JP5182285B2 - Decoding method and decoding apparatus - Google Patents

Description

  The present invention relates to a decoding method and a decoding apparatus, and more particularly to a decoding method and a decoding apparatus for expanding image data compressed by a moving image compression method for performing inter-frame prediction.

  Usually, in a moving image, image information similar in time is similar to each other, and therefore an image close in time can be used as an original image of the current processing region. Such a concept is adopted in a moving picture compression method that performs inter-frame prediction represented by standards such as MPEG-2, MPEG-4, and H.264.

  In a decoding device that decompresses data compressed by a moving image compression method that performs inter-frame prediction, image data that has been decoded once is usually stored in an external storage device, that is, an area necessary for processing another frame, that is, The image data of the reference image is read from the external storage device and used.

  However, in recent compression schemes such as H.264, the reference image read unit has been minimized, and the number of reference image reads from the external storage device and the amount of read image data as a whole have also increased. For example, in H.264, the number of times of reading is 16 times at the maximum and the amount of read image data as a whole is 5 times at maximum as compared with MPEG-2.

As a method of reducing access to the external storage device, a detection method is selected according to a motion vector of a block located at a corner of a decoded macroblock, and a motion vector is detected by the selected method. There is one that generates a predicted image of a motion compensation target block based on (see Patent Document 1).
JP 2004-194283 A

  In recent compression methods such as H.264, in order to handle high-definition images, the number of times of reference image reading from the external storage device and the amount of read image data are increasing, and in order to satisfy the required performance of the standard, It is necessary to take measures such as increasing the operating frequency of the memory controller or the control unit for controlling access to the external storage device, or increasing the operating frequency or the number of the external storage device. For this reason, there has been a problem that the circuit scale of the decoding device increases, and the manufacturing cost and power consumption increase accordingly.

  An object of the present invention is to provide a decoding method and a decoding apparatus capable of reducing the number of accesses to a memory and the amount of accessed data, reducing the circuit scale, and reducing the manufacturing cost and power consumption. To do.

The above-described problem is a decoding method in which a video stream compressed by a moving image compression method for performing inter-frame prediction is read from a storage device, and a decoded frame is used as a reference image for decoding processing, which is extracted from the video stream. Whether the reference image is calculated based on the motion vector information, a predicted image is generated based on the reference image, and each region in the processing unit of the reference image is read from the storage device by integrating a plurality of regions. Is determined by comparing the bus occupancy time when reading the multiple areas together with the bus occupancy time when reading each area individually, and the bus occupancy time when reading each area individually is , Calculated by adding the read cycle of each area and the overhead for each access of the storage device, and reading the integrated multiple areas Bus occupation time to case, calculated by adding the overhead of accessing the storage device and read cycle of integration areas, short plurality towards the bus occupation time of reading by integrating the plurality of regions If it is determined that read by integrating the area can be achieved by decoding method characterized in that also be read out simultaneously a plurality of regions to be integrated from the storage device if the amount of data read from the storage device increases .

The above problem is a decoding device that reads a video stream compressed by a moving image compression method for performing inter-frame prediction from a storage device, and uses the decoded frame as a reference image for decoding processing, and is extracted from the video stream. A calculation unit that calculates the reference image based on the motion vector information, a generation unit that generates a prediction image based on the reference image, and a plurality of regions within each processing unit of the reference image are integrated. Including an integrated determination unit that determines whether to read from the storage device by comparing the bus occupation time when the plurality of areas are read together and the bus occupation time when each area is read individually; The integrated determination unit determines the bus occupancy time when each area is individually read out as an overhead for each area read cycle and each access of the storage device. A bus occupancy time when the plurality of areas are read together and calculated by adding a read cycle of the integrated area and an access overhead of the storage device. If it is determined that the bus occupancy time in the case of reading and integrating the plurality of areas is shorter and the plurality of areas are read together, the plurality of areas to be integrated are also selected even when the amount of data read from the storage device increases. It can be achieved by the decoding apparatus characterized by be read out simultaneously from the storage device.

