KR101485219B1 - Apparatus and method for video decoding - Google Patents

Abstract

An image decoding apparatus and method are disclosed. The disclosed image decoding apparatus includes a decoding unit that processes a plurality of threads in parallel to decode image data, wherein the decoding of the image data comprises a first decoding including entropy decoding and a second decoding performed after the entropy decoding , A thread of a plurality of threads to be processed in parallel in the N-th time interval by the decoding unit is allocated to the first decoding for the N-th image data, and the rest of the threads are allocated to the N- And is assigned to the second decoding.

Description

[0001] APPARATUS AND METHOD FOR VIDEO DECODING [0002]

Embodiments of the present invention relate to an image decoding apparatus and method, and more particularly, to an image decoding apparatus and method capable of efficient parallel decoding of an image.

As higher performance digital devices become more popular, there is a growing demand for high resolution video services, and image compression techniques are being actively researched to meet these demands.

Recently, image processing technology having excellent compression ratio is widely used in multimedia services such as digital broadcasting, multimedia player, and video conference. Since high-complexity algorithms are used in image processing to support high compression rates, high-performance processors must be used to encode / decode high-resolution image data.

On the other hand, the conventional single-core-based processor has a problem that the processing performance can not be increased beyond a certain level due to an increase in the heat generation and power consumption in the processor because the processing speed is improved by increasing the clock speed. In order to solve such a problem, a multi-core based processor that processes data in parallel using a plurality of cores has been actively studied.

Various studies for improving the decoding performance using the above-mentioned multi-core process have also been carried out in image processing decoders. Typically, the data-level parallelization technique divides image processing data into a plurality of threads and simultaneously processes a plurality of threads in parallel while maintaining the inter-data dependency of image processing.

However, in the data-level parallelization method, since the entropy-encoded portions must be decoded sequentially, it is not easy to perform parallel decoding. Particularly, as the demand for a high-resolution video service increases, CABAC (Context Adaptive Binary Arithmetic Coding) entropy encoding having a good compression ratio is preferred, but it has a disadvantage in that it takes a lot of time for encoding due to high complexity.

Therefore, parallel processing availability of image processing decoding is closely related to parallel processing availability of entropy decoding. Various algorithms for parallelizing entropy decoding have been proposed in the related art. However, most of the conventional algorithms have been studied for improving the performance of hardware decoders and focusing on partial performance enhancement of entropy decoding, There was difficulty.

In order to solve the problems of the prior art as described above, the present invention proposes an image decoding apparatus and method capable of efficient parallel decoding of images.

Other objects of the invention will be apparent to those skilled in the art from the following examples.

According to another aspect of the present invention, there is provided a decoding apparatus for decoding video data, the apparatus comprising: a decoding unit for decoding video data by parallel processing a plurality of threads, wherein the decoding of the video data includes first decoding including entropy decoding, And a second decoding performed after the entropy decoding, wherein a thread of a plurality of threads that are parallel-processed in the Nth time interval by the decoding unit is allocated to the first decoding for the Nth image data, And some of the threads are allocated to the second decoding for the (N-1) th image data.

The second decoding may include at least one of inverse quantization, inverse transform, intraprediction, motion compensation, and deblocking.

The number of the plurality of threads is 5 or more, and the number of the threads of the part may be 3 or more and 5 or less.

Wherein the image data is image data encoded according to a syntax element segmentation technique, the part of the threads includes a first syntax element included in an MBINFO group, a second syntax element included in a PRED group, and a third syntax element included in a CBP group A fourth syntax element included in the SIGMAP group, and one or more threads allocated for decoding the fifth syntax element included in the COEFF group.

The decoding unit may decode the first syntax element, the second syntax element, the third syntax element, the fourth syntax element, and the fifth syntax element by parallel processing the part of the threads according to a pipeline scheme .

The decoding unit may process the remaining part of the threads in parallel by using a data level parallelization technique.

The image data may be image data of a frame unit.

