a set of candidate encoding parameters,wherein at least some candidate encoding parameters are adopted from the encoding parameters of simply a previously decoded block partition such that at least some candidate encoding parameters are equal to the encoding parameters of simply a previously decoded block partition, with the determination being made in such a way that candidate encoding parameters equal to the encoding parameters associated with any of the block partitions, which, when merged with the respective block partition (60), would result in one of the supported partition patterns, are excluded from the set of candidate encoding parameters for the respective block partition (60),characterized in that the determination of the set of candidate encoding parameters for the respective block partition (60) of the current block further comprises deriving at least some additional candidate encoding parameters from a combination of encoding parameters from more than one previously decoded block partition, or from—by modification—encoding parameters of a previously decoded partition.

2. Decoder according to claim 1, wherein the decoder (80) is configured to set, if a number of the non-excluded candidate encoding parameters is non-zero, the encoding parameters associated with the respective block partition (60) depending on one of the non-excluded candidate encoding parameters.

3. Decoder according to claim 1, wherein the decoder (80) is configured to set,If a number of the non-excluded encoding parameter candidates is non-zero, the encoding parameters associated with the respective block partition are equal to one of the non-excluded encoding parameter candidates.

4. Decoder according to any of claims 2 to 3, wherein the decoder is configured to support intra- and inter-prediction modes for the current block and to perform the merging and determination of the set of encoding parameter candidates in such a way that, from the set of encoding parameter candidates for the respective block partition (60), the encoding parameter candidates that match the encoding parameters associated with any of the block partitions, which, when fused with the respective block partition (60), would result in one of the supported partition patterns,are excluded from the set of candidate encoding parameters for the respective block partition (60) simply in the case that the current block (40) is encoded in inter-prediction mode.

5. Decoder according to any of claims 2 to 4, wherein the encoding parameters are prediction parameters and the decoder is configured to use the prediction parameters of the respective block partition (60) in order to derive a prediction signal for the respective block partition (60).

6. Decoder according to any of claims 1 to 5, wherein the decoder is configured to, depending on a number of non-excluded candidate encoding parameters for a respective block partition, simply expect the bitstream (30) to comprise a syntax element specifying which of the non-excluded candidate encoding parameters is employed for fusion.

7. Decoder according to any one of claims 1 to 6, wherein the decoder (80) is configured to determine the set of candidate encoding parameters for the respective block partition based, at least partially, on the encoding parameters associated with previously decoded block partitions neighboring the respective block partition, and located outside and inside the current block, respectively.

8. Decoder according to any one of claims 1 to 7, wherein the decoder (80) is configured to determine the set of candidate encoding parameters for the respective block partition from an initial set of previously decoded blocks.excluding those encoded in an intra-prediction mode.

9. Decoder according to any one of claims 1 to 8, wherein the decoder (80) is configured to subdivide the image (20) into encoding blocks according to the subdivision information contained in the bitstream (30), wherein the encoding blocks include the current block.

10. Decoder according to claim 9, wherein the decoder (80) is configured to further subdivide the current block (40) into one or more transformation blocks according to the additional subdivision information contained in the bitstream and to derive a residual signal from the current block (40) of the bitstream (30) into sets of the transformation blocks.

11. Decoder according to any one of claims 1 to 10,wherein the image has associated with it a depth map as additional information.

12. Encoder (10) configured to encode an image (20) into a bitstream (30), the encoder is configured to signal within a bitstream (30) one of the supported partitioning patterns for a current block (40); and if the signaling of the supported partitioning patterns specifies a subdivision of the current block (40) into two rectangular block partitions (50, 60), for each of the block partitions except a first block partition of the block partitions of the current block in an encoding order (70), determine, for the respective block partition (60) of the current block, a set of candidate encoding parameters,where at least some candidate encoding parameters are adopted from the encoding parameters of simply a previously encoded block partition such that at least some candidate encoding parameters are equal to the encoding parameters of said simply a previously encoded block partition, with the determination being made in such a way that candidate encoding parameters that equal the encoding parameters associated with any of the block partitions, which, when merged with the respective block partition, would result in one of the supported partition patterns, are excluded from the set of candidate encoding parameters for the respective block partition (60),characterized in that the determination of the set of candidate encoding parameters for the respective block partition (60) of the current block further comprises deriving at least some additional candidate encoding parameters from a combination of encoding parameters from more than one previously encoded block partition, or from—by modification—encoding parameters of a previously encoded partition.

13. Encoder according to claim 12, wherein the image has associated with it a depth map as additional information.

14. Method for decoding a bitstream (30) that points to one of the supported partition patterns for a current block (40) of an image (20), the method comprising if the pointing to the supported partition patterns specifies a subdivision of the current block (40) into two rectangular block partitions (50, 60),for each of the block partitions except a first block partition of the block partitions of the current block in an encoding order (70), determine, for the respective block partition (60) of the current block, a set of encoding parameter candidates, wherein at least some encoding parameter candidates are adopted from the encoding parameters of simply a previously decoded block partition such that at least some encoding parameter candidates are equal to the encoding parameters of simply a previously decoded block partition, with performing the determination in such a way that encoding parameter candidates equal the encoding parameters associated with any of the block partitions, which, when merged with the respective block partition (60), would result in one of the supported partition patterns,are excluded from the set of candidate encoding parameters for the respective block partition (60), characterized in that the determination of the set of candidate encoding parameters for the respective block partition (60) of the current block further comprises deriving at least some additional candidate encoding parameters from a combination of encoding parameters from more than one previously decoded block partition, or from—by modification—encoding parameters from a previously decoded partition.

15. Method for encoding an image (20) in a bitstream (30), the method comprises signaling within a bitstream (30) one of the supported partitioning patterns for a current block (40); and if the signaling of the supported partitioning patterns specifies a subdivision of the current block (40) into two rectangular block partitions (50, 60),for each of the block partitions except a first block partition of the block partitions of the current block in an encoding order (70), determine, for the respective block partition (60) of the current block, a set of encoding parameter candidates, wherein at least some encoding parameter candidates are adopted from the encoding parameters of simply a previously encoded block partition such that at least some encoding parameter candidates are equal to the encoding parameters of simply a previously encoded block partition, with performing the determination in such a way that encoding parameter candidates equal the encoding parameters associated with any of the block partitions, which, when merged with the respective block partition (60), would result in one of the supported partition patterns,are excluded from the set of candidate encoding parameters for the respective block partition (60), characterized in that the determination of the set of candidate encoding parameters for the respective block partition (60) of the current block further comprises deriving at least some additional candidate encoding parameters from a combination of encoding parameters from more than one previously decoded block partition, or from – by modification – encoding parameters of a previously encoded partition.

16. A computer program having program code for performing, when executed on a computer, a procedure according to claim 14 or 15.

Images