  ADVANTAGE OF THE INVENTION According to this invention, the decoding method and decoding apparatus which can reduce the frequency | count of access to memory and the amount of data accessed, hold down the scale of a circuit, and can reduce manufacturing cost and power consumption are realizable.

It is a block diagram which shows the structure of the decoding apparatus in one Example of this invention. It is a figure explaining integration of a read-out area. It is a timing chart explaining the access timing of DDR-SDRAM. It is a figure explaining the occupancy time of the bus | bath when a block is read separately, and reading is integrated. It is a figure which shows the area | region already stored in the integrated data memory | storage part, and the area | region required by the present process in the inter-frame prediction part. It is a figure explaining reuse of the data shown in FIG. It is a block diagram explaining reuse of integrated data. It is a block diagram explaining reuse of integrated data. It is a block diagram explaining reuse of integrated data. It is a block diagram explaining reuse of integrated data. It is a flowchart explaining the integrated reading control in an Example. It is a flowchart explaining an integrated determination process. It is a flowchart explaining a data reading range determination process. It is a flowchart explaining a copy process. It is a flowchart explaining a request process. It is a flowchart explaining the read-out control when not diverting data. It is a flowchart explaining the request process when not diverting data.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Decoding apparatus 11 Stream input part 12 Decoding process part 13 Display control part 14 SDRAM
21 Entropy decoder 22 Coefficient decoder 23 Inter-frame prediction unit 24 Image generation unit 25 Read prediction image analysis unit 26 Integrated data storage unit 27 Memory controller 30 External storage device

  The present invention reads a video stream compressed by a moving image compression method for performing inter-frame prediction and stored in a storage device, and decodes an encoded frame using motion prediction. Used as a reference image for decoding processing. A reference image is calculated based on the motion vector information extracted from the video stream, a predicted image is generated based on the reference image data read from the storage device, and a read cycle of each area or each access of the storage device The shape of the reference image is determined based on the overhead or the occupied time of the bus used to access the storage device.

  When the reference image data is read, it is determined whether or not each of the areas in the data processing unit is read out in combination with the data of other areas, and the reference image is determined to be read out in an integrated manner. Sometimes, it may include integration determination for simultaneously reading the data of the areas to be integrated from the storage device. In this case, the integrated determination is performed based on the read cycle of each area, the overhead for each access of the storage device, or the bus occupation time.

  Embodiments of a decoding method and a decoding apparatus according to the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of a decoding apparatus according to an embodiment of the present invention. The decoding apparatus 1 includes a stream input unit 11, a decoding processing unit 12, a display control unit 13, and a memory control unit (hereinafter referred to as an SDRAM (Synchronous Dynamic Random Access Memory) control unit) 14 connected as shown in FIG. . The decoding processing unit 12 includes an entropy decoder (or stream analysis unit) 21, a coefficient decoder 22, an inter-frame prediction unit 23, an image generation unit 24, a read prediction image analysis unit 25, an integrated data storage unit 26, and a memory controller 27. . The decoding device 1 is, for example, a semiconductor device configured by one chip or a semiconductor device configured by a multichip module (MCM) including a plurality of chips.

  The external storage device 30 is connected to the SDRAM control unit 14 of the decoding device 1. In this embodiment, the external storage device 30 is composed of SDRAM and is separate from the decoding device 1. However, it goes without saying that the decoding device 1 and the external storage device 30 may constitute a single module.

  The decoding device 1 decompresses a video stream of a moving image compressed by a moving image compression method that performs inter-frame prediction represented by standards such as MPEG-1, MPEG-2, MPEG-4, H.264, and VC-1. (Decrypt) The decoding device 1 is mounted on a device having a video playback function such as a video player / recorder or a video camera.