The decoding unit may include a multicore processor composed of a plurality of cores, and each of the plurality of threads may be parallel-processed in the plurality of cores.

According to another embodiment of the present invention, there is provided an apparatus for decoding image data by performing a first decoding including entropy decoding and a second decoding including a decoding process after the entropy decoding, A decoding unit for performing parallel processing of a plurality of threads in each time interval to perform decoding; And a thread allocator for allocating the plurality of threads to be processed in parallel in each time interval, wherein the thread allocator allocates a thread of a plurality of threads to be parallel-processed in a predetermined time interval to the first And allocates a thread of a plurality of threads to be parallel-processed in the next time interval of the predetermined time interval to the second decoding of the Nth video data.

According to another embodiment of the present invention, there is provided a method for decoding video data, the method comprising: parallel processing a plurality of threads to decode video data, wherein the decoding of the video data is performed after first decoding including entropy decoding and after entropy decoding Wherein a thread of a part of a plurality of threads that are parallel-processed in the Nth time interval by the decoding step is allocated to the first decoding for the Nth image data, and the rest of the threads are allocated to the first decoding And the second decoding for the (N-1) -th image data is allocated to the second decoding.

According to the present invention, efficient parallel decoding of an image becomes possible. Particularly, it is possible to maximize the decoding efficiency of image data in a quad-core or more multicore processor.

Figs. 1 to 5 are diagrams for explaining the concept of decoding according to the H.264 / AVC standard.
6 is a block diagram showing a schematic configuration of an image decoding apparatus according to an embodiment of the present invention.
FIGS. 7 through 17 are views for explaining the concept of decoding by the image decoding apparatus according to an embodiment of the present invention.
18 is a flowchart illustrating an overall flow of an image decoding method according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

Figs. 1 to 5 are diagrams for explaining the concept of decoding according to the H.264 / AVC standard. Hereinafter, the concept of decoding according to the H.264 / AVC standard will be described in detail with reference to FIGS. 1 to 5. FIG.

는 "OR"을 의미한다. H.264 / AVC is a video compression standard made by ISO / IEC and ITU. It has a higher compression ratio than conventional video compression standards and has characteristics suitable for streaming over a network, thus being used in various multimedia applications. As shown in FIG. 1, the decoding process according to H.264 / AVC is divided into entropy decoding (ED), inverse transform (IT), inverse quantization (IQ) Prediction (IP), motion compensation (MC), and deblocking filter (DF). 1,

Means "OR ".

The image data generated according to the H.264 / AVC has a structure as shown in FIG. Referring to FIG. 2, a video sequence of H.264 / AVC is composed of a group of pictures (GOP), and each GOP is composed of a plurality of frames. Here, the frames constituting the GOP are classified into an I frame using spatial redundancy, a P frame using temporal redundancy, and a B frame using a reference frame in both directions. One frame is composed of one or more slices, one slice is composed of a plurality of macroblocks, and one macroblock is composed of a luminance block (Luma block) and a color difference block (chroma block) .

Meanwhile, a data-level parallelization technique, which is a typical parallel H.264 / AVC-based image data decoding technique, is a technique for dividing and processing image data so as to enable parallelization. The parallel- , Slice-based parallelism, and macroblock-based parallelism.

Among them, the parallelization technique on a macroblock-by-macroblock basis allocates each macroblock that can be simultaneously processed to different threads to perform parallel decoding. At this time, there is a dependency between the macroblocks as shown in FIG. For example, intraprediction for macroblock # 1, # 2 macroblock # 3, # 3 macroblock # 4, and # 4 macroblock should be preferentially performed for intraprediction of "Current MB".

As a representative technique of the parallelism technique on a macroblock-by-macroblock basis, a 2D-wave parallelism technique allocates a thread to each arrow as shown in FIG. 4 and performs decoding on a macroblock-by-macroblock basis according to the direction of an arrow. In other words, parallel processing is performed by allocating threads horizontally according to the arrow direction. For example, in FIG. 4, MB (4,0), MB (2,1), and MB (0,2) are simultaneously processed at the fifth time (T5) while maintaining data dependency.

