JP4594214B2 - Image decoding device - Google Patents

Description

  The present invention relates to an image decoding apparatus that performs high-speed decoding processing of encoded image data, for example.

  Conventionally, as a decoding device that decodes encoded stream data obtained by encoding a moving image using a compression encoding format such as MPEG (moving picture experts group), the encoded stream data is divided into a plurality of decoding units for each predetermined unit. One that performs sorting and decoding in parallel is known. Specifically, for example, in the decoding device shown in FIG. 11 of Patent Document 1, the decoder control unit 103 sequentially inputs stream data to a plurality of decoders 110, and each decoder 110 performs a decoding process on the input data. To do. In the decoding device shown in FIG. 4 of Patent Document 2, the row decoder control circuit 16 distributes and inputs the data accumulated in the stream buffer 14a to the plurality of row decoders 17 to 19, and converts the data into the input data. On the other hand, each of the row decoders 17 to 19 performs a decoding process in parallel.

In this type of image decoding device, each decoding unit outputs a completion signal when the decoding process of each assigned unit of encoded stream data is completed, and the control unit that receives the completion signal, A predetermined unit of encoded stream data is input to the decoding unit that has output the completion signal in the order in which the data is arranged. For example, as shown in FIG. 7 of Patent Document 2, when each of the row decoders 17 to 19 completes the decoding process for one row of data, a completion signal is sent to the row decoder control circuit 16, and the completion signal The row decoder control circuit 16 that has received the data inputs data for one row in the order in which the data are arranged to the row decoders 17 to 19 that have transmitted the completion signal. Also, the decoding device of Patent Document 1 performs the same processing with one slice as a predetermined unit.
JP 2002-325255 A (paragraphs 0003 to 0007, FIG. 11) JP 2003-32679 A (paragraphs 0035 to 0037, FIG. 4, paragraph 0077, FIG. 7, paragraph 0002)

  However, if data is input in the order in which the data is arranged to the decoding unit that has output the completion signal as described above, the time during which the other decoding units are in a pause state may become longer. For example, immediately after the last data in the frame is input to one of the decoding units, the decoding process of the other decoding unit is completed, and it takes a long time to decode the last data in the frame. . That is, the time taken for the decoding unit to be in a dormant state becomes longer, so that the time taken to decode the data of each frame becomes longer.

  In view of the above points, an object of the present invention is to further reduce the time required for decoding processing of encoded data in a decoding device that performs decoding processing in parallel by a plurality of decoding units.

In order to solve the above problems, the invention of claim 1
An image decoding device that decodes input encoded image data in parallel by a plurality of decoding units,
Distribution in which each slice constituting the input encoded image data is distributed to the decoding unit and decoded so that the total number of small unit data included in the slice decoded by each decoding unit is averaged A processing unit is provided.

The invention of claim 2
The image decoding device according to claim 1,
The distribution processing unit calculates the total number of small unit data included in the slices that have already been determined to be distributed for each decoding unit, and distributes the slices to be distributed to the decoding unit having the smallest total It is comprised so that it may do.

The invention of claim 3
The image decoding device according to claim 1,
The distribution processing unit is a decoding unit having the smallest total number of small unit data included in slices that have already been determined to be distributed since each slice to be distributed has a large number of small unit data included. It is comprised so that it may distribute to.

The invention of claim 4
The image decoding device according to claim 1,
An average value calculation unit that calculates an average value obtained by dividing the total number of small unit data included in the slice to be distributed by the number of decoding units;
The distribution processing unit sequentially distributes each slice to each decoding unit until the number of small unit data included in the slices already determined to be distributed to the decoding unit exceeds the average value. It is characterized by that.

The invention of claim 5
The image decoding device according to claim 1,
The small unit data is 1-bit data.

The invention of claim 6
The image decoding device according to claim 1,
The encoded image data has been subjected to compression encoding according to the MPEG standard .
Upper Symbol small unit data is characterized by a coded data of each macro block.

As a result, the decoding processing time of each decoding unit is averaged by averaging the total number of small unit data included in the slices decoded by each decoding unit. Therefore, the time for each decoding unit to be in a pause state is shortened, and the time required for the decoding process of the encoded image data is shortened .

