JP2009212900A - Image decoding device and image decoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce a processing time required for decoding image data compression-coded using DC/AC predictive coding. <P>SOLUTION: An image decoding device 1 is equipped with an inverse quantization section 30 and an inverse DC/AC predictive processing section 40. The inverse quantization section 30 executes inverse scan processing, addition processing and inverse quantization processing to a data sequence obtained by performing variable-length decoding on a coded image signal compression-coded using DC/AC predictive coding. Meanwhile, the inverse DC/AC predictive processing section 40 uses quantized DC/AC coefficients restored by the inverse quantization section 30 for a block prior to a current block to determine a prediction direction of the next block to be processed in succession to the current block concurrently with inverse scan processing, addition processing and inverse quantization processing on the current block. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for decoding image data that has been compression-encoded using DCT transform, quantization, DC / AC predictive encoding, and variable-length encoding, and more particularly to speeding up the decoding process.

  In the next generation DVD (Digital Video Disc) and digital television broadcasting, New codecs such as H.264 / AVC (Advanced Video Coding), VC-1 and MPEG-4 (Moving Picture Experts Group-4) are employed. These new codecs use a new technique that is not used in the conventional MPEG-2 in order to enable high-quality and high-compression coding of moving images. One such technique is DC / AC predictive coding. DC / AC predictive coding is a technique for predictively coding DC coefficients and AC coefficients obtained by orthogonal transform such as DCT (Discrete Cosine Transform) in a frame, paying attention to high pixel correlation between adjacent blocks. It is. More specifically, in the DC / AC predictive coding, after determining an appropriate prediction direction, the DC coefficient and AC coefficient of the current block obtained by orthogonal transform and the reference included in the peripheral blocks of the current block This is a technique for reducing the amount of information of compressed image data by calculating a difference between a DC coefficient and an AC coefficient for use and making the obtained difference a variable length coding target.

  Here, terms used in this specification are defined. The DC / AC coefficient is a general term for a DC coefficient and an AC coefficient obtained by orthogonal transformation such as DCT.

  The quantized DC coefficient means a DC coefficient quantized by any representative value of the quantization step interval obtained by quantizing the DC coefficient. The quantized AC coefficient means an AC coefficient quantized by any representative value of the quantization step interval obtained by quantizing the AC coefficient. The quantized DC / AC coefficient is a general term for the quantized DC coefficient and the quantized AC coefficient.

  The restored DC coefficient means a DC coefficient restored by inverse quantization in which the quantized DC coefficient is multiplied by a quantization step. The restored DC coefficient means an AC coefficient restored by inverse quantization in which a quantized AC coefficient is multiplied by a quantization step. The restored DC / AC coefficient is a general term for the restored DC coefficient and the restored AC coefficient.

  The DC prediction coefficient means a DC coefficient of a peripheral block referred to in DC / AC prediction encoding. The AC prediction coefficient means an AC coefficient of a peripheral block referred to in DC / AC prediction encoding. The DC / AC prediction coefficient is a general term for a DC prediction coefficient and an AC prediction coefficient.

  The DC difference coefficient is a difference value between the quantized DC coefficient calculated in the DC / AC predictive encoding process and the DC prediction coefficient. The AC difference coefficient is a difference value between the quantized AC coefficient calculated in the DC / AC predictive encoding process and the AC prediction coefficient. The DC / AC difference coefficient is a general term for a DC difference coefficient and an AC difference coefficient.

  For example, Patent Literatures 1 to 3 disclose image decoding apparatuses that execute a decoding process on image data compression-coded using DC / AC predictive coding. FIG. 5 is a block diagram illustrating a procedure from variable length decoding to inverse DCT among the image decoding procedures performed by the image decoding device disclosed in Patent Document 1. The image decoding procedure according to the block diagram of FIG. 5 is as follows.

  First, the variable length decoding unit 90 performs variable length decoding of encoded data. Next, the reverse scanning unit 91 extracts a DC / AC difference coefficient related to the current block to be processed from the data string after variable length decoding.

Next, the inverse DC / AC prediction unit 92 executes an inverse DC / AC prediction process. The inverse DC / AC prediction process is roughly divided into a DC / AC prediction coefficient scaling process included in a peripheral block, a prediction direction determination process, and a DC / AC prediction coefficient and a DC / AC difference coefficient according to the prediction direction. Are included. Hereinafter, the inverse DC / AC prediction processing procedure will be described in order. First, the inverse DC / AC prediction unit 92 is included in the three peripheral blocks of the current block, specifically, the upper block A, the upper left block B, and the left block C as shown in FIG. Scaling is performed on the coefficient groups QF _A [0] [i], QF _B [0] [0], and QC _C [j] [0] that are used as the DC / AC prediction coefficients. Then, the inverse DC / AC prediction unit 92 determines the prediction direction using the DC coefficient after scaling. Specifically, the DC coefficient _QF A [0] of the block A [0] and DC coefficient of the block _{B QF B [0] [0} ] and the difference between, the DC coefficient _QF C [0] of the block C [0] And the difference between the block B and the DC coefficient QF _B [0] [0] of the block B is compared, and the direction in which the difference is small is determined as the prediction direction. Further, the inverse DC / AC prediction unit 92 adds the DC / AC prediction coefficient of the upper block A or the left block C corresponding to the determined prediction direction and the DC / AC difference coefficient of the current block to quantize the current block. Restore the DC / AC coefficients.