In addition, as shown in FIG. 5, the 3D-wave parallelization technique that extends the above-described 2D-wave parallelization technique allocates one thread to entropy decoding and performs priority processing, and then performs inverse quantization, inverse transform, intra- (For example, MC + IT / IQ + DF or IP + IT / IQ + DF) in parallel.

However, since the 3D-wave parallelization technique does not process entropy decoding in parallel, it is difficult to expect high performance improvement due to parallelism. In this regard, the next generation video standard recently proposed various techniques for parallelizing entropy decoding. In particular, CABAC-based encoding with a higher compression ratio is used as the resolution of the image increases. However, CABAC-based encoding has a disadvantage in that it takes a long time to perform entropy decoding because the compression rate is good but the complexity is high. Accordingly, various studies have been conducted to parallelize entropy decoding based on CABAC in order to reduce the time required for performing entropy decoding.

Meanwhile, the CABAC-based H.264 / AVC encoding apparatus performs encoding by sequentially performing binarization, context modeling, and arithmetic coding. As a result, It is difficult to perform high-speed encoding through parallelization.

Due to the above problems, entropy slice parallelization is proposed for parallelization of entropy decoding in H.264 / AVC and next generation standards. Entropy slice parallelism has some similarities with existing slice - level parallelism. However, in the conventional slice-level parallelization method, one frame is divided into a plurality of slices and decoding is performed in units of slices. In entropy slicing method, entropy decoding is performed by dividing only the CABAC portion into slices and using the reconstructed data And performs the remaining decoding. However, the entropy slicing technique has a disadvantage in that encoding efficiency is deteriorated.

Accordingly, a CABAC-based entropy decoding parallelization technique using Syntax Element Partitioning has been proposed to improve the efficiency of encoding efficiency and parallelization based on CABAC. The technique divides a syntax element (SE: Syntax Element) into groups (syntax element groups) and performs parallel processing. Here, the syntax elements are classified into MBINFO group, PRED group, CBP group, SIGMAP group, and COEFF group as shown in Table 1 below (hereinafter, the syntax elements included in the MBINFO group A syntax element included in the PRED group is referred to as a "second syntax element", a syntax element included in the CBP group is referred to as a "third syntax element", a syntax element included in the SIGMAP group is referred to as a "Quot; fourth syntax element ", and syntax elements included in the COEFF group will be referred to as "fifth syntax element").

group Syntax Element MBINFO mb_skip_flag, mb_type, sub_mb_type, mb_field_decoded flag,
end of slice flag PRED prev_intra4x4 pred_mode_flag, rem intra4x4_pred_mode, prev_intra8x8_pred_mode_flag, rem intra8x8_pred_mode, intra_chroma pred_mode, ref_idx_l0, ref_idx_l1, mvd_l0, mvd_l1 CBP transform_size_8x8_flag, coded_block_pattern, coded_block_flag SIGMAP significant_coeff flag, last_significant_coeff flag COEFF coeff_abs_level_minus1, coeff_sign_flag

Hereinafter, an image decoding apparatus and method according to an embodiment of the present invention will be described in detail with reference to the above description.

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

Referring to FIG. 6, an image decoding apparatus 600 according to an embodiment of the present invention includes a thread assigning unit 610 and a decoding unit 620. Hereinafter, the function of each component will be described in detail.

The thread allocation unit 610 allocates a plurality of threads to a task for decoding image data. The decoding unit 620 processes the plurality of threads allocated by the thread allocating unit 610 in parallel to decode the image data.

Here, the decoding of the image data may include at least one of entropy decoding (ED), inverse transform (IT), inverse quantization (IQ), intra prediction (IP), motion compensation (MC) and deblocking filter have.

