JP2016131280A - Decoding method and decoding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a decoding method and a decoding device which executes decoding processing parallelly by using plural threads, without using a coding method of a coding device side.SOLUTION: A decoding device 1 receives an input of a bit stream by a bit stream receiver 2, converts the bit stream to VLC data with low dependence between CTUs by an entropy converter 3, assigns a VLC data decoding process to two or more operation threads by the CTU unit, decodes in the two or more operation threads the VLC data of the CTU unit assigned by the entropy converter 3, and recovers a reproduced image.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a decoding method and a decoding apparatus.

  In general, the moving image decoding process is a complicated process, and with the increase in the resolution of the moving image content, a highly efficient moving image decoding process has become difficult. On the other hand, the video decoding processing capability supplied by a single computer device is limited, and there is a tendency to use a plurality of computer devices or a plurality of processing cores. For example, a smartphone usually has a central processing unit (CPU (Central Processor Unit)) that includes two or four cores.

  There is a need to use multiple cores for processing consecutive video bitstreams so that high-resolution content can be decoded fast enough for smooth playback in real time. H., the latest video encoding standard. H.265 / HEVC has a feature (for example, using a technique called Tires or Wavefronts) that configures a bit stream so as to promote the distribution of decoding processing to a plurality of cores. As a decoding method, there is a method in which a bit stream of a bit stream is decoded at a time, the decoded data is stored in a memory, and the decoded data is distributed to a plurality of cores and decoded in parallel. As another decoding method using parallel processor technology, there is known a decoding method in which batches of video blocks to be decoded in parallel are defined and a plurality of batches are decoded in parallel with each other (for example, Patent Documents). 1).

Special table 2009-540680

  However, in the decoding method described in Patent Document 1, it is necessary to define a batch in advance, but not all bit streams are generated using such characteristics by the encoding method. For this reason, it is desirable that the decoding of a bitstream that does not use such features has a function corresponding to processing by a plurality of cores.

  Also, in the method of decoding the bit stream bitstream at a time, storing the decoded data in the memory, and distributing the decoded data to multiple cores in parallel, the memory size for storing the decoded data is very large It gets bigger. For this reason, not only the memory cost increases but also the memory bandwidth and the cache miss increase.

  In order to solve the above-described problem, an object is to provide a decoding method and a decoding apparatus capable of performing decoding processing in parallel using a plurality of threads, regardless of the encoding method on the encoding apparatus side. .

  A decoding method according to the present invention is a decoding method executed by a decoding apparatus, and includes an input step for inputting a first code string and a dependency between encoded blocks on the first code string input in the input step. An entropy conversion step for converting the second code string into a second code string having low reliability and assigning decoding processing of the second code string to two or more work threads in units of coding blocks, and entropy in two or more work threads An image processing step of decoding the second code string in block units allocated in the conversion step and restoring the reproduced image.

  The decoding device according to the present invention converts an input unit that inputs a first code string, and a first code string that is input by the input unit into a second code string that has low dependency between coding blocks. Entropy conversion means for assigning the decoding process of the second code string to two or more work threads in units of encoded blocks, and a second block unit assigned by the entropy conversion means in two or more work threads. Image processing means for decoding the code string and restoring the reproduced image.

  In the entropy conversion step in the decoding method according to the present invention, in the entropy conversion step, the decoding device converts the first code string as a second code string into a second code string having low inter-bit dependency. Also good.

  In the entropy conversion step in the decoding method according to the present invention, the decoding apparatus may convert the first code string as the second code string into a second code string having low dependency between coding elements.

  The entropy conversion step in the decoding method according to the present invention includes a sub-step of decoding the first code string and converting it as a second code string into a second code string that can be separated and decoded for each coding block; Adding a decoding process of the second code string for each coding block to a queue of decoding processes, associating the decoding process with a thread determined based on the coding block string, and associating with each thread A sub-step of performing a waiting process may be included if it is determined whether all the second code strings that have been decoded have been decoded, and there is a second code string that has not been processed.

