KR20120004319A - Apparatus and method for performing parallel entropy encoding and decoding - Google Patents

Abstract

PURPOSE: An apparatus for performing entropy encoding and decoding in parallel is provided to perform entropy encoding and decoding for the selection and update of a context model and for each bin encoding and decoding. CONSTITUTION: An information generator(610) creates entropy encoding slice information. A plurality of binarization units(620) binarizes a plurality of entropy encoding slices. A plurality of context modeling units(630) creates a coding bins. A plurality of bin encoder determination units(640) assigns the coding bins to bin encoders. A plurality of bin encoders(660) creates code words through a bin encoding of the coding bins.

Description

Apparatus and method for performing entropy encoding and decoding in parallel {APPARATUS AND METHOD FOR PERFORMING PARALLEL ENTROPY ENCODING AND DECODING}

The present invention relates to a technique for entropy encoding and entropy decoding of image data such as moving pictures.

Context-Adaptive Binary Arithmetic Coding (CABAC), which is used as entropy coding in video encoding and decoding, uses entropy encoding and encoding by using encoding information of neighboring blocks and occurrence probability of bins encoded or decoded in a previous step. Perform decryption.

In this case, the encoding apparatus selects a context model using encoding information of the neighboring block and information about the bin encoded in the previous step, predicts the occurrence probability of the bin according to the selected model, and performs the encoding of the next bin. Entropy encoding is performed by updating the encoded bean information for the context model of.

However, the context model processing method and the single bin encoder having the serialized characteristics of this method increase the amount of work required when encoding a high resolution image. Accordingly, there is a need for an entropy encoder and decoder that can efficiently process through parallelism of workload.

The present invention provides an encoding apparatus and method, and a decoding apparatus and method capable of performing entropy encoding and decoding in parallel in encoding and decoding an image.

In addition, the present invention can perform entropy encoding in parallel by adjusting the number of probability intervals, the size of the probability interval, and the representative probability based on any one of a frequency of occurrence of a bin, an encoding efficiency, and the number of bin decoders. A coding apparatus and method, and a decoding apparatus and method capable of performing entropy decoding in parallel are provided.

An encoding apparatus according to an embodiment of the present invention includes an information generation unit for dividing image data into a plurality of entropy coded slices and generating entropy coded slice information including a unit of dividing the image data into entropy coded slices. A plurality of binarization units for binarizing a plurality of entropy encoding slices and converting them into a bin sequence, and generating a plurality of coding bins by using bins constituting the bin column and Less Probable Bins (LPBs) of bins A context modeler of a plurality of bin encoders for allocating coding bins to a bin encoder based on LPB probabilities of the bins constituting the bin column, and a plurality of individual elements located before the bin encoder and storing the coding bins And a plurality of bin encoders that bin encode the coding bins to generate codewords.

The contextual modeling unit may determine LPB and LPB probabilities of the bins based on the context information of the bin and the location of the bin, respectively. Then, the plurality of bin encoder determiners each determine a probability section to which the LPB probability of the bins belong among a plurality of probability sections, and allocate the coding bins to a plurality of bin encoders corresponding to each of the determined probability sections, respectively. Can be.

The contextual modeling unit may update the LPBs of the bins and the LPB probabilities of the bins.

In addition, the plurality of individual buffers may have the same number as the plurality of empty encoders.

The apparatus may further include a plurality of partial bitstreams generated by collecting the codewords, and a multiplexer which multiplexes the entropy coded slice information.

The encoding method according to an embodiment of the present invention includes generating entropy coded slice information including a unit for dividing image data into entropy coded slices, and binarizing each of the plurality of entropy coded slices. Converting to, generating the coding bins in parallel using the Bean Probable Bins (LPBs) of the bins and bins constituting the empty column, based on the LPB probability of the bins constituting the empty column. Assigning coding bins to a bin encoder, storing the coding bins in parallel, and bin encoding the coding bins to generate codewords in parallel.

The step of assigning the coding bins to the bin encoder may include determining a probability section to which the LPB probability of the bins belongs among a plurality of probability sections, and a plurality of coding bins corresponding to each of the determined probability sections. And assigning each to an empty encoder.

The method may further include updating the LPB of the bins and the LPB probability of the bins.

The method may further include generating the partial bitstream in parallel by collecting the codewords, and multiplexing the parallel generated bitstream and the entropy encoding slice information.

The method may further include dividing the image data into a plurality of entropy encoding slices by grouping syntax elements constituting the image data or blocks constituting the image data.

A decoding apparatus according to an embodiment of the present invention includes a demultiplexer for demultiplexing a bitstream to separate a plurality of partial bitstreams, and a plurality of bins for bin decoding the plurality of partial bitstreams to generate a plurality of coding bins. A decoder determines a bin decoder using LPB probabilities of bins, a plurality of bin decoder determiners that receive the coding bins from the determined bin decoder, and generates empty columns by using the LPBs of the coding bins and coding bins. A plurality of context modeling units and a plurality of inverse binarization units for inverse binarizing bins constituting the empty columns to restore a plurality of entropy encoding slices.

The apparatus may further include a plurality of individual buffers positioned after the plurality of bin decoders and storing coding bins output from the plurality of bin decoders.

In addition, the decoding method according to an embodiment of the present invention, demultiplexing the bitstream to separate a plurality of partial bitstream, bin decoding the plurality of partial bitstream to generate a plurality of coding bin, bin Determining a bin decoder using an LPB probability of H and delivering a coding bin, storing the coding bins output from the bin decoder, generating an empty column using the LPBs of the coding bins and coding bins, and Inverse binarization of the bins constituting the empty column may restore the plurality of entropy coded slices.

The generating of the empty column may include determining the LPB of the bin and the LPB probability of the bin based on the context information of the bin and the location of the bin, for the bins constituting the empty column. Updating the LPB probabilities of the LPB and bins.

The method may further include reconstructing the image data by sorting the plurality of entropy coded slices according to entropy coded slice information separated from the bitstream.

According to the present invention, entropy encoding / decoding can be performed in parallel for binarization / debinarization, context model selection and updating, and bin encoding / decoding, respectively, as a plurality of slices are encoded in a plurality of bin encoders / bin decoders, respectively. have.