Hereinafter, for convenience of explanation, decoding including entropy decoding will be referred to as "first decoding " and decoding including a decoding process performed after entropy decoding will be referred to as" second decoding ". Accordingly, the second decoding may include at least one of inverse quantization, inverse transform, intraprediction, motion compensation, and deblocking.

Also, according to an embodiment of the present invention, the image data to be decoded may be image data of a frame unit. Further, the image data may be image data encoded (more accurately, entropy-encoded) according to the syntax element division technique.

According to the present invention, the thread allocation unit 610 divides the time for decoding the entire image data into a predetermined number of time intervals, and then allocates threads in units of time intervals. The decoding unit 620 decodes A plurality of threads allocated in the time domain are processed in parallel to decode the image data.

In this case, some threads of the plurality of threads that are processed in parallel in the N (one or more integer) time intervals may be assigned to the first decoding for the Nth image data. If N is 2 or more, a part of the threads other than the thread of the plurality of threads to be processed in parallel in the Nth time interval may be allocated to the second decoding for the (N-1) th image data.

In other words, the thread allocation unit 610 allocates a thread of a plurality of threads to be parallel-processed in a predetermined time interval (for example, the N-th time interval described above) to a first decoding for N-th image data, And may allocate a thread of a plurality of threads to be processed in parallel in a next time interval of the predetermined time interval to a second decoding of the Nth image data. The thread allocation unit 610 allocates the remaining threads of the plurality of threads to be parallel-processed in the predetermined time interval to the second decoding for the (N-1) th image data, May be allocated to the first decoding of the (N + 1) th image data.

For example, when eight threads are allocated to each time interval as shown in FIG. 7, four threads, which are part of eight threads to be processed in parallel in the Nth time interval, are allocated for the first decoding of the Nth video data And the remaining four threads may be allocated for the second decoding of the (N-1) th image data.

That is, four threads are allocated to the first decoding for the Nth video data in the predetermined time interval (Nth time interval), and four threads are allocated to the second decoding for the (N-1) th video data. In this case, the first decoding for the (N-1) th image data is performed in parallel in the previous time interval (N-1) time interval of the predetermined time interval, and the second decoding for the N- Time interval (N + 1 < th > time interval).

Accordingly, the decoding of the (N-1) th image data is completed in the Nth time interval, and the decoding of the Nth image data is completed in the (N + 1) th time interval.

7, it is described that the number of threads is 4. However, according to another embodiment of the present invention, the number of threads may be 3 or more and 5 or less (accordingly, Or more).

As another example, when 16 threads are allocated in each time interval as shown in FIG. 8, four threads, which are part of the 16 threads to be parallel processed in the Nth time interval, are used for the first decoding of the Nth video data And the remaining twelve threads can be allocated for the second decoding of the (N-1) th image data. In this case, the number of threads may be three or more and five or less, and the remaining threads may be fifteen or less and eleven or more.

Meanwhile, as described in the above examples, it is preferable that the number of threads to be processed in parallel in one time interval is 5 or more (8 or 16), and the number of some threads is 3 or more and 5 or less. The decoding unit 620 may include a multicore processor including a plurality of cores, and the plurality of threads may be parallel-processed by a plurality of cores included in the multicore processor. That is, the image decoding apparatus 600 according to an embodiment of the present invention can be easily applied to decoding an image using an octa-core to hexa core processor system.

According to an embodiment of the present invention, when the image data is an image encoded according to the syntax element division technique, some of the threads allocated for performing the first decoding may include a first syntax element included in the MBINFO group, PRED Decoding of the fifth syntax element included in the COEFF group and the fourth syntax element included in the SIGMAP group, two or more threads allocated for decoding of the third syntax element included in the CBP group, Lt; RTI ID = 0.0 > and / or < / RTI >

In this case, the decoding unit 620 may decode the first syntax element, the second syntax element, the third syntax element, the fourth syntax element, and the fifth syntax element by parallel processing a part of the threads according to a pipeline technique . In other words, the decoding unit 620 can start the processing of the part of the threads with a predetermined time gap and perform decoding in parallel.