Also, the invention of claim 7,
An image decoding device that decodes input encoded image data in parallel by a plurality of decoding units,
For each slice of the current screen constituting the input encoded image data, the number of small unit data included in the slice corresponding to the slice of the current screen among the slices of the previous screen is specified, and the current screen A distribution processing unit that distributes and decodes each slice to the decoding unit so that the total number of small unit data specified for the slice of the current screen decoded by each decoding unit is averaged It is characterized by.

Thereby, the number of small unit data included in the slice of the previous screen corresponding to the slice decoded by each decoding unit is averaged, so that the decoding processing time of each decoding unit is also averaged. Therefore, the time for each decoding unit to be in a pause state is shortened, and the time required for the decoding process of the encoded image data is shortened.

  ADVANTAGE OF THE INVENTION According to this invention, in the decoding apparatus which performs a decoding process in parallel by a some decoding part, the time concerning the decoding process of the encoded data can be shortened more.

  Hereinafter, embodiments of a decoding device according to the present invention will be described with reference to the drawings. In each of the following embodiments, components having functions similar to those of the other embodiments are denoted by the same reference numerals and description thereof is omitted.

<< Data structure of MPEG2 bit stream >>
First, the data structure of an MPEG2 bit stream that is a decoding target of the decoding apparatus according to the present embodiment will be described with reference to FIGS.

  In the present embodiment, a case where the present invention is applied to a decoding apparatus that decodes encoded image data called an MPEG2 bit stream compressed and encoded by the MPEG2 method will be described. However, encoding is possible by any image encoding method such as MPEG1 or MPEG4. The present invention can be similarly applied to the case where the encoded image data is decoded.

  The data structure of the MPEG2 bit stream is determined in detail so that the decoding apparatus can correctly receive the data. Specifically, as shown in FIG. 1, a sequence, GOP (Group of Picture), It has a hierarchical structure of pictures, slices, macroblocks, and blocks.

  A sequence includes one or more GOPs and is generally encoded signal data of one entire video program. This sequence begins with a sequence header and ends at the end of the sequence. The sequence header includes information related to the entire sequence, such as information indicating the size of the image, the number of frames to be encoded per second, and the communication speed. Further, following the sequence header, function expansion information indicating that it is an MPEG2 encoded signal is inserted, and an input image signal format or the like is designated.

  GOP is an I picture that can be encoded within a frame, a P picture that performs forward motion compensation using only past frames, and a B picture that performs bidirectional motion compensation using both past and future frames. Of these, one or more pictures are included. Note that an I picture is always inserted as the first picture of each GOP. The GOP header includes time stamp information for enabling time adjustment with audio or the like at the time of image restoration.

  A picture includes one or more slices and is data for one screen (m pixels × n lines) constituting a moving image signal, as shown in FIG. The picture header of this picture includes information for identifying I, P, and B pictures, information for specifying the display order of each picture, and the like. Further, the function extension information following the picture header includes information specifying a function introduced in MPEG2, such as setting of a frame structure / field structure.

  A slice is composed of one or more macroblocks arranged in a line in the left-right direction on the screen, and the slice information includes coding information used in the slice such as information representing quantization characteristics. Note that the arrangement of macroblocks is one column in the left-right direction so that the influence of errors on the communication line does not reach the vertical direction on the screen, and there is no more than two columns.

  As shown in FIG. 2, the macro block is composed of four 8 × 8 Y signal blocks, one 8 × 8 Cr signal block, and one 8 × 8 Cb signal block 1 in the 4: 2: 0 format. It is composed of pieces. Note that the macroblock information includes macroblock address increment (MBAI: indicates the number of macroblocks + 1 from the left end of the image + 1), horizontal and vertical motion vectors for performing encoding control in units of macroblocks, Information such as a coded block pattern (CBP: encoded based on whether or not six blocks are significant blocks) is included.

  As shown in FIG. 2, the block is composed of DCT coefficient data of any of 8 × 8 Y signal, Cr signal, and Cb signal in the 4: 2: 0 format. The DCT coefficient data is continuous variable-length encoded data. At the end of each block, an EOB (End of Block) code is added.

  As described above, the MPEG2 bitstream data has a hierarchical structure including six layers. In each layer of the sequence, GOP, picture, and slice, a 32-bit start code as shown in FIG. 3 is inserted, which is arranged in byte units (arranged at bit positions that are multiples of 8 from the top of each layer). .