According to the present invention, the number of probability intervals, the size of the probability interval, and the representative probability may be adjusted based on any one of a frequency of occurrence of a bin, an encoding efficiency, and a number of bin decoders.

1 is a flowchart provided to explain an operation of performing entropy encoding in parallel according to an embodiment of the present invention.
2 is a flowchart provided to explain a detailed operation of performing entropy encoding in parallel according to an embodiment of the present invention.
3 is a block diagram illustrating a configuration of an encoding apparatus according to an embodiment of the present invention.
4 is a flowchart provided to explain a detailed operation of performing entropy decoding in parallel according to an embodiment of the present invention.
5 is a block diagram showing a configuration of a decoding apparatus according to an embodiment of the present invention.
6 is a block diagram illustrating a configuration of an encoding apparatus including individual buffers according to an embodiment of the present invention.
7 is a block diagram illustrating a configuration of a decoding apparatus including individual buffers according to an embodiment of the present invention.
8 is a block diagram illustrating a configuration of an encoding apparatus including a probability interval allocator according to an embodiment of the present invention.
9 is a block diagram illustrating a configuration of a decoding apparatus including a probability interval allocator according to an embodiment of the present invention.
FIG. 10 is a block diagram illustrating a configuration of an encoding apparatus including an individual buffer and a probability interval allocator according to an embodiment of the present invention.
11 is a block diagram showing a configuration of a decoding apparatus including an individual buffer and a probability interval allocator according to an embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings will be described embodiments of the present invention; However, the present invention is not limited or limited by the embodiments. Also, like reference numerals in the drawings denote like elements.

1 is a flowchart provided to explain an operation of performing entropy encoding in parallel according to an embodiment of the present invention.

First, in step 110, the encoding apparatus may generate entropy encoding slice information for processing entropy encoding in parallel.

For example, referring to FIG. 12, the encoding apparatus may classify image data into a plurality of syntax elements. The encoding apparatus may generate, as entropy encoded slice information, information indicating that an entropy encoded slice is formed by grouping a plurality of syntax elements. Accordingly, the entropy encoding slice may perform a parsing process such as selecting and updating a context model beyond a slice boundary. Of course, an entropy encoded slice cannot perform a process related to a decoding process beyond a slice boundary like a conventional slice. However, the existing slice may not perform context model selection and update beyond the slice boundary. Here, the process related to the decoding process may include intra prediction, motion vector prediction, and the like.

As another example, referring to FIG. 13, the encoding apparatus may divide image data into a plurality of blocks and group the plurality of groups to generate information indicating that an entropy encoding slice is configured as entropy encoding slice information.

In operation 120, the encoding apparatus may process the binarization and the context model selection of the binarized syntax element in parallel with respect to the value of the syntax element to be encoded according to the entropy encoding slice information. For example, when an empty column is used as a binarized syntax element, the encoding apparatus may process a context model selection of the empty column in parallel. In other words, the encoding apparatus may perform binarization and context model selection in units of entropy encoding slices.

In operation 130, the encoding apparatus may output values of syntax elements to be encoded through a plurality of bin encoders in a bitstream. In this case, the encoding apparatus may mix the syntax elements and the entropy encoding slice information and output the bitstream.

2 is a flowchart provided to explain a detailed operation of performing entropy encoding in parallel according to an embodiment of the present invention.

First, in step 210, the encoding apparatus may divide image data into a plurality of entropy encoding slices.

For example, the encoding apparatus may divide the image data into a plurality of entropy encoding slices by grouping the image data into a plurality of syntax elements or a plurality of blocks.

In operation 220, the encoding apparatus may generate entropy encoded slice information including a unit that divides image data into entropy encoded slice units.

In operation 230, the encoding apparatus may convert a plurality of entropy encoding slices into bin sequences.

For example, when image data is divided into entropy coded slices in units of syntax elements, the encoding apparatus may convert syntax element values in the plurality of entropy coded slices into bin sequences. At this time, the empty column means a column composed of a plurality of bins.

In operation 240, the encoding apparatus may determine the bin LPB and the LPB probability of the bin based on the context information of the bin constituting the bin column and the position of the bin. Then, the encoding apparatus may update the LPB of the existing bin and the LPB probability of the existing bin to the LPB of the determined bin and the LPB probability of the determined bin. In this case, the encoding apparatus may update the LPB of the bin and the LPB probability of the bin according to the value of the bin.

For example, when the LPB of the bin and the LPB probability of the bin are determined, as the value of the bin and the LPB of the bin are the same, the LPB probability of the bin may increase. And, if the value of the bin and the LPB of the bin is not the same, the LPB probability of the bin can be reduced. In addition, when the LPB probability is greater than 50%, the LPB value may be changed from 0 to 1 or from 1 to 0.

In this case, the encoding apparatus may generate coding bins in parallel using bins constituting the empty column and LPBs of the bins.

As an example, the encoding apparatus may generate the coding bin 0 by performing an exclusive OR operation on the bin 0 constituting the empty column and the LPB of the bin 0. In the same manner, the encoding apparatus may generate a plurality of coding bins by performing an exclusive OR operation on all bins constituting the empty column and the LPBs of the bins.

In operation 250, the encoding apparatus may allocate the plurality of coding bins to the plurality of bin encoders based on the LPB probability of each of the plurality of coding bins.

For example, the encoding apparatus may determine a probability section to which the LPB probability of each coding bin belongs among the plurality of probability sections. In addition, the encoding apparatus may allocate each coding bin to a bin encoder having a representative probability that matches the representative probability of the determined probability interval. In this case, the encoding apparatus may determine the size and number of the probability intervals based on one of the probability distribution of the probability intervals to which the LPB probability belongs, the coding efficiency, and the number of empty decoders. Then, the encoding apparatus may determine a bin encoder matching the representative probability of the adjusted probability section according to the size and the number of the determined probability section, and allocate a coding bin to the determined bin encoder. That is, each probability interval has a representative probability, and the representative probability may correspond to one empty encoder.

In operation 260, the encoding apparatus may store the coding bins in parallel. That is, the encoding apparatus may store the coding bins in parallel using a plurality of individual buffers.