Also, according to an embodiment of the present invention, the decoding unit 620 may process the remaining part of the threads in parallel by using the data level parallelization technique.

Hereinafter, with reference to FIGS. 9 to 14, it is assumed that 3 to 5 threads are allocated to the first decoding including entropy decoding, and some threads are allocated to the first decoding for the Nth video data in each time interval, The reason for allocating to the second decoding of the (N-1) th image data and the effect therefrom will be described in detail.

1. Reason for allocating 3 to 5 threads for first decoding involving entropy decoding

FIG. 9 shows an example of syntax elements classified into a plurality of groups according to a syntax element division technique. That is, the syntax elements are classified into MBINFO group, PRED group, CBP group, SIGMAP group, and COEFF group as described above. Therefore, in the case of decoding the entropy-encoded image data according to the syntax element division technique, the decoding can be performed in parallel by allocating one thread to each of five syntax element groups theoretically.

However, there is a data dependency between syntax element groups as shown in FIG. Such data dependency is caused by the degree of binding of data according to the encoding order of the data when the image data is encoded according to the syntax element division technique.

Referring to FIG. 10, the second syntax element included in the PRED group is decoded after the first syntax element included in the MBINFO group is decoded. The fourth syntax element included in the SIGMAP group is decoded 2 syntax element and the third syntax element included in the CBP group must be decoded after being decoded, and the fifth syntax element included in the COFF group must be decoded after the fourth syntax element included in the SIGMAP group is decoded.

In order to perform the parallel processing of CABAC decoding while maintaining the dependency of the data, as shown in FIG. 11, the first thread (Task 1), the second thread (Task 1), and the second thread After the fifth thread (Task 5) is sequentially allocated, processing of the first to fifth threads may be started at a predetermined time interval to perform parallel decoding in a pipelined manner. Therefore, according to the first embodiment of the present invention, five threads can be allocated for performing the first decoding. Here, the five sub time periods shown in FIG. 11 are included in one time period described above.

12, significant_coeff_flag and last_significant coeff_flag in the SIGMAP group are encoded when the coding flag (coded_block_flag) is 1 as shown in FIG. 12, and the residual data encoded in the COEFF group The level information values coeff_abs_level minus1 and coeff_sign_flag are encoded. By this encoding process, strong data binding occurs between the fourth syntax element included in the SIGMAP group and the fifth syntax element included in the COEFF group.

11, when a separate thread is allocated to each syntax element group and parallel decoding is performed according to the pipeline technique, decoding of the fourth syntax element included in the SIGMAP group (that is, decoding of the fourth thread And processing of the fifth thread included in the COEFF group (i.e., processing of the fifth thread), the efficiency of parallel decoding according to the pipeline technique can be greatly reduced.

Therefore, the image decoding apparatus 600 according to the embodiment of the present invention defines the decoding operation of the fourth syntax element included in the SIGMAP group and the decoding operation of the fifth syntax element included in the COEFF group as one decoding operation group And allocates one thread to it, and then performs parallel processing with other threads to decode the same, thereby preventing the occurrence of such overhead.

More specifically, according to one embodiment of the present invention, the thread assignment unit 610 assigns a first syntax element included in the MBINFO group, a second syntax element included in the PRED group, and a third syntax element included in the CBP group Two or more threads are allocated for decoding, and one thread is allocated for decoding of the fourth syntax element included in the SIGMAP group and the fifth syntax element included in the COEFF group. In other words, the thread allocation unit 610 defines two or more decoding operations by appropriately combining the MBINFO group, the PRED group, and the CBP group, assigns threads to two or more prescribed decoding operations, and allocates threads to the SIGMAP group and the COEFF group A decoding operation is defined as one decoding operation and one thread is allocated.

In this case, the decoding unit 620 parallel-processes the two or more threads and the one thread according to a pipeline technique to generate a first syntax element, a second syntax element, a third syntax element, a fourth syntax element, Decodes the element.