  This bit pattern of the start code never occurs at a position other than the start code position in the MPEG2 bit stream. For this reason, the decoding device that has received the MPEG2 bitstream can perform decoding processing in parallel in a plurality of processing systems for each slice unit, for example, by detecting the start code.

  Although FIG. 2 shows the data structure at the time of 4: 2: 0 formatting, the present invention is not limited to this, and the same applies to the format at 4: 2: 2. Can do.

Embodiment 1 of the Invention
(Configuration of decoding device)
As shown in FIG. 4, the decoding apparatus according to Embodiment 1 of the present invention includes a start code detection unit 11, an external buffer control unit 12, an external buffer 13, an encoded data distribution processing unit 14 (distribution processing unit), and decoder control. A unit 15, a plurality of decoding units (decoders) 16, and an output memory 17 are provided. The external buffer 13 includes a stream buffer 13a and a distribution processing information buffer (start code buffer) 13b. The external buffer 13 is a memory such as a DRAM (Dynamic Random Access Memory).

  The decoder 16 includes a variable length decoding unit 16a, an inverse quantization unit 16b, an inverse DCT unit 16c, and a motion compensation unit 16d.

  As shown in FIG. 5, the encoded data distribution processing unit 14 includes a macroblock number (MB number) counting unit 14a and a distribution order determining unit 14b.

(Outline of the operation of the decoding device)
First, an MPEG2 bit stream (hereinafter referred to as encoded stream data) input from the outside of the decoding apparatus is stored in the stream buffer 13a. Next, the encoded data distribution processing unit 14 determines a decoder 16 for decoding the accumulated encoded stream data for each slice. More specifically, first, start code information indicating a write position of each start code stored in the start code buffer 13b in the stream buffer 13a is read, and one slice after each start code in which the write position is specified from the start code information. Minute encoded stream data (each slice) is analyzed. Then, the number of macroblocks included in each slice is calculated, and for each slice, the decoder 16 having the smallest total number of macroblocks of slices that have already been determined to be distributed is determined as the decoder 16 to be decoded. Then, decoding processing is executed in the decoder 16 for each slice unit on the encoded stream data under the control of the decoder control unit 15, and the decoded image data is stored in the output memory 17.

(Main operations of each part of the decoding device)
The start code detection unit 11 supplies encoded stream data input from the outside and start code information described below to the external buffer control unit 12. More specifically, the start code of each slice of the encoded stream data shown in FIG. 3 is detected, and the start code including information indicating the position where the start code is written in the stream buffer 13a from the detected start code. Information is generated, and the encoded stream data and start code information are output to the external buffer control unit 12.

  The external buffer control unit 12 writes the encoded stream data and start code information input from the start code detection unit 11 in the stream buffer 13a and the start code buffer 13b of the external buffer 13, respectively.

  The external buffer 13 (stream buffer 13a and start code buffer 13b) holds encoded stream data and start code information input from the external buffer control unit 12.

  The MB number counting unit 14a of the encoded data distribution processing unit 14 counts the number of macroblocks included in each slice of the encoded stream data and outputs it to the distribution order determining unit 14b. The start code information of each slice is read from the start code buffer 13b, and the encoded stream data (each slice) for one slice after the start code whose write position is specified is read and analyzed from each start code information, and is analyzed for each slice. The number of included macroblocks is counted, and the number of macroblocks in each slice is output to the distribution order determining unit 14b.

  The distribution order determination unit 14b of the encoded data distribution processing unit 14 determines which decoder 16 of the plurality of decoders 16 decodes each slice based on the number of macroblocks of each slice input from the MB number counting unit 14a. The determination result is output to the decoder control unit 15. A specific determination operation will be described in detail later.

  The decoder control unit 15 distributes the encoded stream data of the stream buffer 13a to the decoder 16 for each slice unit based on the determination result. Further, the start code information of the slice to be distributed is read from the start code buffer 13b, and information indicating the parameter of the picture layer of the slice and the write position in the stream buffer 13a is sent to the decoder 16 determined by the distribution order determination unit 14b. Output. The parameters of the picture layer are parameters common to one image, specify the type of picture (I / P / B picture) and picture structure (frame / field), and are used for the operation of the decoder. .