  In the decoding method according to the present invention, the process of one work thread in the image processing step determines whether or not there is a decoding process of the second code string associated with the target work thread, and the process is performed until it exists. It is determined whether all the encoding blocks necessary for the restoration of the decoding target encoding block associated with the substep to be waited and the target work thread have been restored, and until all the coding blocks are restored. A sub-step of waiting for processing and a sub-step of decoding the second code string acquired corresponding to the encoded block to be decoded and restoring the reproduced image of the block to be decoded may be included.

  In the decoding method according to the present invention, the first code string is encoded with an arithmetic code, and in the entropy conversion step, the decoding apparatus may convert the first code string into a variable length code.

  In the decoding method according to the present invention, the first code string includes a code of the difference parameter value to which the prediction across the sequence of the encoded block is applied, and in the entropy conversion step, the decoding apparatus converts the first code string to the first code string. The first code string may be converted into a second code string that does not include a difference parameter to which prediction is applied across a sequence of encoded blocks.

  In the decoding method according to the present invention, the first code string is encoded using a variable-length code, and in the entropy conversion step, the decoding apparatus converts a part of the first code string into a fixed-length code. May be.

  In the decoding method according to the present invention, in the first code string, codes of some of the encoding elements that can be restored from the decoded data are omitted, and in the entropy conversion step, the decoding apparatus uses the first code string. May be converted into a code string including the code of the omitted encoding element without being omitted.

  According to the present invention, since the first code string is converted into the second code string having low dependency between the coding blocks, a plurality of threads are used regardless of the coding method on the coding device side. Decoding processing can be executed in parallel.

It is a block diagram of the decoding apparatus 1 which concerns on embodiment of this invention. 2 is a hardware configuration diagram of a decoding device 1. FIG. It is a figure which illustrates the example of CTU conceptually. 10 is a flowchart of processing by the decoding device 1. 4 is a flowchart of processing by an entropy conversion unit 3. 5 is a flowchart of processing by the image processing unit 5.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements will be denoted by the same reference numerals, and redundant description will be omitted.

  FIG. 1 is a block diagram of a decoding apparatus 1 according to an embodiment of the present invention. As shown in FIG. 1, the decoding device 1 includes a bitstream reception unit 2 (input unit), an entropy conversion unit 3 (entropy conversion unit), a storage unit 4, an image processing unit 5 (image processing unit), And a display unit 6.

  Here, the hardware configuration of the decoding device 1 will be described with reference to FIG. FIG. 2 is a hardware configuration diagram of the decoding device 1. As shown in FIG. 2, the decoding device 1 physically includes one or a plurality of CPUs (Central Processing Units) 21, a main storage device such as a RAM (Random Access Memory) 22, and a ROM (Read Only Memory) 23. Hardware such as a communication module 24 that is a data transmission / reception device, an auxiliary storage device 25 such as a semiconductor memory, an input device 26 that receives user input such as an operation panel (including operation buttons) and a touch panel, and an output device 27 such as a display. It can comprise as a computer provided with the wear. Each function of the decoding device 1 in FIG. 2 is, for example, by reading one or a plurality of predetermined computer software on hardware such as the CPU 21 and the RAM 22, thereby controlling the communication module 24 and the input device 26 under the control of the CPU 21. This can be realized by operating the output device 27 and reading and writing data in the RAM 22 and the auxiliary storage device 25.

  The bit stream receiving unit 2 is a part that receives an input of a bit stream (first code string). Specifically, the bit stream reception unit 2 receives an input of a bit stream encoded by the encoding device. The bit stream reception unit 2 inputs, for example, CABAC (Context-based Adaptive Binary Arithmetic Coding) data as a bit stream. The CABAC data is data encoded using an arithmetic code. The CABAC data received by the bitstream receiving unit 2 may be stored in a storage unit (for example, the RAM 22 or the auxiliary storage device 25). After the CABAC data is input, the bit stream reception unit 2 sends the CABAC data to the entropy conversion unit 3. The CABAC data includes a case where a part of the CABAC data (for example, slice header data) is encoded with a variable length code.

  The entropy conversion unit 3 converts the bit stream input by the bit stream reception unit 2 into a code string having low dependency between coding blocks, and decodes the converted code string (restoring the bit stream into a reproduced image) Process) is assigned to two or more work threads in units of encoded blocks. Specifically, when CABAC data is input by the bitstream reception unit 2, the entropy conversion unit 3 analyzes the slice header and decodes the input CABAC data. Further, the entropy conversion unit 3 converts the CABAC data into VLC (Variable Length Coding) data (variable length code) as a code string having low dependency between coding blocks. That is, the entropy conversion unit 3 converts CABAC data, which is an arithmetic code, into a variable length code.