As an example, the thread allocation unit 610 allocates one thread (first thread) for decoding the first syntax element and one thread (second thread) for decoding the second syntax element and the third syntax element, ), And allocate one thread (third thread) for decoding the fourth syntax element and the fifth syntax element. That is, according to the second embodiment of the present invention, three threads can be allocated for performing the first decoding.

In this case, the decoding unit 620 may sequentially start the processing of the first thread, the processing of the second thread, and the processing of the third thread sequentially with a predetermined time gap, and thereby the first thread, the second thread, Decoding can be performed.

In other words, as shown in FIG. 13, the video decoding apparatus 600 allocates a first thread (Thread 1) to the decoding of the first syntax element included in the MBINFO group, And a second thread (Thread 2) for decoding a third syntax element included in the CBP group, and for decoding the fifth syntax element included in the fourth syntax element and the COEFF group included in the SIGMAP group, After the thread (Thread 3) is allocated, processing of the first thread, processing of the second thread, and processing of the third thread are sequentially started with a predetermined time gap (i.e., one sub-time interval) , The second thread and the third thread may be processed in parallel to perform the decoding in accordance with the pipeline technique. Here, the three sub time periods shown in FIG. 13 are also included in one time period described above.

Here, according to an embodiment of the present invention, the decoding unit 620 may process the first thread, the second thread, and the third thread, respectively, in units of sub-time intervals as shown in FIG. In this case, the decoding unit 620 starts processing of the second thread after one sub-time interval from the start of the processing of the first thread, and after a sub-time interval from the start of processing of the second thread, The processing of the section can be started.

On the other hand, when decoding is performed according to the pipeline technique as described above, if only one syntax element is processed in one sub-time interval, the overall decoding speed may be slowed down by increasing the number of times of synchronization.

Accordingly, in accordance with an embodiment of the present invention, the decoding unit 620 may process a plurality of syntax elements within one sub-time interval for each thread, thereby reducing the number of synchronization times to prevent a decrease in the decoding speed have. In other words, the decoding unit 620 may decode two or more first syntax elements in one sub-time interval, decode one or more second syntax elements and one or more third syntax elements in one sub-time interval, The fourth syntax element and the one or more fifth syntax elements may be decoded in one sub-time interval.

As another example, the thread allocation unit 610 may allocate one thread (first thread) for decoding the first syntax element and one thread (second thread) for decoding the second syntax element Allocate one thread (third thread) for decoding the third syntax element, and allocate one thread (fourth thread) for decoding the fourth syntax element and the fifth syntax element. That is, according to the third embodiment of the present invention, four threads can be allocated for performing the first decoding. Here, the four sub-time periods shown in FIG. 14 are also included in one time period described above.

In this case, the decoding unit 620 sequentially starts the processing of the first thread, the processing of the second thread, the processing of the third thread, and the processing of the fourth thread at a predetermined time interval, , The third thread and the fourth thread in parallel to perform decoding.

In other words, according to the second embodiment of the present invention, as shown in FIG. 14, the video decoding apparatus 600 allocates a first thread (Thread 1) to decoding of the first syntax element included in the MBINFO group Allocates a second thread (Thread 2) for decoding a second syntax element included in the PRED group, allocates a third thread (Thread 3) for decoding a third syntax element included in the CBP group, A fourth thread (Thread 4) is allocated for the decoding of the fourth syntax element included in the SIGMAP group and the fifth syntax element included in the COEFF group, processing of the first thread with a predetermined time gap, The processing of the third thread and the processing of the fourth thread may be sequentially started to parallelize the first thread, the second thread, the third thread, and the fourth thread to perform decoding according to the pipeline technique.

Here, according to an embodiment of the present invention, the decoding unit 620 may process the first thread, the second thread, the third thread, and the fourth thread, respectively, in units of sub-time intervals as described above. In this case, the decoding unit 620 starts processing of the second thread after one sub-time interval from the start of the processing of the first thread, and after a sub-time interval from the start of processing of the second thread, The processing of the fourth sub-time period can be started after one sub-time interval from the start of the processing of the third thread.