  Next, the inverse quantization unit 93 restores the DC / AC coefficients of the current block by multiplying each of the quantized DC / AC coefficients of the current block by a corresponding quantization step.

Finally, the inverse DCT unit 94 restores the pixel value of the current block by executing an inverse DCT operation using the restored DC / AC coefficient of the current block.
JP 2001-103472 A (FIGS. 1 and 11) Japanese Patent Laid-Open No. 2002-118853 (FIGS. 1, 5, and 6) JP 2006-262188 A (FIG. 17)

  When decoding image data compression-coded using DC / AC predictive coding, there is a problem that the processing time required for decoding increases. In order to show a specific example, a case where reverse scan processing, reverse DC / AC prediction processing, and reverse quantization processing are executed on the VC-1 macroblock shown in FIG. 7 will be described.

  As shown in FIG. 7, the VC-1 macroblock includes a luminance signal block of 16 × 16 pixels and a color difference signal block of 8 × 8 pixels (hereinafter referred to as Cb block and Cr block). When inverse DC / AC prediction is performed, a 16 × 16 pixel luminance signal block is divided into four 8 × 8 pixel blocks (hereinafter referred to as Y0 to Y3 blocks). The processing order of inverse DC / AC prediction in the entire macroblock is the order of Y0, Y1, Y2, Y3, Cb, and Cr.

  FIG. 8 is a diagram illustrating a process in the case of executing reverse scan processing, reverse DC / AC prediction processing, and reverse quantization processing on a macroblock of VC-1. As can be seen from FIG. 8, not only when the first Y0 block is processed, but also in the subsequent five blocks Y1 to Cr, the DC prediction coefficient scaling process and the prediction direction determination process (FIG. 8). In FIG. 8, these are collectively expressed as “DC”), the AC prediction coefficient scaling process (indicated as “AC” in FIG. 8), the subsequent reverse scan process (indicated as “IScan” in FIG. 8), and the addition process (in FIG. 8). In FIG. 8, it is necessary to carry out prior to “Add”) and inverse quantization processing (indicated as IQ in FIG. 8).

  The addition process here is a part of the inverse DC / AC prediction process executed by the inverse DC / AC prediction unit 92 described above, and includes the DC / AC prediction coefficient corresponding to the prediction direction and the current block. It means a process of adding a DC / AC difference coefficient. That is, the addition process corresponds to the latter half of the entire inverse DC / AC prediction process. On the other hand, the DC prediction coefficient scaling process, the prediction direction determination process, and the AC prediction coefficient scaling process correspond to the first half of the entire inverse DC / AC prediction process.

  As described above, in decoding of image data compression-coded using DC / AC predictive coding, it is necessary to additionally perform reverse DC / AC prediction processing in addition to reverse scan processing and inverse quantization processing. Therefore, the processing time required for decoding increases.

  A first aspect of the present invention includes orthogonal transform, quantization of DC / AC coefficients obtained by the orthogonal transform, DC / AC predictive coding of quantized DC / AC coefficients obtained by the quantization, and the DC / It is an image decoding apparatus that decodes an encoded image signal that has been compression-encoded by variable-length encoding of a DC / AC difference coefficient obtained by AC predictive encoding. The image decoding apparatus includes an inverse quantization unit and an inverse DC / AC prediction processing unit.

  Here, the inverse quantization unit performs inverse scan processing for extracting the DC / AC difference coefficient relating to a current block to be processed from a data sequence after performing variable length decoding on the encoded image signal, the current Addition processing for restoring the quantized DC / AC coefficient by adding a DC / AC prediction coefficient corresponding to the prediction direction of the DC / AC prediction coding to the DC / AC difference coefficient of the block, and the restored current An inverse quantization process of the quantized DC / AC coefficient of the block is executed.

  Meanwhile, the inverse DC / AC prediction processing unit determines the prediction direction necessary for the addition process. Further, the inverse DC / AC prediction processing unit uses the quantized DC / AC coefficient restored by the inverse quantization unit related to the block before the current block, thereby performing the inverse scan processing related to the current block, In parallel with the execution of the addition process and the inverse quantization process, the prediction direction determination process for the next block to be processed subsequent to the current block is executed.

  The first aspect of the present invention described above is the inverse DC / AC prediction of the next block as the DC / AC prediction coefficient among the quantized DC / AC coefficients sequentially calculated in the process of the inverse quantization unit for the current block. It is noted that when the inverse quantization of the coefficient group referred to in the process is completed, the prediction direction determination process for the next block can be started. That is, in the first aspect of the present invention described above, the prediction direction determination process for the next block is executed in parallel with the inverse quantization of the current block. As a result, it is not necessary to wait for the time required to determine the prediction direction at the start of processing for the next block, and the inverse scan process to the inverse quantization process can be started immediately. Therefore, according to the first aspect of the present invention, the processing time required for inverse quantization of a macroblock can be shortened.