  The arithmetic processing load (for example, the amount of calculation) required for reading / writing VLC data is smaller than the arithmetic processing load required for reading / writing CABAC data. Since the entropy conversion unit 3 converts the CABAC data into VLC data, each segment of the VLC data (for example, VLC data corresponding to each CTU (Coding Tree Unit)) can be read independently. The entropy conversion unit 3 converts VLC data in units of CTUs. Here, the CTU is a predetermined coding unit (coding element unit) constituting a picture, and is a block of 64 × 64 pixels, for example. This block size is not limited to 64 × 64 pixels, and may be 8 × 8 pixels, 16 × 16 pixels, and 32 × 32 pixels. Note that the data of each CTU is not necessarily a fixed length. That is, the boundary between two consecutive CTU data is usually not known from the data size. Therefore, it is necessary to store an additional pointer in order to randomly access the boundary between CTU data. For example, one pointer needs to indicate the memory associated with one or a group of CTUs (decoding elements) defined in one element of decoding.

  In this way, the entropy conversion unit 3 converts (re-encodes) the CABAC data into VLC data that is a code string with low dependency between encoding elements. That is, the entropy conversion unit 3 converts the data into VLC data as an intermediate compression format that is an intermediate compressed stream. As a result, the entropy conversion unit 3 can convert data with low dependency between CTUs and low dependency between bits. The entropy conversion unit 3 accumulates the converted VLC data in the storage unit 4. That is, the storage unit 4 stores the VLC data and makes the image processing unit 5 accessible. Therefore, the entropy conversion unit 3 can cause the image processing unit 5 to decode the VLC data separated and converted in units of CTUs.

  Note that data obtained by decoding CABAC data instead of VLC data is also data with low dependency between CTUs, but the amount of data is very large compared to VLC data. By converting the CABAC data into VLC data, the data size accumulated in the storage unit 4 can be greatly reduced, the memory bandwidth of the decoding device 1 can be reduced, and the occurrence of a cache miss can be suppressed.

  Subsequently, a specific example in which the entropy conversion unit 3 performs re-encoding will be described. It is assumed that the bit stream received by the bit stream receiving unit 2 is compressed by a binary arithmetic encoder. In this case, the entropy conversion unit 3 uses H.264 as one method. As seen in the H.265 / HEVC encoding standard system, it is conceivable to acquire a BIN string obtained from a bit stream and store it as a bit string. However, this method is not sufficient because random accessibility is limited. For example, the restoration of the quantization parameter QP depends on the elements decoded in the past (for example, the QP value of the first element of a certain CTU sequence is the QP of the last element of the CTU sequence restored immediately before. Value and dependency). As described above, the quantization parameter QP value is a difference parameter value to which a prediction across a sequence of encoded blocks is applied.

Here, there are some means (means not including a difference parameter) for increasing independence by encoding the QP value by another method in an intermediate compression format (intermediate compressed stream). Enumerate.
(1) All QP values are encoded as absolute values instead of differential QP values.
(2) The QP value itself is encoded instead of the differential QP value for the first element of the CTU sequence.
(3) In the first element of the CTU sequence, a differential QP value from a defined reference value is transmitted. For example, the defined reference value is an intermediate value such as 26 (QP value range is 0 to 51) or an initial value obtained from slice header information.

  The entropy conversion unit 3 converts the bit stream converted by the image processing unit 5 into a coding block because the entropy conversion unit 3 converts the bit stream having low dependency between the CTUs by performing the processes (1) to (3). It becomes possible to separate and decipher in units.

  Subsequently, H.C. Regarding the context of the H.265 / HEVC standard, a method in which the entropy conversion unit 3 simplifies an intermediate compression format (intermediate compressed stream) will be described.