In operation 270, the encoding apparatus may bin encode the coding bins through the allocated bin encoder to generate codewords in parallel. That is, the empty encoder 0 to the empty encoder K-1 may generate a plurality of codewords 0 to a plurality of codewords K-1, respectively. In this case, the plurality of bin encoders may bin encode the coding bins stored in the corresponding individual buffers to generate codewords, respectively.

Subsequently, in step 280, the encoding apparatus may generate partial bitstreams in parallel by collecting codewords. For example, when there are K empty encoders, the encoding apparatus may generate partial bitstream 0 to partial bitstream K-1.

In operation 290, the encoding apparatus may generate one bitstream by multiplexing the partial bitstreams and the entropy encoding slice information. Then, the encoding apparatus may transmit the generated bitstream to the decoding apparatus.

In this case, when the decoding apparatus does not know the probability information of each of the partial bitstreams, the encoding apparatus may multiplex the probability information of each of the partial bitstreams with the partial bitstreams and the entropy encoding slice information. Here, the probability information of each of the partial bitstreams may include a representative probability of the empty encoder that outputs the partial bitstream.

3 is a block diagram illustrating a configuration of an encoding apparatus according to an embodiment of the present invention.

According to FIG. 3, the encoding apparatus 300 includes an information generator 310, a binarizer 320, a context modeler 330, an empty encoder determiner 340, an empty encoder 350, and a multiplexer 360. ) May be included. In FIG. 3, the binarizer 320, the context modeler 330, the bin encoder determiner 340, and the bin encoder 360 may be configured in plural. Accordingly, the encoding apparatus 300 may perform entropy encoding in parallel for each slice through a plurality of binarization units, a plurality of context modeling units, a plurality of bin encoder determiners, and a plurality of bin encoders.

First, the information generator 310 may divide the image data into a plurality of entropy coded slices, and generate entropy coded slice information including a unit that divides the image data into entropy coded slices. Accordingly, the encoding apparatus 300 may perform entropy encoding in parallel for each of the plurality of entropy encoding slices.

For example, the information generator 310 may group the plurality of syntax elements constituting the image data and divide the image data into a plurality of entropy encoding slices. The information generator 310 may generate information indicating that the entropy coded slice is configured by grouping a plurality of syntax elements as entropy coded slice information.

As another example, the information generator 310 may group the plurality of blocks constituting the image data and divide the image data into a plurality of entropy encoding slices. The information generator 310 may generate, as entropy coded slice information, information indicating that an entropy coded slice is formed by grouping a plurality of groups.

Then, the binarization unit 320 may binarize the plurality of entropy coded slices divided according to the entropy coded slice information in parallel. That is, the binarization unit 320 may be configured in plurality. Accordingly, the plurality of binarizers may output a bin sequence by binarizing the plurality of entropy encoding slices, respectively. In this case, the binarization unit 320 may binarize syntax element values in the plurality of entropy encoding slices into bin sequences. Here, the empty column may mean a column composed of a plurality of bins.

For example, when the binarization unit 320 is composed of M pieces, the binarization unit 320 may include a binarization unit 0, a binarization unit 1,... It may include a binarization unit M-1. Then, the binarization unit 0 may binarize the syntax element values in the entropy encoding slice 0 into an empty column 0. The binarization unit 1 may binarize the syntax element values in the entropy encoding slice 1 into an empty column 1. Similarly, the binarization unit M-1 may binarize the syntax element values in the entropy encoding slice M-1 to convert to the empty column M-1. As such, the binarization unit 320 may perform binarization for each entropy coded slice to convert syntax element values in the entropy coded slice into empty columns.

The context modeling unit 330 may determine the LPB (Less Probable Bin) and the LPB probability of the bin constituting the probability model based on the context information of the bin and the position of the bin. That is, the context modeling unit 330 may determine the LPB and the LPB probability of each of the plurality of bins constituting the empty column.

In this case, the context information of the bin refers to the bin in the neighboring block, and the location of the bin refers to the information of the bin encoded in the previous step.

In addition, the context modeling unit 330 may update the LPB of the bin and the LPB probability of the bin. In this case, the context modeling unit 330 may generate a coding bin by using the bin and the LPB of the bin, and output the generated coding bin and the LPB probability. For example, the context modeling unit 330 may generate a coding bean by performing an exclusive OR operation on the bin and the LPB of the bin. Here, the context modeling unit 330 may be configured in plurality.

For example, when the context modeling unit 330 includes M, the context modeling unit 0 may determine the LPB of the bin and the LPB probability of the bin for each bin constituting the empty column 0 output from the binarization unit 0. In this case, the context modeling unit 0 may determine the LPB of the bin and the LPB probability of the bin based on the context information of the bin constituting the empty column 0 and the position of the bin. In addition, the context modeling unit 0 may generate a coding bin by using the determined LPB and the bin, and output the generated coding bin and the LPB probability. That is, the context modeling unit 0 may generate and output coding bins (hereinafter, referred to as “coding bins 0”) corresponding to the bins constituting the blank column 0. For example, when empty column 0 is composed of bins 0 to 10, context modeling unit 0 is coded bin 0 of. , Coding 10 of bin 10 can be generated.

In the same way, the context modeling units 1 to M-1 may generate the coding bins 1 to coding bins M-1 for the empty columns 1 to M1 output from the corresponding binarization units 1 to M-1. . In addition, the context modeling unit 1 may output the LPB probability of each of the bins constituting the coding bins 1 and the blank column 1. Similarly, the context modeling unit M-1 may output the LPB probability of each of the bins constituting the coding bins M-1 and the empty column M-1.

The bin encoder determiner 340 may allocate a plurality of coding bins output from the context modeling unit 330 to the plurality of bin encoders, respectively.

For example, when the LPB probability of a bin is used, the bin encoder determiner 340 may determine a probability section to which the LPB probability of the bin belongs among a plurality of probability sections. The empty encoder determiner 340 may compare the representative probability of the determined probability interval with the representative probability of the empty encoder. In this case, when the representative probability of the probability interval coincides with the representative probability of the empty encoder, the empty encoder determiner 340 may allocate coding bins to the matching empty encoder. Here, a representative probability may be preset in the empty encoder, and the representative probability may have a +/- margin within a predefined error range.

The empty encoder determiner 340 may be configured in plural.