  The decoder 16 reads the encoded stream data from the stream buffer 13a via the external buffer control unit 12 for each slice based on the information indicating the writing position of the encoded stream data input from the decoder control unit 15 for each slice. . Then, according to the picture layer parameters input from the decoder control unit 15, the read encoded stream data in units of slices is decoded and output to the output memory 17. When the decoding process is completed, a completion notification is sent to the decoder control unit 15. This decoding process is performed by the variable length decoding unit 16a, the inverse quantization unit 16b, the inverse DCT unit 16c, and the motion compensation unit 16d. Hereinafter, processing of each processing unit in the decoder 16 will be described.

  The variable length decoding unit 16a decodes the variable length encoded data included in the encoded data to restore the quantized DCT coefficient. More specifically, first, each slice is read from the stream buffer 13a via the external buffer control unit 12 based on information indicating the writing position of the encoded stream data input from the decoder control unit 15 for each slice. Then, the macroblock is separated from the read slice. Further, the quantized DCT coefficient of each macroblock is decoded, and the decoded quantized DCT coefficient is output to the inverse quantization unit 16b. When decoding the variable-length encoded data, a predetermined decoding table is referred to. Note that the variable length decoding unit 16a also decodes parameters such as the prediction mode and the prediction vector, and outputs the decoded prediction mode and prediction vector to the motion compensation unit 16d.

  The inverse quantization unit 16b inversely quantizes the quantized DCT coefficient input from the variable length decoding unit 16a, decodes it to a DCT coefficient, and outputs the decoded DCT coefficient to the inverse DCT unit 16c. In this inverse quantization, a predetermined inverse quantization table is referred to as in the decoding of the variable length encoded data.

The inverse DCT unit 16c performs inverse DCT transform on the DCT coefficient input from the inverse quantization unit 16b, decodes the pixel data before encoding, and outputs the decoded pixel data to the motion compensation unit 16d. Based on the pixel data input from the inverse DCT unit 16c and the prediction mode and prediction vector input from the variable length decoding unit 16a, motion compensation is performed, and the pixel data subjected to motion compensation is output to the output memory 17. Write.

  The pixel data written in the output memory 17 is used for display output and as reference data for other images. That is, when the macroblock input from the inverse DCT unit 16c uses motion compensation, the motion compensation unit 16d determines that the pixel data has luminance according to the prediction vector input from the variable length decoding unit 16a. If it is data, the reference pixel is read from the luminance buffer of the output memory 17, and if the pixel data is color difference data, the reference pixel data is read from the color difference buffer of the output memory 17. Then, motion compensation is performed by adding the read reference pixel data to the pixel data input from the inverse DCT unit 16 c, and the pixel data subjected to the motion compensation is written in the output memory 17.

(Distributed processing of encoded stream data)
Next, the operation of distributing the encoded stream data for one frame by the encoded data distribution processing unit 14 in the decoding apparatus according to the present embodiment will be described with reference to FIG. In addition, while the decoding unit to be distributed is determined, a decoding process according to the determination result may be performed in parallel, or after all the distribution of data for one frame is determined, the decoding for the frame is performed. Processing may be initiated.

(S101) In the encoded data distribution processing unit 14, each register is initialized. Here, L is a value of a register indicating the number of macroblocks of a slice in which it is determined which decoder 16 is distributed. M1 to _{Mn to MN} are register values indicating the accumulation of the number of macroblocks distributed to each of the N decoders 16.

  (S102) The start code of each slice is sequentially input from the head of the encoded stream data to the MB number (number of macroblocks) counting unit 14a, and analyzed by the MB number (macroblock number) counting unit 14a. The number of included macroblocks is calculated.

  (S103) The number of macroblocks of the slice distributed in (S105) is input to L.

(S104) The distribution order determination unit 14b detects the decoder with the smallest cumulative number of macroblocks that have already been determined to be processed, that is, the decoder 16 with the smallest value of _Mn . When there are a plurality of decoders 16 having the smallest value of M _n , for example, the decoder 16 having the smallest value of n is detected. Here, it is assumed that the decoder 16 with the smallest _Mk and n = k is detected.