  For example, when the bit stream input by the bit stream reception unit 2 is encoded using a variable length code, the entropy conversion unit 3 uses a variable length code instead of the variable length code for a certain syntax element (encoding element). A fixed length code may be used. For example, the last (significant) coefficient position in the block (last_sig_coeff_x_prefix, last_sig_coeff_y_prefix, last_sig_coeff_x_suffix and last_sig_coeff_y_suffix syntax elements) can be represented by a fixed length code. Here, the code length depends on the conversion size. By this processing, the stream can be converted into a stream with low dependency between encoded blocks. For example, even when an arithmetic code is used for the intermediate format, by making the last coefficient of the block a fixed-length code, the boundary between the CTUs of the bit string becomes clear and the first bit of the CTU can be accessed at random. Thus, the entropy conversion unit 3 may convert a part of the bit stream into a fixed length code when the bit stream input by the bit stream reception unit 2 is a variable length code.

  Further, the entropy conversion unit 3 does not include the estimation rule for the significance map flag sig_coeff_flag in the intermediate compression format (intermediate compressed stream), and always transmits 16 significance map bits to the 4 × 4 sub-block. (For example, even if the 1st to 15th bit values are 0, the 16th bit is not inferred as 1 and 0 is transmitted). The entropy conversion unit 3 converts the compressed stream into an intermediate compressed stream with less dependency between the encoded elements by the above-described processing, so that the complexity of the decoding process can be reduced. In this way, the entropy conversion unit 3 includes the omitted codes without omitting the codes of some of the encoding elements that can be restored from the encoded data in the input bitstream. Convert to code string.

  Further, the entropy conversion unit 3 may not include the estimation rule for the sign flag coeff_sign_flag in the intermediate compression format (intermediate compressed stream). Under certain conditions, the last sign flag of the 4 × 4 sub-block is self-explanatory and need not be encoded, but the entropy conversion unit 3 includes the sign bits of all the coefficients in the intermediate compression format. With this process, the entropy conversion unit 3 converts the stream into a stream having low dependency between bits and less guessing processing at the time of decoding, and thus the complexity of the decoding process can be reduced. In this way, the entropy conversion unit 3 includes the omitted code without omission even if the codes of some encoding elements that can be restored from the encoded data are saved in the input bitstream. Convert to code string.

  Further, the entropy conversion unit 3 may use a simple and regular variable length code (for example, Exponential Golomb code) for encoding the coefficient (however, adaptation of the Rice parameter is not used). By this processing, it can be converted into a stream with low dependency between bits.

  In addition, the entropy conversion unit 3 may change the order of the syntax elements in order to facilitate the decoding of the intermediate compressed stream. For example, when the entropy conversion unit 3 considers a path at the position of one coefficient in a 4 × 4 block, the entropy conversion unit 3 does not put a sign bit between a flag of a coefficient greater than 1, a coefficient greater than 2, or the remaining coefficient. It may be. This processing can be realized by using a single unary decode operation. Since the entropy conversion unit 3 can convert the stream into a stream with less dependency between the encoding elements by the above-described process, the complexity of the decoding process can be reduced. As described above, the entropy conversion unit 3 converts the bit stream into a replaced code string so that a part of the bit stream can be decoded by one unary operation.

  In addition, the entropy conversion unit 3 adds a decoding process (decoding process) of VLC data corresponding to each encoded block to a queue (queue), and also performs a decoding process based on the sequence of encoded blocks. Are associated (mapped) and notified to the image processing unit 5. The queue is a queue for a plurality of work threads executed by the image processing unit 5 described later. Here, the work thread is a thread for decoding the VLC data and executing a corresponding element (CTU) restoration process. Also, there are a plurality of work threads, and each work thread operates independently and in parallel.

  Here, an example in which the entropy conversion unit 3 associates work threads based on a sequence of encoded blocks will be described with reference to FIG. FIG. 3 shows a sequence of CTUs (encoded block sequence composed of side-by-side encoded blocks). Each CTU is labeled (an identifier that uniquely identifies each CTU is determined). In the example of FIG. 3, each CTU is labeled with each column and the order of the same column (number given from left to right in FIG. 3). For example, the leftmost CTU in column A is labeled “A1”.

  The entropy conversion unit 3 determines a CTU to be processed by each work thread based on the CTU column and the work thread number (a number uniquely determined for each work thread). That is, the entropy conversion unit 3 associates the CTU and the work thread.