For example, when the bin encoder determiner 340 includes M, the bin encoder determiner 0 may allocate a plurality of coding bins output from the context modeler 0 to the plurality of bin encoders, respectively. In this case, the bin encoder determiner 0 may allocate the coding bins to the plurality of bin encoders based on the LPB probability of the bin corresponding to the coding bin. For example, when the coding bins output from the context modeling unit 0 are configured as coding bins 0 to 10, the bin encoder determining unit 0 is a bin whose representative probability matches the representative probability of the probability section to which the LPB probability of the bin 0 belongs. Encoder k-1 may be determined, and coding bin 0 may be assigned to the determined bin encoder k-1. Similarly, the bin encoder determiner 0 may assign the coding bin 1 to the coding bin 10 to the bin encoder 0 to the bin encoder k-1 based on the LPB probabilities of the bins 1 to 10, respectively.

In the same manner, the bin encoder determiner 1 to bin encoder determiner M-1 may allocate a plurality of coding bins output from each of the context modeling unit 1 to the context modeler M-1 to each of the plurality of encoders.

The bin encoder 350 may generate a codeword by bin encoding the coding bins allocated by the bin encoder determiner 340. The empty encoder may collect the generated codewords and output a partial bitstream. Here, the bin encoder 350 may be configured in plurality.

For example, when the empty encoder 350 is configured with K, the empty encoder 0 may include one or more coding bins allocated in the empty encoding allocator 0, one or more coding beans allocated in the empty encoding allocator 1, and empty encoding allocating units 2 through 2. One or more coding bins allocated in each of M-1 may be bin encoded to generate codewords. The empty encoder 0 may collect the generated codewords and output a partial bitstream 0.

In the same manner, each of the empty encoders 1 to empty encoder K-1 may bin encode one or more coding bins allocated in each of the empty encoding allocator 0 to the empty encoding allocator M-1 to generate codewords. The empty encoder 1 to the empty encoder K-1 may collect the generated codewords and output the partial bitstream 1 to the partial bitstream K-1, respectively.

The multiplexer 360 may multiplex the entropy encoding slice information and the partial bitstreams and output the multiplexed bitstreams as one bitstream. In this case, when the decoding apparatus does not know the probability information of each of the partial bitstreams, the multiplexer 360 multiplexes one bitstream by multiplexing the entropy encoding slice information, the partial bitstreams, and the probability information of each of the partial bitstreams. You can output Here, the probability information of each of the partial bitstreams may include a representative probability of the empty encoder that outputs the partial bitstream.

Hereinafter, an operation of performing entropy decoding in parallel will be described in detail. In this case, when decoding a specific syntax element, the inverse binarization unit may request the bins from the bin columns corresponding to the values of the syntax elements in the order of the bins to the bin decoder. Then, the context modeling unit may obtain the LPB probability of the coding bin to be decoded by using the context information of the requested bin and the probability model according to the location of each bin. As such, the decoding apparatus may have a structure of determining a probability section including the LPB probability using the bin decoder determiner and decoding the coding bin through a bin decoder corresponding to the representative probability of the determined probability section.

4 is a flowchart provided to explain a detailed operation of performing entropy decoding in parallel according to an embodiment of the present invention.

First, in operation 410, the decoding apparatus may demultiplex a bitstream received from an encoding apparatus to separate a plurality of partial bitstreams and entropy encoding slice information. In this case, when probability information is included in the bitstream, the decoding apparatus may separate probability information of each of the partial bitstreams from the bitstream. Here, the probability information may include the representative probability of the empty encoder that outputs the partial bitstream. The partial bitstream may consist of codewords generated through empty encoding in the empty encoder.

Here, the entropy coded slice information may include information indicating that an entropy coded slice is formed by grouping a plurality of syntax elements or an entropy coded slice is configured by grouping a plurality of blocks.

In operation 420, the decoding apparatus may bin-decode codewords in the plurality of partial bitstreams, respectively, to generate a plurality of coding bins.

In this case, the decoding apparatus may calculate 1) the LPB of the requested bin and 2) the LPB probability of the requested bin based on the context information of the requested bin and the location of the requested bin. Here, the LPB of the requested bin is the LPB of the coding bin, and the LPB probability of the requested bin is the LPB probability of the coding bin.

In operation 430, the decoding apparatus may determine a bin decoder corresponding to each of the plurality of bins using the LPB probability of the bin. In addition, the decoding apparatus may receive and transmit a coding bin from the determined bin decoder.

In one example, the decoding apparatus may determine a probability section to which the LPB probability of the bin belongs. The decoding apparatus may determine the bin decoder among the plurality of bin decoders by comparing the representative probability of the determined probability interval with the representative probability of the bin decoder. In addition, the decoding apparatus may receive and transmit a coding bin from each of the determined bin decoders. That is, the decoding apparatus may transmit the input coding bin to step 440 below.

In operation 440, the decoding apparatus may generate a bin sequence using the LPB of the coding bin and the coding bin. In one example, the decoding apparatus may restore the bin by performing an exclusive OR operation on the coding bin and the LPB of the coding bin. In addition, the decoding apparatus may generate an empty column composed of the restored bins.

In this case, the decoding apparatus may determine the LPB of the bin and the LPB probability of the bin based on the context information of the bin and the location of the bin, for the restored bins constituting the bin column. The decoding apparatus may update the LPB of the bin and the LPB probability of the bin.

In operation 450, the decoding apparatus may reconstruct bins constituting the plurality of empty columns to restore a plurality of entropy encoding slices.

As an example, the decoding apparatus may restore the values of syntax elements by inverse binarizing bins constituting the empty column, for a plurality of empty columns. The decoding apparatus may reconstruct the entropy encoding slice by collecting values of the recovered syntax elements. Accordingly, the decoding apparatus may restore the plurality of entropy coded slices through inverse binarization, respectively, for the plurality of empty columns.

In operation 460, the decoding apparatus may reconstruct image data by aligning the plurality of entropy coded slices reconstructed based on the entropy coded slice information.