  The video stream input to the stream input unit 11 is temporarily stored in the external storage device 30 via the SDRAM control unit 14. The decode processing unit 12 reads the video stream from the external storage device 30 and performs a decoding process, and stores the decoded image in the external storage device 30 again. In addition, when decoding a frame encoded using motion prediction, the decode processing unit 12 reads data of a previously decoded frame from the external storage device 30 and performs the decoding process using the frame as a reference image. Use. The display control unit 13 reads out the decoded images stored in the external storage device 30 in the display order, and outputs them as display image signals to a video interface circuit or the like (not shown). As a result, the display image signal is displayed on the screen of a display device (not shown).

  In the decoding processing unit 12, the entropy decoder 21 reads a video stream stored in the external storage device 300 via the SDRAM control unit 14 that functions as an interface to the external storage device 30, and motion vectors, DCT (Discrete Cosine Transform). Various information necessary for decoding processing such as coefficients is extracted. This embodiment is particularly characterized in that a read prediction image analysis unit 25 and an integrated data storage unit 26 are provided in the decode processing unit 12.

  The inter-frame prediction unit 23 calculates a reference image based on information such as a motion vector extracted from the video stream, reads the reference image data from the external storage device 30, and generates a predicted image based on the read data. . As will be described later, when the reference image data is read, the read predicted image analysis unit 25 determines whether it is better to read each region in the data processing unit in combination with the data of other regions. If it is determined that the occupancy time of the bus used for accessing the external storage device 30 (ie, the bus connecting the decoding device 1 and the external storage device 30) is shorter when it is determined and integrated, Data is read from the external storage device 30 simultaneously. The integrated data storage unit 26 stores integrated area data read from the external storage device 30, that is, integrated data. If the integrated data already stored in the integrated data storage unit 26 can be used as part of the reference image data, it is read from the integrated data storage unit 26 and reused. In this specification, such read control is referred to as “integrated read control”.

  The coefficient decoder 22 performs inverse quantization and inverse DCT transform using the coefficient information extracted by the entropy decoder 21 to restore a prediction error. The image generation unit 24 adds the prediction error restored by the coefficient decoder 22 to the prediction image data generated by the inter-frame prediction unit 23, generates a decoded image, and stores the decoded image in the external storage device 30.

  In general, even if the reference image is subdivided, the same rectangular area (macroblock) or an adjacent macroblock is often read as a reference image. Therefore, the read predicted image analysis unit 25 integrates a plurality of adjacent areas as one read area. The integrated area is collectively read from the external storage device 30 and stored in the integrated data storage unit 26. Later, when calculating the inter-frame prediction, only the necessary part in the inter-frame prediction unit 23 is read out from the integrated data storage unit 26, and the data is also read out from the external storage device 30 as necessary. 23, inter-frame prediction is performed.

  In the present embodiment, the inter-frame prediction unit 23 reads a reference image based on motion vector information extracted from the video stream, and reads the reference image data from the external storage device 30, and based on the read data. Means for generating a predicted image. Further, as will be described later, the read predicted image analysis unit 25 determines the read cycle of each area, the overhead for each access of the external storage device 30, or the occupied time of the bus used to access the external storage device 30. A determining unit configured to determine a shape of the reference image based on the reference image; When the reference image data is read, the determining unit determines whether or not each area in the data processing unit is to be integrated with the data of the other areas, and when it is determined to be integrated and read, Integrated determination means for simultaneously reading out data of the area to be read from the external storage device 30.

  FIG. 2 is a diagram for explaining integration of readout areas. FIG. 2 shows a case where there are seven areas (blocks) B having a read request (request) per macroblock. In this case, the read predicted image analysis unit 25 includes one read area so that the left two adjacent blocks B among the seven blocks B indicated by (a) include the areas indicated by hatching (b). It is integrated into IR, and the adjacent five blocks B on the right side are integrated into one readout area IR so as to include the area indicated by hatching in (b). Also, the two integrated read areas IR shown in (b) are stored in the integrated data storage unit 26. Therefore, when the inter-frame prediction unit 23 calculates inter-frame prediction, only a necessary part (block B) is read out of the integrated read area IR in the integrated data storage unit 26 shown in (c). It reads out and uses through the prediction image analysis part 25. FIG.