  According to the present invention, it is possible to provide an image decoding device capable of reducing the processing time required for decoding image data compressed and encoded using DC / AC predictive encoding.

  Hereinafter, specific embodiments to which the present invention is applied will be described in detail with reference to the drawings. In the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted as necessary for the sake of clarity.

<Embodiment 1 of the Invention>
FIG. 1 is a block diagram showing an image decoding apparatus 1 which is one embodiment of the present invention. The image decoding device 1 is a device that decodes encoded image data encoded by VC-1. In FIG. 1, a variable length decoding unit 10 performs variable length decoding of encoded image data. The variable length decoding unit 10 corresponds to the variable length decoding unit 90 described above.

  The input buffer 11 is a buffer memory that holds a DC / AC coefficient group (of course, subjected to quantization and DC / AC predictive coding) subjected to variable length decoding by the variable length decoding unit 10. The input buffer 11 is a supply source of processing target data to a pipeline processing unit 12 described later.

  The pipeline processing unit 12 performs reverse scan processing, reverse DC / AC prediction processing, and reverse quantization processing under the order control by the sequence processing unit 13. The sequence processing unit 13 is a block for mainly controlling execution of pipeline processing for each block by the pipeline processing unit 12.

  The sequence processing unit 13 supplies the pipeline processing unit 12 with a start signal for starting the pipeline for each block and a parameter signal including parameters necessary for the inverse quantization processing for each block. In addition, the sequence processing unit 13 receives a finish signal indicating completion of inverse quantization processing for one block from the pipeline processing unit 12.

The parameter signal includes a parameter indicating whether or not AC prediction is performed for each macroblock, and a parameter related to the quantization step size of the current block and peripheral blocks. Here, the parameter relating to the quantization step size may be the quantization step size itself, or may be dobule_quant calculated using the quantization parameter MQUANT or MQUANT used in VC-1. MQUANT is a parameter associated with a quantization step size having a value between 1 and 31, and is encoded for each macroblock. The quantization step size (DCStepSize) and double_quant of the DC coefficient can be calculated by the following equations using the quantization parameter MQUANT.
(1) DCStepSize
For MQUANT equal to 1 or 2
DCStepSize = 2 * MQUANT
For MQUANT equal to 3 or 4
DCStepSize = 8
For MQUANT greater than or equal to 5
DCStepSize = MQUANT / 2 + 6
(2) double_quant
double_quant = 2 * MQUANT + HALFQP
Here, HALFQP is a 1-bit value of 0 or 1.

  The parameter signal supplied from the sequence processing unit 13 to the pipeline processing unit 12 includes parameters specified for each current block in addition to the parameters for each macroblock. The parameters specified for each current block are an index value for specifying the current block, a parameter indicating whether the peripheral block can be used for AC / DC prediction, and the like.

  The pipeline processing unit 12 includes a pipeline control unit 20, an inverse quantization unit 30, an inverse DC / AC prediction processing unit 40, in order to execute an inverse scan process, an inverse DC / AC prediction process, and an inverse quantization process. And a peripheral DC / AC coefficient buffer 50. Among these, the pipeline control unit 20 functions as an interface with the sequence processing unit 13 for transmission / reception of the start signal, the parameter signal, and the finish signal, and performs overall control of the pipeline processing unit 12.

  The inverse quantization unit 30 includes an inverse scan processing unit 301, an addition processing unit 302, and an inverse quantization processing unit 303. The inverse scan processing unit 301 extracts the DC / AC difference coefficient related to the current block related to the current block from the input buffer 11.

  The addition processing unit 302 executes an addition process corresponding to the latter half of the entire inverse DC / AC prediction process.

  The inverse quantization processing unit 303 receives the output data of the addition processing unit 302, that is, the quantized DC / AC coefficient of the current block for which the inverse DC / AC prediction processing has been completed, and performs inverse quantization to thereby perform a process related to the current block. A restored DC / AC coefficient is generated. The restored DC / AC coefficient generated by the inverse quantization processing unit 303 is stored in the output buffer 15.

  The inverse DC / AC prediction processing unit 40 executes parameter calculation, scaling of DC / AC coefficients, and determination of a prediction direction, which are processes in the first half of the entire inverse DC / AC prediction process. For this purpose, the inverse DC / AC prediction processing unit 40 includes a control unit 401, a parameter calculation unit 402, a scaling unit 403, and a prediction direction determination unit 404. The inverse DC / AC prediction processing unit 40 has an interface with the pipeline control unit 20, the inverse quantization unit 30, and the peripheral DC / AC coefficient buffer 50.

  The control unit 401 receives the dcstart signal and the acstart signal from the inverse quantization unit 30 and transmits the preddir signal and the dcacfinish signal to the inverse quantization unit 30. Here, the dcstart signal is a signal for instructing the start of inverse prediction of DC coefficients, and the acstart signal is a signal for instructing the start of inverse prediction of AC coefficients. The preddir signal is a signal for notifying the inverse quantization unit 30 of the prediction direction determined by the prediction direction determination unit 404. The dcacfinish signal is a signal for notifying the inverse quantization unit 30 that the DC / AC coefficient scaling for one block and the prediction direction determination processing have been completed.