Meanwhile, in step 420, the probability information may be loaded in the bitstream and transmitted from the encoding apparatus to the decoding apparatus, and the encoding apparatus and the decoding apparatus may know the probability information of each bin encoder / bin decoder in advance. As such, when the encoding and decoding apparatuses know the probability information in advance, the encoding apparatus and the decoding apparatus may know in advance how many empty encoders / bin decoders and representative probabilities of the empty encoders / bin decoders.

5 is a block diagram showing a configuration of a decoding apparatus according to an embodiment of the present invention.

Referring to FIG. 5, the decoding apparatus 500 includes a demultiplexer 510, a bin decoder 520, a bin decoder determiner 530, a context modeler 540, an inverse binarizer 550, and an alignment unit ( 560 may include. In FIG. 5, the bin decoder 520, the bin decoder determiner 530, the context modeler 540, and the inverse binarizer 550 may be configured in plural. Accordingly, the decoding apparatus 500 may reconstruct a plurality of slices by performing entropy decoding in parallel through a plurality of bin decoders, a plurality of bin decoder determiners, a plurality of context modelers, and a plurality of inverse binarization units. .

First, the demultiplexer 510 demultiplexes a bitstream received from the encoding apparatus 300 to separate a plurality of partial bitstreams, entropy coded slice information, and probability information of each of the partial bitstreams from the bitstream. Can be. Here, the probability information may include the representative probability of the empty encoder that outputs the partial bitstream. The partial bitstream may consist of codewords generated through empty encoding in the empty encoder.

The bin decoder 520 may then bin decode the codewords constituting the partial bitstream. As such, the bin decoder 520 may reconstruct a plurality of coding bins by bin decoding the codewords. In this case, the plurality of empty decoders 520 may be configured.

For example, when K bin decoders are configured with K, bin decoder 0 may bin-decode codewords constituting partial bitstream 0 to restore coding bins. In the same manner, the bin decoders 1 to bin decoder K-1 may reconstruct the coding bins by bin decoding the codewords constituting the partial bitstream 1 to the partial bitstream K-1.

The bin decoder determiner 530 may determine a probability section to which the LPB probability of the bin belongs based on the LPB probability of the bin. For example, when the bin binarizer 760 requests a bin, the context modeling unit 750 may calculate the LPB of the requested bin and the LPB probability of the requested bin. Then, the bin decoder determiner 530 may determine a bin decoder corresponding to the requested bin by using a probability section to which the LPB probability of the requested bin belongs. The bin decoder determiner 530 may receive a coding bin from each of the determined bin decoders and transmit the coded bin to the context modeling unit 540. Here, the LPB of the requested bin is the LPB of the coding bin, and the LPB probability of the requested bin means the LPB probability of the coding bin.

In detail, the bin decoder determiner 530 may determine a bin decoder having a representative probability of a probability section determined among the plurality of bin decoders. In this case, the bin decoder determiner 530 may compare the representative probabilities of the coding bins with the representative probabilities of the plurality of bin decoders, respectively, and determine a bin decoder having a representative probability that matches the representative probabilities of the coding bins for each coding bin. The bin decoder determiner 530 may receive coding bins output from the determined bin decoders. In this case, the bin decoder determiner 530 may be configured in plural.

As an example, the bin decoder determiner 0 may receive a coding bin output from each of the bin decoder 0 to the bin decoder K-1. In addition, the bin decoder determiner 0 may receive a coding bin from each of the bin decoder 0 to the bin decoder K-1 and transfer the coding bin to the context modeler 0. In the same way, the bin decoder determiner 1 to bin decoder M-1 may receive the coding bins output from the bin decoders 0 to bin decoder K-1, respectively. The bin decoder determiner 1 to bin decoder M-1 may deliver the input coding bins to the context modeler 0 to the context modeler M-1, respectively.

The context modeling unit 540 may generate an empty column using the coding bins input from the bean decoder determiner 530. For example, the context modeling unit 540 may restore the bin by performing an exclusive OR operation on the coding bin and the LPB of the coding bin. Then, the context modeling unit 540 may generate an empty column composed of the restored bins.

In addition, the context modeling unit 540 may update the LPB of the bin constituting the probability model and the LPB probability of the bin. In this case, the context modeling unit 540 may be configured in plurality.

For example, when coding bins 0 to 10 are input from the bin decoder determiner 0, the context modeling unit 0 may restore the bin 0 by performing an exclusive OR operation on the LPBs of the coding bin 0 and the coding bin 0. have. The context modeling unit 0 may restore the bin 1 by performing an exclusive OR operation on the LPBs of the coding bin 1 and the coding bin 1. Then, the context modeling unit 0 may determine the LPB of the bin 1 and the LPB probability of the bin 1 based on the context information of the restored bin 1 and the position of the restored bin 1. The context modeling unit 0 may update the LPB probability of the coding bin 1 and the LPB probability of the coding bin 1 to the LPB probability of the bin 1 and the LPB probability of the bin 1 determined using the restored bin 1. That is, the context modeler 0 may update the LPB of the bin requested by the inverse binarization unit and the LPB probability of the requested bin to the LPB of the determined bin and the LPB probability of the determined bin.

In the same way, the context modeling unit 0 may restore the bins 2 to 10 by performing an exclusive OR operation on the coding bins 2 to 10 and the LPBs of the coding bins 2 to 10. In addition, the context modeling unit 0 may update 1) the LPB of each of bins 2 to 10 and 2) the LPB probability of each of bins 2 to 10. Subsequently, the context modeling unit 0 may output a blank column 0 including the restored bins 0 to 10.

Similarly, the context modeling unit 1 to the context modeling unit M-1 may output the empty columns 1 to the empty column M-1 by performing an exclusive OR operation on the coding bin and the LPB of the coding bin. Here, the empty column 1 may include a plurality of bins restored by the context modeling unit 1, and the empty column M-1 may include a plurality of bins restored by the context modeling unit M-1. In this case, the context modeling unit 1 to the contextual modeling unit M-1 may update the LPB probability of the bins constituting the empty column 1, the LPB probability of the bins, and the LPB probability of the bins constituting the empty column M-1 and the LPB probability of the bins. .

The inverse binarization unit 550 may reconstruct an entropy encoding slice by inverse binarizing a plurality of bins constituting an empty column. In this case, the inverse binarization unit 550 may restore the values of syntax elements by inverse binarizing the plurality of bins constituting the bin column. The inverse binarization unit 550 may reconstruct an entropy encoding slice by collecting values of reconstructed syntax elements. Here, the reverse binarization unit 550 may be configured in plurality.