  Next, before explaining the calculation performed by the read predicted image analysis unit 25, a general DDR-SDRAM access timing that can be used as the external storage device 30 will be described with reference to FIG. FIG. 3 is a timing chart for explaining the access timing of the DDR-SDRAM, and shows the clock, address designation and data output input to the DDR-SDRAM.

  When a DDR-SDRAM is used for the external storage device 30, the row address (Adr) of the DDR-SDRAM accessed in the first cycle is specified first, and the column address (Adr) is specified in the next cycle. In these two cycles, an access area in the DDR-SDRAM to be accessed is determined. On the other hand, in the example of FIG. 3, data is output after 2.5 cycles. In the DDR-SDRAM, data is output at both the rising and falling edges of the clock, so reading of data for four words is completed in two cycles. That is, it takes 2 + 2 + 2.5 + (4/2) = 6.5 cycles to read 4 words of data from the address issuance point. Even if data reading is completed at the falling edge of the clock, the next address cannot be specified at the next rising edge, so an overhead of +5 cycles is added for reading that ends in two cycles. Will be.

  In addition to the DDR-SDRAM access preparation period, several circuit blocks are interposed before the read request from the read predicted image analysis unit 25 reaches the DDR-SDRAM, so that the overhead further increases. . Here, for convenience of explanation, it is assumed that this increased overhead is +4 cycles. Therefore, the overhead in this case is 9 cycles in total. Note that the total overhead in this case varies depending on the type of DDR-SDRAM used and the hardware configuration, so the above-mentioned nine cycles are merely examples, although there is a high possibility of being a value unique to the hardware.

The bus occupancy time when reading the two blocks Bi and Bj from the DDR-SDRAM is defined as a function such as a data read cycle as follows:
Bi data read cycle (= Bi word number / 2) = ReadCycle (Bi)
Bj data read cycle (= number of words in Bj / 2) = ReadCycle (Bj)
Overhead for each SDRAM access = Overhead
Merge Bi and Bj as one rectangular area = merge (Bi, Bj)
The bus occupancy time when Bi and Bj are read separately is as follows:
ReadCycle (Bi) + ReadCycle (Bj) + Overhead × 2
The bus occupation time when Bi and Bj are integrated and read is as follows.

ReadCycle (merge (Bi, Bj)) + Overhead
FIG. 4 is a diagram for explaining the bus occupation time when the blocks Bi and Bj are individually read and the bus occupation time when the blocks Bi and Bj are read together. In FIG. 4, it is assumed that ReadCycle (Bi) = 4, ReadCycle (Bj) = 4, Overhead = 9, and ReadCycle (merge (Bi, Bj)) = 9. In the case of the example shown in FIG. 4, the bus occupation time (access time) when the blocks Bi and Bj are read individually is 4 + 4 + 9 × 2 = 26 cycles, and the buses when the blocks Bi and Bj are read together are read. The occupation time (access time) is 10 + 9 = 19 cycles. In this case, it can be seen that the bus occupation time is shorter when the blocks Bi and Bj are integrated and read.

  Thus, by considering the overhead, it can be seen that even if the amount of data read from the DDR-SDRAM increases, it is more efficient to read adjacent blocks in an integrated manner.

  Next, a method of further reducing the access amount by reusing integrated data obtained by integrating adjacent blocks, that is, integrated read control will be described.

  First, reuse of integrated data stored in the integrated data storage unit 26 will be described with reference to FIGS. Here, for convenience of explanation, an example will be described in which an algorithm that performs integrated reading for each macroblock is employed. In the present embodiment, the processing of one macroblock is set as one unit, but it goes without saying that the same method can be used even if the processing unit is changed, for example, two macroblocks are set as one unit.