(S105) It is determined that the slice is distributed to the decoder 16 detected in S104, and the number of macroblocks included in the slice is added to the value of the register indicating the accumulation of the number of macroblocks to be decoded by the decoder 16. The That is, L is added to _Mk .

  (S106) Whether or not there is a slice to be distributed in the frame is determined. If slices remain, the processes of S103 to S105 are performed for the remaining slices, and if not, the process ends.

  With the above operation, as an example of the distribution determination result, a decoder that performs decoding processing is determined as shown in FIG.

  While the distribution processing operation is being performed or after the distribution processing is completed, the encoded stream data is output to the decoder for each slice according to the distribution determination result, and the decoding operation is performed by each decoder.

When there are a plurality of decoders 16 having the smallest value of _Mn in S104, the decoder 16 having the smallest value of n is selected, but any of them may be selected.

(effect)
As described above, according to the decoding apparatus according to the first embodiment, the number of macroblocks of each slice calculated by start code analysis is used, and the encoded data distribution processing unit 14 performs the decoding order of slices in the decoder 16. Is determined so that the number of macroblocks of data processed by each decoder is averaged, the processing time of each decoder 16 can be averaged, thereby speeding up the decoding process. . That is, by averaging the processing times of a plurality of decoders, the time during which some of the decoders are processing but some of the decoders are idle (t0 is shown as an example in FIG. 8). As a whole, the time required for the decoding process is shortened.

<< Embodiment 2 of the Invention >>
(Outline of decoding device)
The decoding apparatus according to the second embodiment of the present invention is characterized in that each slice is distributed to the decoder 16 that has issued a completion notification indicating that the processing for the previous slice is completed, since the number of bits included is large. It is said. The configuration is obtained by changing the configuration of the encoded data distribution processing unit 14 in the decoding apparatus according to the first embodiment to the configuration of the encoded data distribution processing unit 24 shown in FIG. The encoded data distribution processing unit 24 includes a bit number counting unit 24a that counts the number of bits included in each slice, instead of the MB number counting unit 14a of the encoded data distribution processing unit 14.

(Main operations of the encoded data distribution processing unit 24)
The bit number counting unit 24a of the encoded data distribution processing unit 24 reads information indicating the position where the start code of each slice is written to the stream buffer 13a from the start code buffer 13b, and codes for one slice after each start code. Stream data (each slice) is analyzed, the number of bits of each slice is counted, and the number of bits of each slice is output to the distribution order determination unit 24b.

The distribution order determination unit 24b of the encoded data distribution processing unit 24 performs the following operation for each frame of data.
Based on the number of bits input from the bit number counting unit 24 a, it is determined in which order the decoder 16 is to decode each slice, and the determination result is output to the decoder control unit 15.

(Distributed processing of encoded stream data)
Next, the operation of distributing the encoded stream data for one frame by the encoded data distribution processing unit 24 in the decoding apparatus according to the present embodiment will be described with reference to FIG. Here, a register value D indicating which of the plurality of decoders 16 is to decode the slice, a register value M or L in which the number of bits of a predetermined slice is set, and all decoders provided in the decoding device This will be described using a value N indicating the number.

  (S201) In the encoded data distribution processing unit 24, the registers are initialized. Here, since slices are allocated from the decoder 1, D = 1 is set.

  (S202) The start code is sequentially input from the head of the encoded stream data to the bit number counting unit 24a, and the number of bits included in each slice is calculated by analysis in the bit number counting unit 24a.

  (S203) The distribution order determination unit 24b rearranges the slices in the order of the number of bits calculated in (S202).

  (S204) It is determined whether or not there remains a slice that has not been determined by which decoder to be decoded (distribution has not been determined). If it remains, the process proceeds to (S205), and if not, the process ends.

  (S205) Of the slices for which distribution is not determined, the slice having the maximum number of bits is distributed to the decoder D.

  (S206) If D is equal to N, proceed to (S208), otherwise proceed to (S207).

  (S207) 1 is added to D, and the process returns to (S204).

  (S208) Wait for a completion notification from one of the decoders 16. If there is a completion notification from any of the decoders 16 (S209), the process proceeds.

  (S209) The value of D is a number indicating the decoder 16 that has notified completion.

  (S210) It is determined whether a slice whose distribution has not been determined remains in the frame. If it remains, the process proceeds to (S211), and if not, the process ends.