  The parameter calculation unit 402 calculates parameters for scaling the DC coefficient and the AC coefficient. For example, when the double_quant value is supplied to the pipeline processing unit 12 by the parameter signal described above, the parameter calculation unit 402 uses the double_quant value to determine the value related to the MQUANT, DC coefficient step size (DCStepSize), and AC coefficient step size. STEP (STEP = double_quant-1) and the quantization scale are calculated. Here, as described later, the quantization scale may be acquired with reference to a lookup table (not shown) in which the quantization scale value is stored.

  The scaling unit 403 has a case where the quantization step size or the quantization scale is different between the current block that is the processing target block of the inverse quantization unit 30 and the peripheral blocks (that is, the blocks A to C illustrated in FIG. 7) ( In other words, when MQUANT is different), scaling is performed on DC / AC prediction coefficients with different quantization step sizes or quantization scales. Note that scaling with respect to the AC prediction coefficient may be performed only for the neighboring blocks in the prediction direction determined by the prediction direction determination unit 404.

ここで、SQF_C[0][0]は、スケーリング後のＤＣ予測係数である。DCSTEP_P は、ＤＣ予測係数F_C[0][0]の量子化ステップサイズである。DCSTEP_C は、カレントブロックのＤＣ係数の量子化ステップサイズである。DQSclale[m]は、mで示されるステップサイズに対応した量子化スケールである。なお、DQSclale[m]は、量子化スケール値を格納したルックアップテーブル（いわゆる量子化スケールテーブル）である。ＶＣ−１の場合、DQSclale[m]には６３通りの量子化ステップサイズに対応した量子化スケール値が格納されている。なお、量子化スケールの具体的な値は、ＶＣ−１（ＳＭＰＴＥ ４２１Ｍ）に定義されているものを使用すればよい。量子化スケールテーブルから対応する１つの値を取得する処理は、上述したパラメータ計算部４０２が行えばよい。演算子"＞＞k"は、右方向へのkビットシフトを表す。 For example, the scaling for the left DC prediction coefficients _QF C [0] of the block C [0] may be performed by the following equation (1).

Here, SQF _C [0] [0] is a DC prediction coefficient after scaling. DCSTEP _P is the quantization step size of the DC prediction coefficient F _C [0] [0]. DCSTEP _C is the quantization step size of the DC coefficient of the current block. DQSclale [m] is a quantization scale corresponding to the step size indicated by m. DQSclale [m] is a look-up table (so-called quantization scale table) that stores quantization scale values. In the case of VC-1, quantization scale values corresponding to 63 quantization step sizes are stored in DQSclale [m]. In addition, what is necessary is just to use what is defined in VC-1 (SMPTE 421M) as a specific value of a quantization scale. The process for obtaining one corresponding value from the quantization scale table may be performed by the parameter calculation unit 402 described above. The operator “>> k” represents a k-bit shift to the right.

ここで、SQF_C[ｊ][0]は、スケーリング後のＡＣ予測係数である。ACSTEP_P は、ＡＣ予測係数F_C[j][0]の量子化ステップサイズである。ACSTEP_C は、カレントブロックのＡＣ係数の量子化ステップサイズである。 Further, the scaling for the AC prediction coefficients in the left block _{C QF C [j] [0} ] may be performed by the following equation (2).

Here, SQF _C [j] [0] is an AC prediction coefficient after scaling. ACSTEP _P is the quantization step size of the AC prediction coefficient F _C [j] [0]. ACSTEP _C is the quantization step size of the AC coefficient of the current block.

  The prediction direction determination unit 404 calculates the difference value of the DC coefficient after scaling of the three peripheral blocks A to C, and determines the prediction direction of the current block.

  The inverse DC / AC prediction processing unit 40 stores the DC / AC prediction coefficient scaled by the scaling unit 403 in the peripheral DC / AC coefficient buffer 50, and receives the preddir signal including the prediction direction determined by the prediction direction determination unit 404. It transmits to the inverse quantization unit 30.

  The peripheral DC / AC coefficient buffer 50 is a DC / AC of peripheral blocks (specifically, upper, upper left, and left three blocks) supplied from the external memory 14 for use in inverse DC / AC prediction of the current block. A prediction coefficient is stored. Also, the data stored in the peripheral DC / AC coefficient buffer 50 is updated with the DC / AC prediction coefficient after scaling. The DC / AC prediction coefficient held in the peripheral DC / AC coefficient buffer 50 is used for addition processing with the DC / AC difference coefficient in the addition processing unit 302 described above.

  A specific example of the peripheral DC / AC coefficient buffer 50 is shown in FIG. The DC / AC coefficient buffer 50 shown in FIG. 2 includes four DC coefficient buffers 500 to 503 and six DCAC coefficient buffers 510 to 515. The DC coefficient buffer 500 stores the DC prediction coefficient of the upper left block of the current block of the luminance signal. The DC coefficient buffer 501 stores the DC prediction coefficient of the left block of the current block of the luminance signal. The DC coefficient buffer 502 stores the DC prediction coefficient of the upper left block of the current block of the color difference signal Cb. The DC coefficient buffer 503 stores the DC prediction coefficient of the upper left block of the current block of the color difference signal Cr.