  FIG. 5 is a diagram showing a region P already stored in the integrated data storage unit 26 and a region Bi necessary for the current processing in the inter-frame prediction unit 23. In FIG. 5, the region P is indicated by a thick solid line, and the region Bi is indicated by a thick alternate long and short dash line. In this embodiment, in the case shown in FIG. 5, some data has already been read from the external storage device 30 and is reused.

  FIG. 6 is a diagram for explaining the reuse of data shown in FIG. In FIG. 6, P is an image holding area that has already been read in the past macroblock processing, Bi is a current required area that is required in the current macroblock processing as a result of the integration process, and Ci is the above two areas P and Bi. , That is, a diversion area that can be diverted, Ri indicates a new read area that needs to be newly read from the external storage device 30.

  7 to 10 are block diagrams for explaining the reuse of the integrated data. The same parts as those in FIG. 7 to 10, reference numeral 26A denotes an integrated data storage area for storing even-numbered macroblocks in the integrated data storage section 26, and 26B denotes an integrated data storage area for storing odd-numbered macroblocks in the integrated data storage section 26. Indicates.

  In order to reuse the integrated data, the integrated data storage unit 26 is divided into two integrated data storage areas 26A and 26B as shown in FIG. 7, and even-numbered macroblock data is stored in one integrated data storage area 26A. The odd-numbered macroblock data is stored in the other integrated data storage area 26B. Needless to say, the integrated data storage areas 26A and 26B may be configured by physically separate storage units.

  At the time of processing the first macroblock, since there is no data of the previous macroblock, the data cannot be reused. Therefore, as shown by the broken arrow Din in FIG. Data is read from the external storage device 30 and stored in the integrated data storage area 26A.

  At the time of processing the next macroblock, if the data of the previous macroblock can be reused, the data of the previous macroblock is transferred from the integrated data storage area 26A as indicated by the broken arrow Dcopy in FIG. Data copying to be performed on the integrated data storage area 26B is performed.

  Further, regarding the macro block data to be newly read, it is necessary to input data to be read from the external storage device 30 and stored in the integrated data storage area 26B, as indicated by the broken line arrow Din in FIG.

  By repeating the above operation for each macroblock, the data used in the previous macroblock can be reused even when the subsequent macroblock is processed.

  FIG. 11 is a flowchart for explaining the above series of processing, that is, integrated reading control in the present embodiment. In FIG. 11, in step S <b> 1, the inter-frame prediction unit 23 calculates a reference image (that is, a region before integration) for each motion vector for one macroblock. In step S <b> 2, the read prediction image analysis unit 25 performs integration determination processing for determining whether or not each block in one macroblock should be integrated with data of other blocks. In step S <b> 3, the read predicted image analysis unit 25 performs a data read range determination process for determining a data read range from the external storage device 30 and the integrated data storage unit 26 based on the result of the integration determination process. In step S4, the read predicted image analysis unit 25 performs a copy process of copying data from the integrated data storage unit 26 based on the data read range determination process. In step S5, the read predicted image analysis unit 25 performs a request process for requesting data from the integrated data storage unit 26 based on the data read range determination process, and the process ends. In this way, a reference image for each motion vector in one macroblock is obtained, and an integration determination process, a data read range determination process, a copy process, and a request process are performed.

  FIG. 12 is a flowchart illustrating the integration determination process in step S2. In FIG. 12, step S201 sets the number of target blocks for integration determination to N (N is an integer equal to or greater than 1), and step S202 sets i to i = 1. The number N of target blocks for integration determination is equal to the number of motion vectors in one macroblock. In step S203, it is determined whether or not the current necessary area (block) Bi is null. If the decision result in the step S203 is YES, a step S204 sets j to j = 1. In step S205, it is determined whether i = j, the current necessary block Bi is Null, and the reference images of the blocks Bi and Bj are the same. If the decision result in the step S205 is YES, a step S206 decides whether or not the access time at the individual reading is longer than the access time at the integrated reading. That is, step S206 determines whether {ReadCycle (Bi) + ReadCycle (Bj) + Overhead × 2}> {ReadCycle (merge (Bi, Bj)) + Overhead}.