  (S211) Of the slices for which distribution is not determined, the slice having the maximum number of bits is distributed to the decoder D, and the process returns to (S208).

  By the above operation, as shown in FIG. 11 as an example of the determination result, the decoder that performs the decoding process can be determined.

  In this embodiment, D = 1 is initialized so that slices are allocated from the decoder 1, but it may be set so that slices are allocated from other decoders.

  The initialization in (S201) may be performed after (S202) or (S203).

  In addition, the rearrangement in (S203) may be performed by a method in which, for example, a bit indicating the reading order is added to the head of the start code information, and then slices are read in that order. In such a method, the slices do not need to be recorded again in another memory in the order after rearrangement.

(effect)
As described above, according to the decoding device according to the second embodiment, the encoded data distribution processing unit 14 uses the number of bits of each slice calculated by the analysis of the start code to determine the slices in the stream in the decoder 16. By determining the decoding order so that the number of bits of data processed by each decoder is averaged, the processing time of each decoder 16 can be averaged, thereby speeding up the decoding process. it can. That is, by averaging the processing times of a plurality of decoders, the time during which some of the decoders are processing but some of the decoders are idle (t0 is shown as an example in FIG. 8). As a whole, the time required for the decoding process is shortened.

<< Embodiment 3 of the Invention >>
(Configuration of decoding device)
As shown in FIG. 12, the decoding apparatus according to Embodiment 3 of the present invention includes a previous frame decoding information detection storage unit 38 that stores information of a frame that has been subjected to decoding processing. This is different from the decoding apparatus according to the second embodiment. The encoded data distribution processing unit 34 does not include the MB number (macroblock number) counting unit 14a or the bit number counting unit 24a for analyzing the start code, and the distribution order determining unit 34b is a frame to be decoded. Based on which decoder each slice of (current screen) is decoded in which order, the time actually required to decode the slice corresponding to the slice of the previous frame held in the previous frame decoding information detection storage unit 38 And the determination result is output to the decoder control unit 15.

  The previous frame decoding information detection storage unit 38 detects the decoding time of each slice and stores it for one frame, and supplies encoded data from the decoder control unit 15 to the decoder 16 as shown in FIG. The timer unit 38a that counts the time from the start of the process until the decoder 16 notifies the decoder control unit 15 of the completion, and the storage buffer 38b that stores the slice number and the count time of each slice. Here, the slice number is a number for identifying a slice. For example, when the start code of the slice is detected, the slice number may be sequentially given to the head of the start code information.

(Operation of Previous Frame Decoding Information Detection Storage Unit 38)
Next, the operation of the previous frame decoding information detection storage unit 38 in the decoding device according to the present embodiment will be described with reference to FIG. Here, a description will be given using a number n for identifying a plurality of slices in the previous frame, and a register value T _n indicating the cumulative decoding time of the decoder after decoding for the slice indicated by n is started. This operation is performed for each frame. The previous frame decoding information detection storage unit 38 is provided for each decoder.

(S301) The register is initialized. Specifically, T _n = 0 is set.

  (S302) The decoder control unit 15 determines whether or not the transfer of the encoded stream data to the decoder 16 has been started.

(S303) If the transfer is started, the decoding process time is counted by the timer unit 38a. Specifically 1 to T _n are added.

  (S304) The timer unit 38a determines whether or not there is a completion notification from the decoder 16 to the decoder control unit 15. Until there is a completion notification, the process of (S303) and the determination of (S304) are repeated.

(S305) If there is a completion notification, the count of the decoding processing time ends, the count result (T _n ), for example, a value indicating the slice number and the processing time is written in the storage buffer 38b, and the processing ends.

(Distributed processing of encoded stream data)
Next, the operation of distributing the encoded stream data for one frame by the encoded data distribution processing unit 34 in the decoding apparatus according to the present embodiment will be described with reference to FIG. Here, a value SUM indicating the total decoding processing time in the previous frame, a decoding processing time M in the previous frame already distributed to a predetermined decoder, a DIV indicating the average decoding processing time in the previous frame of each decoder, a plurality of decoders Description will be made using a value D for identifying 16 and a value K for identifying a slice. The total number of slices is S, and the total number of decoders is N. Further, the decoding time of each slice in the previous frame held in the coded data distribution processing unit 34 and T _K.