  The ACDC coefficient buffer 510 stores the DC / AC prediction coefficient of the upper left or upper right block in addition to the upper block of the current block of the luminance signal. The ACDC coefficient buffer 511 stores the DC / AC prediction coefficient at the upper left or lower left in addition to the left block of the current block of the luminance signal. The ACDC coefficient buffer 512 stores the DC / AC prediction coefficient of the upper block of the current block of the color difference signal Cb. The ACDC coefficient buffer 513 stores the DC / AC prediction coefficient of the left block of the current block of the color difference signal Cb. The ACDC coefficient buffer 514 stores the DC / AC prediction coefficient of the upper block of the current block of the color difference signal Cr. Finally, the ACDC coefficient buffer 515 stores the DC / AC prediction coefficient of the left block of the current block of the color difference signal Cr.

  Returning to FIG. The external memory 14 is a memory for storing the decoded image data, and stores the pixel value of the upper macroblock of the current macroblock necessary for the inverse DC / AC prediction process.

  The output buffer 15 stores the output data of the pipeline processing unit 12, that is, the DC / AC coefficient of the current block after inverse quantization, and supplies this to the inverse DCT unit 16.

  The inverse DCT unit 16 restores the pixel value of the current block by inputting the restored DC / AC coefficient of the current block from the output buffer 15 and executing an inverse DCT operation.

  Subsequently, the operation procedure when the pipeline processing unit 12 performs from the inverse scan process to the inverse quantization process will be specifically described below.

(STEP0)
The pipeline processing unit 12 performs DC / AC coefficient processing on the plurality of blocks in the macroblock in the order shown in FIG. 7, that is, the order of Y0, Y1, Y2, Y3, Cb, and Cr. The sequence processing unit 13 manages the current position of the current block.

(STEP1)
In response to an instruction from the sequence processing unit 13, the values of the peripheral blocks of the current block are stored from the external memory 14 into the peripheral DC / AC coefficient buffer 50. Of the peripheral block values stored in the peripheral DC / AC coefficient buffer 50, the value related to the left macroblock is the result of the processing of the previous macroblock.

(STEP2)
When the data for the current block is accumulated in the input buffer 11, the sequence processing unit 13 gives the start signal and the parameter signal to the pipeline control unit 20, and the inverse quantization unit 30 and the inverse DC / AC prediction processing unit 40 Start processing.

(STEP3)
The pipeline control unit 20 that has received the start signal supplies the parameters given from the sequence processing unit 13 to the inverse quantization unit 30 and the inverse DC / AC prediction processing unit 40, and also indicates the start of processing for the current block. The signal is supplied to the inverse quantization unit 30.

(STEP4)
The inverse quantization unit 30 that has received the blkstart signal causes the inverse DC / AC prediction processing unit 40 to perform dcstart in order to perform DC / AC prediction processing of the Y0 block, specifically, coefficient scaling and prediction direction determination processing. Signal and acstart signal. This is because the pipeline processing unit 12 of the present embodiment performs inverse quantization in units of macroblocks, and coefficient scaling and prediction direction determination are not performed for the Y0 block that is the first block in one current block. This is because. The inverse DC / AC prediction processing unit 40 that has received the dcstart signal and the acstart signal performs parameter calculation, executes a DC coefficient scaling process when the quantization scales of the current block and the peripheral blocks are different, and sets the prediction direction. Make a decision. Then, the inverse DC / AC prediction processing unit 40 includes the calculated prediction direction in the preddir signal and notifies the inverse quantization unit 30. Subsequently, the inverse DC / AC prediction processing unit 40 performs scaling of the AC coefficient in the calculated prediction direction, and stores the processing result in the peripheral DC / AC coefficient buffer 50. The inverse DC / AC prediction processing unit 40 notifies the inverse quantization unit 30 of a dcacfinish signal when all the DC / AC prediction processing (however, coefficient scaling and prediction direction determination processing) regarding the Y0 block is completed.

(STEP5)
When the inverse quantization unit 30 receives the dcacfinish signal indicating the end of the processing related to the Y0 block in the inverse DC / AC prediction processing unit 40, the inverse quantization, the DC / AC prediction coefficient according to the prediction direction, and the DC / AC difference coefficient Addition processing and inverse quantization are performed on the 64 coefficients in the Y0 block. The inverse quantization unit 30 stores the obtained restored DC / AC coefficient of the Y0 block in the output buffer 15. Note that the upper side coefficient and the left side coefficient of the Y0 block after the addition processing are used as the peripheral prediction pixels of the subsequent Y1 block and Y2 block. For this reason, the inverse quantization unit 30 stores these coefficients in the ACDC coefficient buffers 510 and 511 included in the peripheral DC / AC coefficient buffer 50.

(STEP6)
The inverse quantization unit 30 causes the inverse DC / AC prediction processing unit 40 to start coefficient scaling and prediction direction determination processing for the Y1 block in parallel with the inverse quantization processing of the Y0 block in STEP 5 as follows. Operate. That is, the inverse quantization unit 30 provides the dcstart signal of the Y1 block to the inverse DC / AC prediction processing unit 40 as soon as the addition process to the DC coefficient of the Y0 block is completed in the execution process of STEP5 described above. The inverse DC / AC prediction processing unit 40 that has received the dcstart signal of the Y1 block performs parameter calculation, performs DC coefficient scaling processing as necessary, and determines the prediction direction of the Y1 block. Then, the inverse DC / AC prediction processing unit 40 includes the calculated prediction direction of the Y1 block in the preddir signal and notifies the inverse quantization unit 30.