For example, when the inverse binarization unit includes M, the inverse binarization unit 0 may receive an empty column 0 from the context modeling unit 0. The inverse binarization unit 0 may restore the values of syntax elements by inverse binarizing the plurality of bins constituting the blank column 0. In this case, when the empty column 0 includes bins 0 to 10, the inverse binarization unit 0 may restore the values of syntax elements 0 to syntax elements 10, respectively. The inverse binarization unit 0 may recover the entropy encoding slice 0 by collecting values of syntax elements 0 to syntax elements 10.

In the same manner, the inverse binarization unit 1 to the inverse binarization unit M-1 may restore the values of syntax elements by inverse binarizing bins constituting the empty column 0 to the empty column M-1. The inverse binarization unit 1 to the inverse binarization unit M-1 may reconstruct the entropy coded slice 1 to entropy coded slice M-1 by collecting values of reconstructed syntax elements, respectively.

The alignment unit 560 may sort the plurality of entropy coded slices reconstructed through inverse binarization based on entropy coded information. For example, the alignment unit 560 may reconstruct image data by aligning entropy coded slices 0 to entropy coded slices M-1.

6 is a block diagram showing a configuration of an encoding apparatus according to an embodiment of the present invention.

FIG. 6 includes a plurality of individual buffers in the encoding apparatus of FIG. 3. Accordingly, description overlapping with FIG. 3 among the components of FIG. 6 will be omitted.

According to FIG. 6, the encoding apparatus 600 includes an information generator 610, a binarizer 620, a context modeler 630, an empty encoder determiner 640, an individual buffer 650, and an empty encoder 660. , And multiplexer 670. In FIG. 6, the binarization unit 620, the context modeling unit 630, the empty encoder determiner 640, the individual buffer 650, and the empty encoder 660 may be configured in plural numbers.

The individual buffer 650 has the same number as the empty encoder 660 and may be located before the empty encoder 660. As such, as located before the bin encoder, the individual buffer 650 may store the coding bins assigned to the bin encoder 660 among the coding bins output from the bin encoder determiner 640. That is, the individual buffer 650 may temporarily store the coding bins input to the bin encoder 660. In addition, the individual buffer 650 may deliver the stored coding bins to the assigned bin encoder.

For example, when the individual buffer 650 is composed of K, the individual buffer 0 is positioned before the empty encoder 0, and the coding bins allocated to the empty encoder 0 may be received and stored from the empty encoder determiners 0 to M-1. . And, individual buffer 0 can pass the stored coding bins to empty encoder 0. Similarly, the individual buffer 1 may receive and store the coding bins allocated to the empty encoder 1 from the empty encoder determiners 0 to M-1. And, individual buffer 1 can deliver the stored coding bins to empty encoder 1.

In the same way, the individual buffers 2 to K-1 may receive and store the coding bins assigned to each of the empty encoders 2 to empty encoder K-1. In addition, the individual buffers 2 to K-1 may transfer the stored coding bins to the empty encoders 2 to the empty encoder K-1.

The bin encoder 660 may then bin encode the coding bins input from the individual buffer 650 to generate codewords. The empty encoder 660 may collect the generated codewords and output the partial codestream.

For example, when coding bins 0 to 10 are input in individual buffers 0, the bin encoder 0 may encode the coding bins 0 to 10 to generate codewords 0 to 10. The empty encoder 0 may collect the codewords 0 to 10 to output the partial bitstream 0.

In the same manner, each of the empty encoders 1 to empty encoder K-1 may bin encode coding bins inputted from each of the individual buffers 1 to individual buffer K-1 to generate codewords, respectively. Each of the encoders 1 to empty encoder K-1 may collect the codewords and output the partial bitstream 1 to the partial bitstream K-1, respectively.

Then, the multiplexer 670 may output one bitstream by multiplexing the partial bitstreams and the entropy coded slice information output from the empty encoder 660.

In this case, in the decoding apparatus, when the probability information of each of the partial bitstreams is not already known, the multiplexer 670 multiplexes the partial bitstreams, the entropy encoding slice information, and the probability information of each of the partial bitstreams. It can output a bitstream of. Here, the probability information of each of the partial bitstreams may include a representative probability of the empty encoder that outputs the partial bitstream. In one example, the multiplexer 670 multiplexes 1) partial bitstream 0 to partial bitstream K-1, 2) entropy coded slice information, and 3) partial bitstream 0 to partial bitstream K-1. can do.

7 is a block diagram showing the configuration of a decoding apparatus according to an embodiment of the present invention.

FIG. 7 includes a plurality of individual buffers in the decoding apparatus of FIG. 5. Accordingly, description overlapping with FIG. 5 among the components of FIG. 7 will be omitted.

According to FIG. 7, the decoding apparatus 700 includes a demultiplexer 710, an empty decoder 720, an individual buffer 730, an empty decoder determiner 740, a context modeler 750, and an inverse binarizer 760. ), And an alignment unit 770. In FIG. 7, the bin decoder 720, the individual buffer 730, the bin decoder determiner 740, the context modeler 750, and the inverse binarizer 760 may be configured in plural.

The individual buffer 730 has the same number as the bin decoder 720 and may be located after the bin decoder 720. As such, as located after the bin encoder, the individual buffer 720 may store the coding bins output from the bin decoder 720. That is, the individual buffer 730 may temporarily store the coding bins output from the plurality of bin decoders.

In one example, if K individual buffers 730 are configured, K 0 may be located after empty decoder 0. In addition, the individual buffer 0 may temporarily store coding bins output from the bin decoder 0. Similarly, the individual buffers 1 to K-1 may each temporarily store the coding bins output from the bin decoders 1 to empty decoder K-1.

The bin decoder determiner 740 may determine a probability section to which the LPB probability of the bin belongs. The bin decoder determiner 740 may determine a bin decoder having a representative probability of the determined probability interval. Then, the bin decoder determiner 740 may receive a coding bin output from an individual buffer corresponding to the determined bin decoder.