Also, according to an embodiment of the present invention, the decoding unit 620 may decode two or more first syntax elements in one sub-time interval and decode two or more syntax elements in one sub-time interval as described above , Two or more third syntax elements may be decoded within one sub-time interval, and one or more fourth syntax elements and one or more fifth syntax elements may be decoded in one sub-time interval. Accordingly, it is possible to solve the problem that the overall decoding speed is slowed down by processing only one syntax element in one sub-time interval and increasing the number of times of synchronization.

Accordingly, the image decoding apparatus 600 according to the present invention can efficiently parallel-decode data according to an image without a large overhead.

2. The reason why some threads in each time interval are allocated to the first decoding for the Nth image data and the remaining threads are allocated to the second decoding for the (N-1) th image data

In a case where image data is decoded by processing four threads in parallel using a quad-core processor including four cores, as shown in FIG. 15, the image decoding apparatus 600 first performs a first decoding including entropy decoding Four threads are assigned to four threads, and four threads are allocated to a second decoding including at least one of inverse quantization, inverse transform, intraprediction, motion compensation and deblocking, and parallel processing is performed to decode video data Can be performed.

 However, when decoding is performed in a manner as shown in FIG. 15 using a multicore processor including five or more cores such as a quad-core processor or a hexa core processor, as shown in FIGS. 16 and 17, (FIG. 16 shows an example of allocating and processing four threads for the first decoding using an octa core processor, and FIG. 17 shows an example of using a hexa core processor And four threads are assigned to the first decoding to perform processing). When there is a thread that is not allocated for a specific task, a problem arises in that data processing efficiency of the multicore processor is reduced.

On the other hand, in order to decode the image data as described above, after the first decoding (including entropy decoding) is preferentially performed (including at least one of inverse quantization, inverse transform, intraprediction, motion compensation, and deblocking) A second decoding has to be performed.

Therefore, in order to prevent the first decoding and the second decoding from being performed sequentially and the redundant thread to be present, as described with reference to FIGS. 7 and 8, a plurality of threads Some of the threads are allocated to the first decoding of the N-th image data for parallel processing, and the rest of the threads are allocated to the second decoding for the (N-1) -th image data for parallel processing.

Accordingly, the video decoding apparatus 600 according to the present invention can maximize the efficiency of decoding video data especially in a multi-core processor of a quad-core or more.

18 is a flowchart showing an overall flow of an image decoding method according to an embodiment of the present invention. Hereinafter, a process performed in each step will be described.

First, in step S1810, a plurality of threads are allocated to a task for decoding image data.

Next, in step S1820, the plurality of threads allocated in step S1810 are processed in parallel to decode the image data.

According to an embodiment of the present invention, the decoding of the image data may consist of a first decoding including entropy decoding and a second decoding performed after the entropy decoding. In this case, the second decoding may include at least one of inverse quantization, inverse transform, intraprediction, motion compensation, and deblocking.

Also, according to an embodiment of the present invention, a thread of a plurality of threads that are processed in parallel in the Nth time interval in step S610 is allocated to the first decoding for the Nth image data, The thread may be assigned to the second decoding for the (N-1) th image data.

Embodiments of the image decoding method according to the present invention have been described so far, and the configuration of the image decoding apparatus 600 described above with reference to FIG. 6 can be applied to this embodiment as it is. Hereinafter, a detailed description will be omitted.

In addition, embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Examples of program instructions, such as magneto-optical and ROM, RAM, flash memory and the like, can be executed by a computer using an interpreter or the like, as well as machine code, Includes a high-level language code. The hardware devices described above may be configured to operate as one or more software modules to perform operations of one embodiment of the present invention, and vice versa.

As described above, the present invention has been described with reference to particular embodiments, such as specific elements, and limited embodiments and drawings. However, it should be understood that the present invention is not limited to the above- Various modifications and variations may be made thereto by those skilled in the art to which the present invention pertains. Accordingly, the spirit of the present invention should not be construed as being limited to the embodiments described, and all of the equivalents or equivalents of the claims, as well as the following claims, belong to the scope of the present invention .