  If the decision result in the step S206 is NO, a step S207 sets the current necessary block Bi to Bi = merge (Bi, Bj) and sets the block Bj to Bj = NULL, and the process advances to a step S208. On the other hand, if the determination result of step S205 is NO, or if the determination result of step S206 is YES, the process proceeds to step S208. In step S208, j is set to j = j + 1, and in step S209, it is determined whether j> N. If the determination result in step S209 is NO, the process returns to step S205, and if YES, the process proceeds to step S210. Step S210 sets i to i = i + 1, and step S211 determines whether i = N. If the determination result in step S211 is NO, the process returns to step S203, and if YES, the process ends.

  In this way, it is determined whether or not it is better to integrate each block in one macroblock with the data of other blocks, and the integrated time of the bus between the decoding device 1 and the external storage device 30 is determined. When it is determined that the data can be shortened, the data of the blocks to be integrated are read simultaneously. Further, by repeatedly determining whether the block data is individually read out or integratedly read out, the shape in which the reference image is finally read out, that is, the reference image acquisition range is determined.

  FIG. 13 is a flowchart for explaining the data read range determination processing in step S3. In FIG. 13, step S301 sets i to i = 1, and step S302 determines whether or not the current necessary area (block) Bi = Null. If the decision result in the step S302 is YES, a step S303 decides whether or not the reference image data of the current necessary block Bi is used in the previous macro block. If the decision result in the step S303 is YES, a step S304 sets the new reading area Ri to the new reading area Ri = the current necessary area Bi−the image holding area P, and the process advances to the step S305. On the other hand, if the decision result in the step S302 or S303 is NO, the process advances to a step S310 described later.

  Step S305 determines whether or not the new reading area Ri = currently required area Bi. If the determination result is NO, step S306 determines whether or not the new reading area Ri is a rectangle. If the decision result in the step S306 is NO, since the data is not reused, the new reading area Ri is set to Ri = Bi and the diversion area Ci is set to Ci = Null in the step S307, and the process proceeds to the step S310. . If the decision result in the step S306 is YES, since the data is reused, the diversion area Ci is set to Ci = Bi-Ri in a step S308, and the process advances to a step S310. If the decision result in the step S305 is YES, since there is no area that can be reused, the diversion area Ci is set to Ci = Null in a step S309, and the process proceeds to a step S310.

  Step S310 sets i to i = i + 1, and step S311 determines whether i = N. If the determination result in step S311 is NO, the process returns to step S302, and if YES, the process ends.

  In this way, the diversion area Ci and the new reading area Ri are determined, and the data is copied from the integrated data storage unit 26 and the data is read from the external storage device 30.

  Thus, the data read range determination process is performed in units of read blocks after the integration determination process. If the data used in the previous macroblock processing cannot be reused, the copy process is not performed and the diversion area Ci = NULL is set. Even if the data can be reused, if the new read area Ri is not rectangular, it is necessary to generate a plurality of read requests, which complicates the processing. For this reason, in this embodiment, when the new read area Ri is not rectangular, the new read area Ri = the current required area Bi, and data is not used. The new reading area Ri is not limited to a rectangle.

  FIG. 14 is a flowchart for explaining the copy processing in step S4. In FIG. 14, step S401 sets i to i = 1, and step S402 determines whether the diversion area | region Ci is Ci = Null. If the decision result in the step S402 is NO, a step S403 copies the diversion area Ci from the integrated data storage unit 26, and the process advances to a step S404. If the decision result in the step S402 is YES, the process advances to a step S404. Step S404 sets i to i = i + 1, and step S405 determines whether i = N. If the decision result in the step S405 is NO, the process returns to the step S402, and if it is YES, the process is ended.