For example, the bin decoder determiner 0 may receive coding bins output from the individual buffer 0 to the individual buffer K-1. The bin decoder determiner 1 may receive the coding bins output from the individual buffer 0 to the individual buffer K-1. Similarly, each of the empty decoder determiner 2 to the empty decoder determiner M-1 may receive coding bins output from the individual buffers 0 to individual buffers K-1.

Then, the context modeling unit 750 may restore the plurality of empty columns by using the coding bins output from the bin decoder determining unit 740 and the LPBs of the bins corresponding to the corresponding coding bins. In addition, the context modeling unit 750 may update the LPB and the LPB probability of the bin constituting the probability model. In this case, the context modeling unit 750 may be configured in plurality.

In addition, the inverse binarization unit 760 may inverse binarize the plurality of bins constituting the empty column to restore values of syntax elements. The inverse binarization unit 760 may reconstruct a plurality of entropy coded slices by collecting values of reconstructed syntax elements. In this case, the reverse binarization unit 760 may be configured in plurality.

As an example, the inverse binarization unit 0 may restore the values of syntax elements by inverse binarizing bins constituting the empty column 0 output from the context modeling unit. The inverse binarization unit 0 may recover the entropy coded slice 0 by collecting values of the recovered syntax elements. In the same manner, inverse binarization unit 1 restores entropy coded slice 1. The inverse binarization unit M-1 may reconstruct the entropy encoding slice M-1.

The alignment unit 770 may sort the plurality of entropy coded slices reconstructed through inverse binarization based on entropy coded information. For example, the alignment unit 770 may reconstruct image data by aligning entropy coded slices 0 to entropy coded slices M-1.

8 is a block diagram illustrating a configuration of an encoding apparatus according to an embodiment of the present invention.

8 further includes a probability interval allocator in the encoding apparatus of FIG. 3. Accordingly, description overlapping with FIG. 3 among the components of FIG. 8 will be omitted.

According to FIG. 8, the encoding apparatus 800 includes an information generator 810, a binarizer 820, a context modeler 830, an empty encoder determiner 840, an empty encoder 850, and a multiplexer 860. , And a probability interval allocator 870. In FIG. 8, the binarizer 820, the context modeler 830, the bin encoder determiner 840, and the bin encoder 850 may be configured in plural.

The probability interval allocator 870 may manage the empty encoder determiner 840 corresponding to each of the plurality of entropy coded slices.

For example, the probability interval allocator 870 may determine the size of the probability interval based on any one of 1) probability distribution of the probability interval including the representative probability, 2) coding efficiency according to the probability, and 3) the number of empty decoders. And the number of probability intervals.

The probability section allocator 870 may adjust the probability section of the empty encoder determiner and the representative probability of the probability section based on the size and number of the determined probability section. For example, the probability section allocator 870 adjusts the probability section of each of the empty encoder determiner 0 to the empty encoder determiner M-1, and adjusts the representative probability of each of the empty encoder determiner 0 to the empty encoder determiner M-1. I can regulate it. In this case, the probability section allocator 870 may calculate a probability distribution of the probability section by using a frequency of occurrence of bins.

In detail, referring to FIG. 14, when the LPB probability 1410 is divided into a plurality of probability intervals 1401 at random intervals, the probability interval allocator 870 may determine the probability distribution, the coding efficiency, and the empty decoder. Based on any one of the numbers, a probability section of each bin encoder determiner and a representative probability of each probability section may be adaptively adjusted 1403. In this case, as the size of the probability interval is adjusted, the representative probability 1404 belonging to the adjusted probability interval may also be adaptively adjusted.

The probability section allocator 870 may insert information indicating the adjusted probability section and the representative probability into the bitstream and transmit the information to the decoding apparatus as the size of the probability section and the representative probability are adjusted.

9 is a block diagram showing the configuration of a decoding apparatus according to an embodiment of the present invention.

9 further includes a probability interval allocator in the decoding apparatus of FIG. 5. Accordingly, description overlapping with FIG. 5 among the components of FIG. 9 will be omitted.

According to FIG. 9, the decoding apparatus 900 includes a demultiplexer 910, a bin decoder 920, a bin decoder determiner 930, a context modeler 940, an inverse binarizer 950, and an alignment unit 960. And a probability interval allocator 970. In FIG. 9, the bin decoder 920, the bin decoder determiner 930, the context modeler 940, and the inverse binarizer 950 may be configured in plural.

The probability interval allocator 970 may control an operation of a plurality of empty decoder determiner units. In this case, the probability interval allocator 970 may adjust the representative probability of each of the probability intervals and the probability intervals of the plurality of bin decoder determiners based on one of the probability distribution of the bin, the coding efficiency, and the number of the bin decoders.

The decoding apparatus may decode information indicating the adjusted probability interval and the representative probability inserted into the bitstream by the encoding apparatus. The decoding apparatus may adjust the size and the representative probability of the probability interval using the decoded information.

Then, the bin decoder 920 and the bin decoder determiner 930 may perform bin decoding by using the newly adjusted probability interval and the representative probability.

As an example, the bin decoder 920 may bin decode codewords in the partial bitstreams based on the adjusted probability interval and the adjusted representative probability, respectively.

The bin decoder determiner 930 may determine a probability section to which the LPB probability of the bin requested by the inverse binarization unit 950 belongs based on the adjusted probability section and the adjusted representative probability. The bin decoder determiner 930 may receive and output a coding bin from a bin decoder having a representative probability of a probability section determined among a plurality of bin decoders.

Up to now, the probability interval allocator which adaptively controls the empty encoder determiner and the empty decoder determiner has been described with reference to FIGS. 8 and 9. In this case, the encoding apparatus may adaptively control the bin encoder determiner corresponding to each of the plurality of entropy coded slices even when the bins are temporarily stored using a plurality of individual buffers. In other words, as shown in FIG. 10, the encoding apparatus may further include a probability interval allocator 1010 in the encoding apparatus of FIG. 6. Similarly, the decoding apparatus may further include a probability interval allocator 1110 in the decoding apparatus of FIG. 7.