  Further, in order to cause the inverse DC / AC prediction processing unit 40 to perform the scaling processing of the AC coefficient necessary for the execution of the addition processing of the Y1 block in parallel with the quantum processing of the Y0 block, the inverse quantization unit 30 is as follows. Operate. In other words, the inverse quantization unit 30 promptly outputs the acstart signal of the Y1 block immediately after the addition processing relating to the left side coefficient of the Y0 block (that is, the indexes 8, 16, 24, 32, 40, 48, and 56 in FIG. 3) is completed. Transmit to the inverse DC / AC prediction processing unit 40. The end of the addition processing related to the left-side coefficient of the Y0 block may be determined by, for example, acquiring the coefficient at index 49 subsequent to the coefficient at index 56 by the reverse scan processing unit 301. The inverse DC / AC prediction processing unit 40 that has received the acstart signal of the Y1 block scales the AC coefficient (that is, the AC prediction coefficient) of the peripheral block corresponding to the prediction direction of the Y1 block, and the result is the peripheral DC / AC coefficient. The data is stored in the buffer 50 and the dcacfinish signal is transmitted to the inverse quantization unit 30.

(STEP7)
The inverse quantization unit 30 receives the dcacfinish signal related to the Y1 block from the inverse DC / AC prediction processing unit 40, and when the inverse quantization processing of the Y0 block is completed for all 64 coefficients, the processing of the Y0 block is completed. In order to notify this, a pipedone signal is transmitted to the pipeline control unit 20.

(STEP8)
The pipeline control unit 20 recognizes that the inverse quantization processing of the Y0 block has been completed by receiving the pipedone signal, and transmits a finish signal to the sequence processing unit 13.

(STEP9)
When the sequence processing unit 13 receives the finish signal, the sequence processing unit 13 obtains parameters necessary for processing of the Y1 block that is the next processing target block, and sends a parameter signal and a start signal indicating the start of processing of the Y1 block to the pipeline control unit 20. Send.

(STEP10)
The pipeline control unit 20 that has received the start signal indicating the start of processing of the Y1 block supplies the parameters given from the sequence processing unit 13 to the inverse quantization unit 30 and the inverse DC / AC prediction processing unit 40, and at the same time, the Y1 block A blkstart signal indicating the start of the process is transmitted to the inverse quantization unit 30.

(STEP11)
The inverse quantization unit 30 that has received the blkstart signal promptly starts the DC coefficient addition process. This is because the scaling of the DC / AC coefficient and the determination of the prediction direction regarding the peripheral blocks of the Y1 block have already been completed in STEP 6 described above. The inverse quantization unit 30 performs inverse scanning on the Y1 block, addition processing of DC / AC prediction coefficients and DC / AC difference coefficients according to the prediction direction, and inverse quantization on 64 coefficients in the Y1 block. The inverse quantization unit 30 stores the obtained restored DC / AC coefficient of the Y1 block in the output buffer 15. Note that the upper side coefficient of the Y1 block after the addition processing is used as a peripheral prediction pixel of the subsequent Y3 block. For this reason, the inverse quantization unit 30 stores these coefficients in the ACDC coefficient buffer 510 included in the peripheral DC / AC coefficient buffer 50.

(STEP12)
The inverse quantization unit 30 causes the inverse DC / AC prediction processing unit 40 to start coefficient scaling and prediction direction determination processing related to the Y2 block in parallel with the above-described inverse quantization processing of the Y1 block in STEP 11 as described below. To work. That is, the inverse quantization unit 30 provides the dcstart signal of the Y2 block to the inverse DC / AC prediction processing unit 40 as soon as the addition process for the DC coefficient of the Y1 block is completed. Further, the inverse quantization unit 30 may transmit the acstart signal in accordance with the dcstart signal. As can be seen from FIG. 2, the AC coefficient of the Y1 block is not the AC prediction coefficient of the Y2 block, so there is no need to wait for the AC coefficient addition process of the Y1 block.

  The inverse DC / AC prediction processing unit 40 that has received the dcstart signal and the acstart signal of the Y2 block uses the DC prediction coefficients included in the peripheral blocks of the Y2 block to perform parameter calculation, DC coefficient scaling, and prediction direction determination of the Y2 block. Execute. Then, the inverse DC / AC prediction processing unit 40 includes the calculated prediction direction of the Y2 block in the preddir signal and notifies the inverse quantization unit 30. Subsequently, the inverse DC / AC prediction processing unit 40 scales the AC coefficient (that is, AC prediction coefficient) of the peripheral block corresponding to the calculated prediction direction of the Y2 block, and stores the result in the peripheral DC / AC coefficient buffer 50. Then, the dcacfinish signal is transmitted to the inverse quantization unit 30.