  In this way, the diversion area Ci is copied from the integrated data storage unit 26 storing the data of the previous macroblock one by one.

  FIG. 15 is a flowchart for explaining the request processing in step S5. In FIG. 15, step S501 sets i to i = 1, and step S502 determines whether or not the new reading area Ri is Ri = Null. If the decision result in the step S502 is NO, a step S503 requests a new read area Ri to the external storage device 30, and the process advances to a step S504. If the decision result in the step S502 is YES, the process advances to a step S504. Step S504 sets i to i = i + 1, and step S505 determines whether i = N. If the decision result in the step S505 is NO, the process returns to the step S502, and if it is YES, the process is ended.

  In this way, by making a request in turn for the blocks other than the block that has been integrated and null, a request is made to the external storage device 30 that stores the data of the previous macroblock for each new read area Ri. And copy.

  FIG. 16 is a flowchart for explaining read control when data is not used. In FIG. 16, the processes of steps S11 and S12 are the same as the processes of steps S1 and S2 shown in FIG. The request process in step S13 includes steps as shown in FIG.

  FIG. 17 is a flowchart for explaining the request processing in step S13 when data is not used. In FIG. 17, step S1301 sets i to i = 1, and step S1302 determines whether or not the current necessary area Bi is Bi = Null. If the decision result in the step S1302 is NO, a step S1303 requests and copies the current necessary area Bi to the external storage device 30, and the process advances to a step S1304. If the decision result in the step S1302 is YES, the process advances to a step S1304. In step S1304, i is set to i = i + 1, and in step S1305, it is determined whether i = N. If the determination result in step S1305 is NO, the process returns to step S1302, and if YES, the process ends.

  As described above, in this embodiment, it is determined whether or not it is better to integrate each block in one macro block with the data of other blocks by the integration determination process. When it is determined to be good, the data of the blocks to be integrated are read out simultaneously. In this way, when data is read out simultaneously, if the data already stored in the integrated data storage unit 26 can be used, the data in the integrated data storage unit 26 is reused without being read out from the external storage device 30 again. To do. This determination is not limited to one macroblock, but may be adaptively set for several macroblocks or a vector of one image. However, if the target macroblocks are separated from each other in time, there is a low possibility that the bus occupation time will be shortened, and the calculation itself becomes complicated, so one macroblock or two to three macroblocks, It is desirable to perform the above operation on vectors of at most 10 macroblocks.

  Further, the integration determination process is limited to the process shown in FIG. 12 as long as the determination is performed in consideration of the block read cycle, the overhead for each access of the external storage device 30, or the bus occupation time. is not. Similarly, the data read range determination process, the copy process, and the request process are not limited to the processes shown in FIGS.

  The present invention is represented by standards such as MPEG-1, MPEG-2, MPEG-4, H.264, and VC-1, which are mounted on devices equipped with video playback functions such as video players / recorders and video cameras. The present invention is applicable to a decoding apparatus that decompresses (decodes) a video stream of a moving image compressed by a moving image compression method that performs inter-frame prediction.

  While the present invention has been described with reference to the embodiments, it is needless to say that the present invention is not limited to the above-described embodiments, and various modifications and improvements can be made within the scope of the present invention.