Meanwhile, in the encoding and decoding apparatus according to an embodiment of the present invention, the number of the plurality of empty encoder determiners and the number of the empty encoders may be different from each other, and the number of the plurality of empty decoder determiners and the number of empty decoders may be different from each other. Can be. In this case, the structures of the plurality of empty encoder determiners used in the encoding apparatus and the structures of the plurality of empty decoder determiners used in the decoding apparatus may be different from each other. For example, when the encoding apparatus and the decoding apparatus mutually promise that the structures of the plurality of empty encoder determination units and the structures of the plurality of empty decoder determination units are mutually different from each other, the encoding apparatus specifies the structures of the plurality of empty encoder determination units in the bitstream. You can't. In this case, when the encoding and decoding devices do not promise each other, the encoding device may insert information indicating the structures of the plurality of empty encoder determination units into the bitstream and transmit the information to the decoding device. Then, the decoding apparatus may configure the plurality of bin decoder determiners based on the information indicating the structures of the plurality of bin encoder determiners.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

310: information generating unit
320: binarization
330: context modeling unit
340: empty encoder determiner
350: empty encoder
360: multiplexer

Claims (20)

An information generation unit for dividing image data into a plurality of entropy coded slices and generating entropy coded slice information including a unit for dividing the image data into entropy coded slices;
A plurality of binarizers for binarizing the plurality of entropy coded slices into bin sequences;
A plurality of context modeling units generating coding bins using bins constituting the bin column and Less Probable Bins (LPBs) of bins;
A plurality of bin encoder determination units for allocating coding bins to a bin encoder based on LPB probabilities of bins constituting the bin column;
A plurality of individual buffers located before the bin encoder, the individual buffers storing the coding bins; And
A plurality of bin encoders that bin encode the coding bins to generate codewords
Encoding apparatus comprising a. The method of claim 1,
The plurality of context modeling unit,
Determine LPB probabilities of the bins based on the context information of the bin and the location of the bin,
The plurality of empty encoder determiner,
And a probability section to which the LPB probabilities of the bins belong among the plurality of probability sections, respectively, and allocate the coding bins to a plurality of bin encoders corresponding to the determined probability sections, respectively. The method of claim 1,
The plurality of context modeling unit,
And an LPB of the bins and an LPB probability of the bins. The method of claim 1,
The plurality of context modeling unit,
And an encoding OR operation on the bins constituting the empty column to generate an exclusive OR operation on a LP and a Least Probable Bin (LPB) of the bin. The method of claim 1,
The plurality of individual buffers,
And the same number as the plurality of empty encoders. The method of claim 1,
A multiplexer for multiplexing the plurality of partial bitstreams generated by collecting the codewords and the entropy coded slice information;
Encoding apparatus further comprising. The method of claim 1,
The information generator,
And encoding the syntax elements constituting the image data or the blocks constituting the image data to divide the image data into a plurality of entropy encoding slices. Generating entropy coded slice information including a unit of dividing image data into entropy coded slices;
Binarizing each of the plurality of entropy coded slices into a bin sequence;
Generating coding bins in parallel using bins constituting the bin column and Less Probable Bins (LPBs) of bins;
Assigning coding bins to a bin encoder based on LPB probabilities of bins constituting the bin column;
Storing the coding bins in parallel; And
Bin encoding the coding bins to generate codewords in parallel
Encoding method comprising a. The method of claim 8,
Generating the coding bins in parallel,
And determining LPB probabilities of the bins based on the context information of the bin and the position of the bin. 10. The method of claim 9,
Assigning the coding bins to a bin encoder,
Determining a probability interval among the plurality of probability intervals to which the LPB probability of the bins belong; And
Allocating the coding bins to a plurality of bin encoders respectively corresponding to the determined probability intervals.
Encoding method comprising a. The method of claim 8,
Updating the LPB of the bins and the LPB probability of the bins
Encoding method further comprising. The method of claim 8,
Collecting the codewords to generate a partial bitstream in parallel; And
Multiplexing the parallel generated partial bitstream and the entropy encoded slice information
Encoding method further comprising. The method of claim 8,
Dividing the image data into a plurality of entropy encoding slices by grouping syntax elements constituting the image data or blocks constituting the image data
Encoding method further comprising. A demultiplexer for demultiplexing the bitstream to separate a plurality of partial bitstreams;
A plurality of bin decoders for bin decoding the plurality of partial bitstreams to generate a plurality of coding bins;
A plurality of bin decoder determination units that determine a bin decoder using LPB probabilities of bins, and receive and transfer the coding bins from the determined bin decoders;
A plurality of context modeling units configured to generate empty columns using the coding bins and LPBs of the coding bins; And
A plurality of inverse binarization units for inverse binarizing the bins constituting the empty column to restore a plurality of entropy coded slices
Decoding apparatus comprising a. The method of claim 14,
The plurality of context modeling unit,
Determining the LPB of the bin and the LPB probability of the bin based on the context information of the bin and the location of the bin, and updating the LPB of the bin and the LPB probability of the bin, based on the bins constituting the bin column; Decryption device. The method of claim 14,
A plurality of individual buffers located after the plurality of bin decoders and storing coding bins output from the plurality of bin decoders
Decoding apparatus further comprising. The method of claim 16,
The plurality of individual buffers,
The same number as the plurality of bin decoders,
The plurality of empty decoder determiner,
And a coding bin output from an individual buffer corresponding to the determined bin decoder. The method of claim 14,
The plurality of inverse binarization unit,
And inverse binarization of bins constituting the empty column to restore values of syntax elements, and collect values of the restored syntax elements to recover a plurality of entropy encoding slices. Demultiplexing the bitstream to separate the plurality of partial bitstreams;
Bin decoding the plurality of partial bitstreams to generate a plurality of coding bins;
Determining a bin decoder using the LPB probability of the bin and delivering a coding bin;
Storing the coding bins output from the bin decoder;
Generating an empty column using the coding bin and the LPB of the coding bin; And
Inverse binarizing the bins constituting the empty column to restore a plurality of entropy coded slices
Decryption method comprising a. 20. The method of claim 19,
Restoring image data by sorting the plurality of entropy coded slices according to entropy coded slice information separated from the bitstream.
Further comprising:
Generating the empty column,
Determining LPB of the bin and LPB probability of the bin based on the context information of the bin and the location of the bin, for the bins constituting the bin column; And
Update the bin's LPB and the bin's LPB probability
Decryption method comprising a.

Images