(STEP 13)
The inverse quantization unit 30 receives the dcacfinish signal for the Y2 block from the inverse DC / AC prediction processing unit 40, and the Y1 block processing is completed when the inverse quantization processing for the Y1 block is completed for all 64 coefficients. In order to notify this, a pipedone signal is transmitted to the pipeline control unit 20.

(STEP14)
The pipeline control unit 20 recognizes that the inverse quantization process of the Y1 block has been completed by receiving the pipedone signal, and transmits a finish signal to the sequence processing unit 13.

(STEP 15)
When the sequence processing unit 13 receives the finish signal, the sequence processing unit 13 obtains parameters necessary for processing the Y2 block that is the next processing target block, and sends a parameter signal and a start signal indicating the start of processing of the Y1 block to the pipeline control unit 20. Send.

(STEP 16)
The pipeline processing unit 12 and the sequence processing unit 13 perform the inverse quantization process on the Y2, Y3, Cb, and Cr blocks by repeating the same processes as in STEPs 10 to 14 regarding the Y1 block described above. Thereby, the inverse quantization of one macroblock is completed.

  As described above, the pipeline processing unit 12 included in the image decoding apparatus 1 according to the present embodiment relates to the next block with respect to the inverse AC / DC prediction processing from the remaining Y1 block to the Cr block excluding the head block Y0. The DC / AC coefficient scaling and the prediction direction discrimination processing are executed in parallel with the current block inverse quantization. This eliminates the need to wait for the time required for scaling the DC / AC coefficient and determining the prediction direction at the start of processing for the next block (for example, Y1 block) after dequantization of the current block (for example, Y0 block). For this reason, the pipeline processing unit 12 can quickly start the inverse scan process, the addition process, and the inverse quantization process for the next block. Therefore, the image decoding apparatus 1 according to the present embodiment can reduce the processing time required for inverse quantization of one macroblock.

  Here, the effect of shortening the processing time according to the present embodiment will be described by comparing FIG. 4 and FIG. FIG. 4 is a diagram for explaining processing time when the VC-1 macroblock inverse scan processing, inverse DC / AC prediction processing, and inverse quantization processing are executed in the image decoding apparatus 1 according to the present embodiment. FIG.

4 and 8, there is no difference in the processing time of the first Y0 block. In FIG. 4 and FIG. 8, the time T _A, the processing time required for the determination process of DC coefficient scaling and prediction direction. The time _{T B} is the processing time required for the AC coefficients scaling. The time T _C, the reverse scan process, a processing time required for the addition processing and the inverse quantization process.

  However, there is a difference between FIG. 4 and FIG. 8 regarding the processing of subsequent Y1-Cr blocks. As is clear from the operation of the image decoding device 1 described above and FIG. 4, the image decoding device 1 is configured to perform DC coefficient scaling, prediction direction determination, which is part of the inverse DC / AC prediction processing of the Y1 to Cr blocks, In addition, the scaling of the AC coefficient is executed in parallel with the inverse scan processing, addition processing, and inverse quantization processing of the preceding block (Y0 block in the case of Y1 block) of each block. For this reason, in FIG. 4, the processing time until the inverse quantization of the Y1 to Cr blocks is completed is shortened compared to FIG.

For example, the DC coefficient scaling and processing clock number C _A prediction directions of _{discrimination,} the processing clock number C _B DC coefficient _scaling, and inverse scan processing, addition processing, and inverse quantization processing of the processing clock number C _C respectively following values Assume that
C _A : 9 clocks C _B : 11 clocks C _C : 42 clocks In this case, the number of clocks required for processing of one conventional macroblock shown in FIG. 8 is 6 (C _A + C _B + C _C ), that is, 372 clocks. Become. On the other hand, the number of clocks required for processing one macroblock by the image decoding apparatus 1 shown in FIG. 4 is C _A + C _B + 6C _C , that is, 272 clocks.

The ability to reduce the number of processing clocks required for inverse encoding in this way also contributes to reducing the operating frequency of the image decoding apparatus. For example, in the example of FIG. 8, the operating frequency required to decode 1920 × 1080 pixel full HD (High-Definition) size image data at 1 × speed based on the values of C _A , C _B and C _C described above. In contrast to 91.0 MHz, the example in FIG. 4 is reduced to 66.6 MHz.

  In the above-described embodiment, the example in which the DC coefficient scaling of the next block, the determination of the prediction direction, and the AC coefficient scaling are executed in parallel with the inverse quantization process of the current block is shown. However, for example, the DC block scaling of the next block and the determination of the prediction direction may be executed in parallel with the inverse quantization processing of the current block. Even with such a configuration, it is possible to shorten the overall processing time required to complete the inverse quantization of the macroblock by the amount of parallelization of the inverse prediction processing of the DC coefficient.

  In the above-described embodiments, the decoding of images encoded using DCT used in many codecs such as VC-1 and MPEG-4 has been described. However, DCT is only one method for expressing an image signal in spatial frequency, and the present invention is not limited to DCT and can be applied to decoding processing of an image encoded using another orthogonal transform method. is there

  Further, the above-described image decoding device 1 is applicable not only to decoding VC-1, but also to decoding image data that has been compression-encoded using another codec technique that performs AC / DC predictive encoding.