Claims (10)

A decoding method for reading a video stream compressed by a moving image compression method for performing inter-frame prediction from a storage device, and using the decoded frame as a reference image for decoding processing,
Calculating the reference image based on motion vector information extracted from the video stream;
Generating a predicted image based on the reference image;
Whether to read out each area within the reference image processing unit from the storage device by integrating a plurality of areas, or to read the bus occupation time and each area separately when reading the plurality of areas together Judging by comparing the bus occupancy time of
The bus occupation time when reading each area individually is calculated by adding the read cycle of each area and the overhead for each access of the storage device,
The bus occupation time when the plurality of areas are read together is calculated by adding the read cycle of the integrated area and the access overhead of the storage device,
If it is determined that the bus occupancy time in the case of reading out the plurality of areas in an integrated manner is shorter and the plurality of areas are read out in a unified manner, a plurality of data to be integrated may be integrated even when the amount of data read from the storage device increases . decoding method characterized in that to read out simultaneously an area from the storage device. An integrated data storage unit stores the area read from the storage device,
When calculating the reference image based on the motion vector information, if the integrated area stored in the integrated data storage unit can be used as part of the data of the reference image, the stored image is stored. 2. The decoding method according to claim 1, wherein the integrated area is read from the integrated data storage unit. The integrated data storage unit has a first integrated data storage area for storing even-numbered macroblocks, and a second integrated data storage area for storing odd-numbered macroblocks,
When all data of the first macroblock is read from the storage device and stored in the first integrated data storage area, and the data of the previous macroblock can be reused when the next macroblock is processed Performs a data copy of copying the data of the previous macroblock from the first integrated data storage area to the second integrated data storage area, and the newly read macroblock data is read from the storage device and the second integrated data Store it in the storage area,
3. The data used in the previous macroblock is reused when processing the subsequent macroblock by repeating the storage or data copy of the data of the macroblock for each macroblock. Decoding method.   The decoding method according to any one of claims 1 to 3, wherein the moving image compression method is one of MPEG-1, MPEG-2, MPEG-4, H.264, and VC-1.   5. The decoding method according to claim 1, wherein the processing unit of the reference image is 1 to 3 rectangular macroblocks. A decoding device that reads a video stream compressed by a moving image compression method that performs inter-frame prediction from a storage device, and uses the decoded frame as a reference image for decoding processing,
A computing unit that computes the reference image based on motion vector information extracted from the video stream;
A generating unit that generates a predicted image based on the reference image;
Whether to read out each area within the reference image processing unit from the storage device by integrating a plurality of areas, or to read the bus occupation time and each area separately when reading the plurality of areas together An integrated determination unit that determines by comparing the bus occupation time of
The integrated determination unit
The bus occupation time when reading each area individually is calculated by adding the read cycle of each area and the overhead for each access of the storage device,
The bus occupancy time when the plurality of areas are read together is calculated by adding the read area read cycle and the access overhead of the storage device,
If it is determined that the bus occupancy time in the case of reading out the plurality of areas in an integrated manner is shorter and the plurality of areas are read out in a unified manner, a plurality of data to be integrated may be integrated even when the amount of data read from the storage device increases . decoding apparatus characterized by be read out simultaneously an area from the storage device. A storage unit that stores the integrated data read from the storage device in the integrated data storage unit;
When the integrated area stored in the integrated data storage section can be used as a part of the reference image data, the arithmetic section reads the integrated area from the integrated data storage section. 7. The decoding device according to claim 6, wherein The integrated data storage unit has a first integrated data storage area for storing even-numbered macroblocks, and a second integrated data storage area for storing odd-numbered macroblocks,
The storage unit reads all data of the first macroblock from the storage device, stores the data in the first integrated data storage area, and reuses the data of the previous macroblock at the time of processing the next macroblock If possible, the data of the previous macroblock is copied from the first integrated data storage area to the second integrated data storage area, and the newly read macroblock data is read from the storage device. Stored in a second integrated data storage area;
The data used in the previous macroblock is reused when processing the subsequent macroblock by repeating the storage or data copy of the data of the macroblock every macroblock. Decoding device.   9. The decoding apparatus according to claim 6, wherein the moving image compression method is any one of MPEG-1, MPEG-2, MPEG-4, H.264, and VC-1.   The decoding apparatus according to claim 6, wherein the processing unit of the reference image is 1 to 3 rectangular macroblocks.

Images