  When the entropy conversion unit 3 assumes that the number of work threads is N, a thread number is assigned to a CTU in the CTU sequence R (for example, a remainder value obtained by dividing the sequence R by N). For example, when the number of work threads constituting the image processing unit 5 described later is two, the entropy conversion unit 3 assigns odd-numbered CTUs to the first work thread, and assigns even-numbered CTUs to the second Assign to a work thread. As described above, when the entropy conversion unit 3 associates each work thread with the CTU, the associated result is registered in the storage unit 4 (even if the entropy conversion unit 3 and the image processing unit 5 store the result). Good). The image processing unit 5 performs the decoding process for each work thread based on this association. That is, in the example of FIG. 3, the image processing unit 5 performs restoration processing of the reproduction images of the CTUs indicated by A1, A2, A3, A4, and C1 as the first work thread, and B1 as the second work thread. And CTU reconstructed image data indicated by B2.

  Also, the entropy conversion unit 3 determines whether all the second code strings (VLC data) associated with each thread have been decoded, and when there is unprocessed VLC data, wait processing I do. This process is a process for synchronizing the process by the entropy conversion unit 3 and the process by the above-described work thread. The entropy conversion unit 3 is realized by executing a known sleep command as a waiting process. Further, the entropy conversion unit 3 may perform a waiting process using a condition variable by the pthread library.

  The storage unit 4 is a part that stores the VLC data converted by the entropy conversion unit 3. The storage unit 4 is a part that stores the association between each thread and the CTU defined by the entropy conversion unit 3. Based on the association between each work thread stored in the storage unit 4 and the CTU, the image processing unit 5 performs restoration processing of the reconstructed image of the CTU in units of each work thread (the image processing unit 5 is provided with each work thread and CTU). Record the association).

  The image processing unit 5 is a part that decodes VLC data in block units allocated by the entropy conversion unit 3 and restores a reproduced image (that is, performs decoding processing) by two or more work threads. The image processing unit 5 performs, in parallel, a restoration process of elements assigned to each work thread for a plurality of work threads. The image processing unit 5 is implemented by, for example, a plurality of CPUs, and each of the CPUs performs a restoration process of the reproduced image of the CTU assigned to the work thread. The image processing unit 5 performs a decoding process in response to the notification from the entropy conversion unit 3.

  The image processing unit 5 determines whether there is VLC data associated with the target processing thread, and waits for processing until it exists. Specifically, the image processing unit 5 refers to the storage unit 4 to determine whether there is a CTU assigned to its own thread, and waits for processing until it exists. As this waiting process, the above-mentioned known sleep command is executed, or a polling process is executed. Further, the image processing unit 5 may perform a waiting process based on the condition variable.

  When there is VLC data associated with the target processing thread, the image processing unit 5 has restored all the encoding blocks necessary for restoring the decoding target encoding block associated with the target work thread. And waits until all the encoded blocks are restored.

  Specifically, the image processing unit 5 determines whether all the elements (CTUs) related to (that is, dependent on) the CTU restored by decoding the VLC data to be processed have been restored. Such dependencies can arise from other processes involving intra prediction, motion vector reconstruction, or intersecting element operations.

  H. In the H.265 / HEVC standard, restoration of CTUs is required after restoration of neighboring CTUs (when belonging to the same tile). Reconstructed adjacent CTUs indicate the CTUs immediately to the left, upper left, immediately above, and upper right of the current CTU. This will be described using the example of the CTU shown in FIG. For example, assume that 3 threads have restored 3 CTU sequences and that elements A1, A2, A3 and B1 (blocks darkly drawn in the figure) are marked as restored. Thread 1 restores element A4 and thread 2 processes element B2 (all dependent elements A1, A2, A3 and B1 are marked as processed). Since the dependency of the element C1 on B2 has not yet been resolved, the thread 3 is waiting for the restoration process of the element C1. If B2 is marked as processed by thread 2, thread 3 restores element C1.

  If the image processing unit 5 determines that all elements (CTUs) related to (relied upon) the CTU restored by decoding the VLC data to be processed have been restored, the VLC acquired corresponding to the CTU to be decoded The data is decoded, and the reproduced image of the decoding target CTU is restored. Specifically, as the decoding processing, the image processing unit 5 analyzes target VLC data, performs motion compensation, intra prediction, inverse quantization, inverse transformation, residual signal and prediction signal Addition is performed. The image processing unit 5 sends the restored reproduced image to the display unit 6. The display unit 6 displays the reproduced image acquired from the image processing unit 5.