  Furthermore, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the gist of the present invention described above.

It is a block diagram of the image decoding apparatus concerning embodiment of this invention. It is a figure which shows the specific example of a periphery DC / AC coefficient buffer. It is a figure which shows the scanning direction by the zigzag scan at the time of the compression encoding of the block of 8x8 pixels and the said block. The figure for demonstrating the processing time at the time of performing the reverse scan process of a VC-1 macroblock, a reverse DC / AC prediction process, and a reverse quantization process in the image decoding apparatus concerning embodiment of this invention. is there. It is a block diagram of the conventional image decoding apparatus. It is a figure which shows the positional relationship of the periphery lock | rock for determining the prediction direction of DC / AC prediction, and a current block. It is a figure which shows the some block in a VC-1 macroblock, and the process order with respect to these blocks. It is a figure for demonstrating the processing time at the time of performing the reverse scan process of a VC-1 macroblock, a reverse DC / AC prediction process, and a reverse quantization process in the conventional image decoding apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image decoding apparatus 10 Variable length decoding part 11 Input buffer 12 Pipeline process part 13 Sequence process part 14 External memory 15 Output buffer 16 Inverse DCT part 20 Pipeline control part 30 Inverse quantization part 301 Inverse scan process part 302 Addition process Unit 303 inverse quantization unit 40 inverse DC / AC prediction processing unit 401 control unit 402 parameter calculation unit 403 scaling unit 404 prediction direction determination unit 50 peripheral DC / AC coefficient buffer

Claims (6)

Orthogonal transform, quantization of DC / AC coefficient obtained by the orthogonal transform, DC / AC predictive coding of the quantized DC / AC coefficient obtained by the quantization, and DC / obtained by the DC / AC predictive coding An image decoding apparatus that performs decoding on an encoded image signal that has been compression-encoded by variable-length encoding of an AC difference coefficient,
Inverse scan processing for extracting the DC / AC difference coefficient related to the current block to be processed from the data string after performing variable length decoding on the encoded image signal, and the DC / AC difference coefficient of the current block Addition processing for restoring the quantized DC / AC coefficient by adding a DC / AC prediction coefficient corresponding to a prediction direction of DC / AC predictive coding, and the quantization DC / AC coefficient of the restored current block An inverse quantization unit that performs an inverse quantization process;
An inverse DC / AC prediction processing unit for determining the prediction direction required for the addition processing,
The inverse DC / AC prediction processing unit uses the quantized DC / AC coefficients restored by the inverse quantization unit related to the block before the current block, thereby performing the inverse scan process and the addition process related to the current block. In parallel with the execution of the inverse quantization process, the prediction direction determination process for the next block to be processed following the current block is executed.
Image decoding device. The quantization unit is
When the result of the addition process of at least the DC coefficient of the DC / AC coefficients related to the current block is the DC prediction coefficient of the next block, the inverse DC is determined according to the end of the addition process of the DC coefficient. / Instruct the AC prediction processing unit to start processing,
When the result of the DC coefficient addition process is not used as the DC prediction coefficient of the next block, the inverse DC / AC prediction processing unit is instructed to start the process before the end of the DC coefficient addition process To
The image decoding apparatus according to claim 1.   The image decoding device according to claim 1, wherein the inverse quantization unit pipelines the inverse scan process, the addition process, and the inverse quantization process for each block. The inverse DC / AC prediction processing unit performs scaling processing of the DC / AC prediction coefficient, and supplies the DC / AC prediction coefficient after scaling to the inverse quantization unit.
The inverse quantization unit performs the addition process using the scaled DC / AC prediction coefficient supplied from the inverse DC / AC prediction processing unit.
The image decoding device according to any one of claims 1 to 3. Orthogonal transform, quantization of DC / AC coefficient obtained by the orthogonal transform, DC / AC predictive coding of the quantized DC / AC coefficient obtained by the quantization, and DC / obtained by the DC / AC predictive coding An image decoding method for an encoded image signal that is compression-encoded by variable-length encoding of an AC difference coefficient,
Inverse scan processing for extracting the DC / AC difference coefficient related to the current block to be processed from the data string after performing variable length decoding on the encoded image signal, and the DC / AC difference coefficient of the current block Addition processing for restoring the quantized DC / AC coefficient by adding a DC / AC prediction coefficient corresponding to a prediction direction of DC / AC predictive coding, and the quantization DC / AC coefficient of the restored current block A step (a) of performing an inverse quantization process;
By using the quantized DC / AC coefficients related to the block before the current block, the current block is executed in parallel with the execution of the inverse scan processing, the addition processing, and the inverse quantization processing regarding the current block. (B) executing the prediction direction determination process for the next block to be processed subsequent to
A decoding method comprising: When the result of the addition process of at least the DC coefficient of the DC / AC coefficients related to the current block is the DC prediction coefficient of the next block, the addition process of the DC coefficient in the step (a) is completed. In response, the parallel processing of the step (b) is started,
When the result of the DC coefficient addition process is not used as the DC prediction coefficient of the next block, the parallel processing of the step (b) is performed before the end of the DC coefficient addition process in the step (a). Start,
The image decoding method according to claim 5.

Images