Claims (17)

A decoder for decoding image data by parallel processing a plurality of threads including a plurality of first threads and at least one second thread, the decoding unit including a hexa core processor or an octa core processor,
Wherein the decoding of the image data comprises a first decoding including entropy decoding and a second decoding including inverse quantization, inverse transform, intra prediction, motion compensation, and deblocking performed after the entropy decoding,
Wherein when the decoding section is constituted by a hex core processor, the number of threads is 6, the number of the first threads is 3 to 5, the number of the second threads is 3 to 1,
Wherein when the decoding unit is composed of an octa core processor, the number of threads is 8, the number of the first threads is 3 to 5, the number of the second threads is 5 to 3,
The decoding unit processes a plurality of threads in parallel in each of an (N-1) th time interval, an (N) -th time interval, and an (N + 1) -th time interval, Wherein the first thread performs the first decoding on the Nth video data in the Nth time interval and the second thread performs the first decoding on the Nth video data on the In the (N + 1) < th > time interval, the second thread performs the second decoding on the Nth video data,
Wherein the decoding of the (N-1) th image data is completed in the Nth time interval, and the decoding of the Nth image data is completed in the (N + 1) th time interval. delete delete The method according to claim 1,
Wherein the image data is image data encoded according to a syntax element division technique,
The first thread includes a first syntax element included in the MBINFO group, a second syntax element included in the PRED group, two or more threads allocated for decoding of the third syntax element included in the CBP group, and a second syntax element included in the SIGMAP group 4th syntax element and the fifth syntax element included in the COEFF group. 5. The method of claim 4,
Wherein the decoding unit decodes the first syntax element, the second syntax element, the third syntax element, the fourth syntax element, and the fifth syntax element by parallel-processing the first thread according to a pipeline technique And outputs the decoded image. The method according to claim 1,
Wherein the decoding unit parallel-processes the second thread using a data level parallelization technique. The method according to claim 1,
Wherein the image data is image data of a frame unit. delete There is provided an apparatus for decoding image data by performing first decoding including entropy decoding and second decoding including inverse quantization, inverse transform, intra prediction, motion compensation, and deblocking performed after the entropy decoding,
A hexa core processor or an octa core processor and performs parallel processing of a plurality of threads including a plurality of first threads and at least one second thread in each time interval in units of a predetermined time interval to perform decoding A decoding unit; And
And a thread allocation unit allocating the plurality of threads to be processed in parallel in each time interval,
Wherein when the decoding section is constituted by a hex core processor, the number of threads is 6, the number of the first threads is 3 to 5, the number of the second threads is 3 to 1,
Wherein when the decoding unit is composed of an octa core processor, the number of threads is 8, the number of the first threads is 3 to 5, the number of the second threads is 5 to 3,
Wherein the thread allocation unit allocates a plurality of threads in parallel in the (N-1) th time interval, the N-th time interval, and the (N + 1)
And allocating the first thread to perform the first decoding on the (N-1) th video data in the (N-1) th time interval, and allocating the first thread to the And allocates the second thread to perform the second decoding on the (N-1) th video data, and allocates the second thread to the N Th image data to perform the second decoding on the second image data,
Wherein the decoding of the (N-1) th image data is completed in the Nth time interval, and the decoding of the Nth image data is completed in the (N + 1) th time interval. delete delete delete 10. The method of claim 9,
Wherein the image data is image data encoded according to a syntax element division technique,
The first thread includes a first syntax element included in the MBINFO group, a second syntax element included in the PRED group, two or more threads allocated for decoding of the third syntax element included in the CBP group, and a second syntax element included in the SIGMAP group 4th syntax element and the fifth syntax element included in the COEFF group. 10. The method of claim 9,
Wherein the image data is image data of a frame unit. delete delete delete

Images