  (S401) SUM, M, D, K, and DIV are initialized.

  (S402) For the previous frame, the sum (SUM) of decoding processing times of the slices counted by the timer unit 38a is calculated.

  (S403) The ideal average processing time (DIV) in each decoder 16 is calculated by dividing the sum obtained in (S402) by the number of decoders N.

(S404) The decoding processing time T _{K of the} slice identified by K in the previous frame is added to M.

  (S405) It is determined whether M indicating the cumulative decoding processing time of the decoder D has exceeded the ideal average processing time (DIV). If it is determined that it has been exceeded, the process proceeds to (S406), and if it is determined that it has not been exceeded, the process proceeds to (S408).

  (S406) It is determined whether D = N. If D = N, the process proceeds to (S408), and if D ≠ N, the process proceeds to (S407).

  (S407) 1 is added to D so that the decoder that performs the decoding process on the slice is changed to another decoder. Further, M indicating the cumulative decoding processing time is returned to 0.

  (S408) It is determined that the Kth slice is distributed to the decoder D.

  (S409) 1 is added to K.

  (S410) If the value of K is larger than the value of the total number of slices S, the process ends. If not, the process returns to (S404) and the distribution process is continued.

  With the above operation, as shown in FIG. 16 as an example of the distribution processing result, the decoder that performs the decoding processing is determined.

  In (S410), the presence / absence of a slice to be distributed is determined based on the value in the register K, but may be determined based on information included in the slice.

  Further, the example in which the storage buffer 38b is included in the decoding device has been described, but an external buffer or the like may be used.

(effect)
As described above, according to the decoding apparatus according to the third embodiment, the decoding time of each slice in the previous frame is stored, and the encoded data distribution processing unit 14 in the stream in the decoder 16 stores the decoding time based on the decoding time. By changing the decoding order of the slices, the processing time of each decoder 16 can be averaged, thereby speeding up the decoding process. That is, by averaging the processing times of a plurality of decoders, the time during which some of the decoders are processing but some of the decoders are idle (t0 is shown as an example in FIG. 8). As a whole, the time required for the decoding process is shortened.

<Others>
In the decoding apparatus according to the first embodiment, each slice is distributed to the decoder 16 having the smallest total number of macroblocks of slices that have already been determined to be distributed. Instead, the time and the number of bits actually taken for decoding the slice of the previous frame corresponding to the slice may be used. For example, each slice may be distributed to the decoder 16 having the smallest total number of bits of slices that have already been determined to be distributed.

  In the decoding apparatus according to the second embodiment, each slice is distributed to the decoder that has notified the completion because the number of bits included is large, but is already distributed to each decoder. It is also possible to calculate the total number of bits of the determined slices and distribute each slice to the decoder having the smallest total number from the one having the largest number of bits. In this case, while the decoding unit to be distributed is determined, the decoding process according to the determination result may be performed in parallel, or after all the distribution of the data for one frame is determined, The decryption process may be started. Also, instead of the number of bits, the time actually taken for decoding the slice of the previous frame corresponding to the number of macroblocks and slices may be used. For example, each slice may be distributed to the decoder that has notified the completion from the one having a large number of macroblocks.

  Further, in the decoding device of the third embodiment, the decoding processing time of the previous frame is averaged, but the number of macroblocks calculated by the MB number counting unit 14a of the first embodiment and the bit number counting unit 24a of the second embodiment. You may comprise so that the calculated bit number may be averaged. Further, the number of macroblocks and the number of bits in the previous frame may be averaged.

  In the first, second, and third embodiments, the distribution process of the encoded stream data for one frame to the decoder has been described. However, the same process may be performed in units of fields instead of in units of frames. Further, in the decoding device of the third embodiment, the decoding processing time of the previous frame is averaged, but the decoding processing time of the previous field may be averaged.

  Further, although different distribution procedures have been described in the first, second, and third embodiments, any distribution procedure may be used in the decoding device according to the first, second, and third embodiments, and a plurality of procedures may be combined. It may be used.

  The image decoding apparatus according to the present invention has an effect of shortening the time required for decoding the encoded data in a decoding apparatus that performs decoding processing in parallel by a plurality of decoding units. The present invention is useful as an image decoding device that performs high-speed data decoding processing.