  Subsequently, processing and operations of the decoding method executed by the decoding device 1 according to the present embodiment will be described using the flowcharts of FIGS. The flowchart illustrated in FIG. 4 is a flowchart of the entire decoding process performed by the decoding device 1.

  First, the bitstream accepting unit 2 accepts input of CABAC data (step S1: input step). Then, the entropy conversion unit 3 analyzes the CABAC data and converts it into VLC data (step S2: entropy conversion step). Subsequently, the entropy conversion unit 3 associates the converted VLC data with the work thread in units of CTU (step S3: entropy conversion step). Subsequently, the image processing unit 5 decodes the VLC data of each CTU, restores the decoded image (reproduced image) (step S4: image processing step), and ends the process. The image processing unit 5 sends the decoded image to the display unit 6, and the decoded image is displayed on the display unit 6.

  Next, details of the process performed by the entropy conversion unit 3 among the processes and operations executed by the decoding device 1 according to the present embodiment will be described using the flowchart of FIG. 5. The flowchart shown in FIG. 5 is a flowchart showing details of the processing in steps S2 and S3 of the flowchart shown in FIG. First, when the entropy conversion unit 3 receives input of CABAC data by the bitstream reception unit 2, the entropy conversion unit 3 analyzes the slice header (step S11). Subsequently, the entropy conversion unit 3 analyzes the CABAC data (step S12). The amount of data analyzed at this time corresponds to one element (CTU). Then, the entropy conversion unit 3 re-encodes the CABAC data as VLC data, and accumulates the VLC data in the storage unit 4 (step S13). Then, the entropy conversion unit 3 adds a restoration process of an element (CTU) corresponding to the converted (recoded) VLC data to the queue (step S14). The entropy conversion unit 3 determines whether there is additional data (step S15). If there is additional data (step S15; YES), the process proceeds to step S12. When there is no additional data (step S15; NO), the entropy conversion unit 3 has restored all the elements (CTU) necessary for restoring the element (CTU) to be decoded by the image processing unit 5 It is determined whether or not (step S16). When all the elements have not been restored (step S16; NO), the entropy conversion unit 3 performs a waiting process (step S17), and determines again whether all the elements have been restored (step S16). On the other hand, when all the elements have been restored (step S16; YES), the process ends.

  Next, details of processing performed by the image processing unit 5 among processing and operations executed by the decoding device 1 according to the present embodiment will be described using the flowchart of FIG. 6. The flowchart shown in FIG. 6 is a flowchart showing details of the process in step S4 of the flowchart shown in FIG. First, the image processing unit 5 determines whether or not there is an element to be processed (CTU) in the queue (step S21). When there is no element to be processed in the queue (step S21; NO), the image processing unit 5 performs a waiting process until the element to be processed exists in the queue (step S22), and the process proceeds to step S21 again.

  When there is an element to be processed in the queue (step S21; YES), the image processing unit 5 acquires VLC data corresponding to the element (CTU) from the queue (step S23). Subsequently, the image processing unit 5 determines whether or not all the elements necessary for restoring the element have been restored (step S24). If all the elements necessary for restoring the processing target element have not been restored (step S24; NO), the image processing unit 5 executes a waiting process (step S25), and proceeds to step S24 again.

  When all the elements related to the element to be processed have been restored (step S24; YES), the image processing unit 5 analyzes the VLC data related to the element (step S26). Subsequently, the image processing unit 5 stores the decoded motion data (step S27) and performs motion compensation (step S28). Subsequently, the image processing unit 5 performs prediction signal addition, restoration of a residual signal (error signal) by intra prediction, inverse quantization, and inverse transformation processing, and image restoration by adding the residual signal and the prediction signal. (Step S29). Subsequently, the image processing unit 5 performs a deblocking process (step S30). Subsequently, the image processing unit 5 performs an in-loop filter process (step S31). Subsequently, the image processing unit 5 marks that the element has been restored (step S32). Subsequently, the image processing unit 5 determines whether or not all the processes in the thread have been completed (step S33). If the process has not been completed (step S33; NO), the process proceeds to step S21, where all the processes are completed. If so (step S33; YES), the process is terminated.