It is explanatory drawing which shows the data structure of an MPEG2 bit stream. It is explanatory drawing which shows the correspondence of an MPEG2 bit stream and a screen. It is explanatory drawing which shows the start code of an MPEG2 bit stream. 1 is a block diagram illustrating a configuration of a decoding device according to Embodiment 1. FIG. 3 is a block diagram showing a configuration of an encoded data distribution processing unit 14 in the same manner. FIG. 4 is a flowchart showing an operation of distribution processing of encoded stream data by the encoded data distribution processing unit 14; FIG. 6 is an explanatory diagram illustrating an example of a distribution processing result by the encoded data distribution processing unit 14; It is explanatory drawing which shows time t0 of a state in which some decoders are processing, but some decoders are dormant. 6 is a block diagram illustrating a configuration of an encoded data distribution processing unit 24 of a decoding device according to Embodiment 2. FIG. 4 is a flowchart showing an operation of encoded stream data distribution processing by the encoded data distribution processing unit 24. FIG. FIG. 6 is an explanatory diagram illustrating an example of a distribution processing result by the encoded data distribution processing unit 24; It is a block diagram which shows the structure of the decoding apparatus which concerns on Embodiment 3. 3 is a block diagram showing a configuration of a previous frame decoded information detection storage unit 38. FIG. 4 is a flowchart showing the operation of the previous frame decoded information detection storage unit 38. FIG. 4 is a flowchart showing an operation of a distribution process of encoded stream data by the encoded data distribution processing unit. 4 is an explanatory diagram showing an example of a distribution processing result by the encoded data distribution processing unit 34. FIG.

11 Start Code Detection Unit 12 External Buffer Control Unit 13 External Buffer 13a Stream Buffer 13b Start Code Buffer (Distribution Processing Information Buffer)
14 Coded data distribution processing unit 14a MB number (number of macroblocks) counting unit 14b Distribution order determining unit 15 Decoder control unit 16 Decoder 16a Variable length decoding unit
16b Inverse quantization unit
16c Reverse DCT section
16d motion compensation unit
17 Output Memory 24 Encoded Data Distribution Processing Unit 24a Bit Number Counting Unit 24b Distribution Order Determination Unit 34 Encoded Data Distribution Processing Unit 34b Distribution Order Determination Unit 38 Previous Frame Decoding Information Detection Storage Unit 38a Timer Unit 38b Storage Buffer

Claims (7)

An image decoding device that decodes input encoded image data in parallel by a plurality of decoding units,
Distribution in which each slice constituting the input encoded image data is distributed to the decoding unit and decoded so that the total number of small unit data included in the slice decoded by each decoding unit is averaged An image decoding apparatus comprising a processing unit. The image decoding device according to claim 1,
The distribution processing unit calculates the total number of small unit data included in the slices that have already been determined to be distributed for each decoding unit, and distributes the slices to be distributed to the decoding unit having the smallest total An image decoding apparatus configured to The image decoding device according to claim 1,
The distribution processing unit is a decoding unit having the smallest total number of small unit data included in slices that have already been determined to be distributed since each slice to be distributed has a large number of small unit data included. An image decoding apparatus configured to be distributed to The image decoding device according to claim 1,
An average value calculation unit that calculates an average value obtained by dividing the total number of small unit data included in the slice to be distributed by the number of decoding units;
The distribution processing unit sequentially distributes each slice to each decoding unit until the number of small unit data included in the slices already determined to be distributed to the decoding unit exceeds the average value. An image decoding apparatus characterized by that. The image decoding device according to claim 1,
The image decoding apparatus according to claim 1, wherein the small unit data is 1-bit data. The image decoding device according to claim 1,
The encoded image data has been subjected to compression encoding according to the MPEG standard .
Upper Symbol small unit data, the image decoding apparatus which is a coded data of each macro block. An image decoding device that decodes input encoded image data in parallel by a plurality of decoding units,
For each slice of the current screen constituting the input encoded image data, the number of small unit data included in the slice corresponding to the slice of the current screen among the slices of the previous screen is specified, and the current screen A distribution processing unit that distributes and decodes each slice to the decoding unit so that the total number of small unit data specified for the slice of the current screen decoded by each decoding unit is averaged An image decoding apparatus characterized by the above.

Images