  As described above, the decoding device 1 converts the CABAC data into VLC data, re-encodes the bit stream that has received the input, and stores the re-encoded data in the storage unit 4. And can be stored with a small memory size. In addition, the decoding apparatus 1 associates the decoded processing of the VLC data with the work threads in units of CTUs so that a plurality of work threads operate in parallel, and can reduce the dependency between the data as much as possible.

  In the above-described embodiment, the H.264 standard. Although described based on the H.265 / HEVC video encoding standard system, The present invention may be applied to other encoding schemes such as H.264 / AVC. In this case, a macroblock is applied instead of the CTU. In the above-described embodiment, the case where the element corresponding to the VLC data converted by the entropy conversion unit 3 or the element to be processed in the image processing unit 5 is set as one CTU unit is described. It is also possible to perform image signal processing as a group CTU.

  DESCRIPTION OF SYMBOLS 1 ... Decoding apparatus, 2 ... Bit stream reception part, 3 ... Entropy conversion part, 4 ... Memory | storage part, 5 ... Image processing part, 6 ... Display part.

Claims (10)

A decoding method executed by a decoding device,
An input step of inputting a first code string;
The first code string input in the input step is converted into a second code string having low dependency between coding blocks, and the decoding process of the second code string is performed in two or more coding block units. An entropy conversion step assigned to the work thread;
In the two or more work threads, an image processing step of decoding a second code string in block units allocated by the entropy conversion step and restoring a reproduced image;
A decryption method.   2. The decoding method according to claim 1, wherein, in the entropy conversion step, the decoding apparatus converts the first code string into a second code string having low dependency between bits as the second code string.   The said entropy conversion step WHEREIN: The said decoding apparatus converts a 1st code sequence into a 2nd code sequence with low dependence between encoding elements as said 2nd code sequence. Decryption method. The entropy conversion step includes
A sub-step of decoding a first code string and converting the second code string into a second code string that can be separated and decoded for each coding block;
Adding a decoding process of the second code string for each coding block to a queue of decoding processes and associating the decoding process with a thread determined based on the coding block string;
It is determined whether all the second code strings associated with each thread have been decoded. If there is a second code string that has not been processed, a sub-step for performing a waiting process;
The decoding method according to any one of claims 1 to 3, further comprising: The processing of one work thread in the image processing step is as follows:
A sub-step of determining whether there is a decoding process of the second code string associated with the target work thread and waiting for the process until it exists;
A sub-step of determining whether all the coding blocks necessary for restoring the decoding target coding block associated with the target work thread have been restored, and waiting for processing until all the coding blocks are restored; and ,
A sub-step of decoding a second code string acquired corresponding to the encoded block to be decoded and restoring a reproduced image of the block to be decoded;
The decoding method according to any one of claims 1 to 3, further comprising: The first code string is encoded with an arithmetic code;
In the entropy conversion step, the decoding device comprises:
The decoding method according to any one of claims 1 to 5, wherein the first code string is converted into a variable length code. The first code sequence includes a sign of a difference parameter value to apply prediction across a sequence of encoded blocks;
In the entropy conversion step, the decoding device comprises:
The decoding method according to any one of claims 1 to 5, wherein the first code string is converted into a second code string that does not include a difference parameter to which prediction across the coded block string is applied. The first code string is encoded using a variable-length code;
In the entropy conversion step, the decoding device comprises:
The decoding method according to any one of claims 1 to 5, wherein a part of the first code string is converted into a fixed-length code. In the first code string, codes of some coding elements that can be restored from the decoded data are omitted,
In the entropy conversion step, the decoding device comprises:
The decoding method according to any one of claims 1 to 5, wherein the first code string is converted into a code string including a code of the omitted encoding element without being omitted. Input means for inputting the first code string;
The first code string input by the input means is converted into a second code string having low dependency between coding blocks, and the decoding process of the second code string is performed in two or more coding block units. An entropy conversion means assigned to a work thread;
A decoding apparatus comprising: image processing means for decoding a second code string in block units allocated by the entropy conversion means and restoring a reproduced image in the two or more work threads.