JP2005318559A - Inverse prediction apparatus and decoding apparatus using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a highly diversified inverse prediction apparatus that reduces hardware costs at the time of mounting and is not limited by image size or the like.
SOLUTION: Inverse prediction means 12 receives decoding information 1 that is variable-length decoded data of a target macroblock and target macroblock information 2 that is macroblock information of the target macroblock, and luminance / color difference reference value storage means 3, a reference value initialization unit 4, an inverse prediction calculation unit 5, a luminance / color difference prediction value storage unit 6, and a next reference value update unit 7. During the processing of the macro block of interest, the reference value initialization unit 4 determines whether or not the reference block needs to be initialized based on five patterns based on a certain standard. Since the next reference value update unit 7 calculates reference block data used for the subsequent and subsequent prediction processes, it is not necessary to store the block information of the reference block until the prediction process of the target macroblock of the next line. Hardware cost during mounting can be reduced.
[Selection] Figure 1

Description

  The present invention relates to a reverse prediction apparatus that performs reverse prediction processing on data of a macroblock of interest that has been subjected to variable length decoding, and a decoding apparatus that uses the reverse prediction apparatus.

  In recent years, information devices and information terminal devices that handle moving images have spread dramatically, and techniques for compressing moving images are being standardized.

  MPEG-4 defined by MPEG (Moving Picture Experts Group) is one standard for image compression technology.

  The book: “All about MPEG-4” (Miki et al., 1998, published by Industrial Research Co., Ltd.) explains the MPEG-4 standard in principle.

  One screen data according to MPEG-4 is called VOP (Video Object Plane).

  The VOP is configured by arranging a plurality of macro blocks (rectangular regions of 16 × 16 pixels) vertically and horizontally. A macroblock is a basic unit of VOP processing.

  Macroblocks include intra macroblocks (macroblocks that are subject to intra-frame encoding) and intermacroblocks (macroblocks that are not intra macroblocks).

  Each of the intra macro book and the inter macro block is composed of 6 blocks (4 luminance component blocks and 2 color difference component (Cb, Cr) blocks) each having 8 × 8 pixels.

  In order to prevent error propagation at the time of decoding, MPEG-4 allows a synchronization word (a fixed length signal composed of a specific bit pattern) to be placed in a bit stream. Even if a decoding error occurs in the middle of the bitstream, if a synchronization word is detected, the error is not propagated after the synchronization word, and correct decoding can be performed after the synchronization word.

  A VP (Video Packet) is a unit that is composed of one or more macroblocks and holds a synchronization word and encoded information that follows the synchronization word. In other words, one VOP can be divided into arbitrary VPs, and one VP can be divided into arbitrary macroblocks.

  However, the macroblocks have the following order. That is, starting from the upper left macroblock of one VOP, the process proceeds to the right, and when the rightmost macroblock is reached, the process proceeds to the leftmost macroblock in the row immediately below. Going further to the right, the end is the macroblock at the bottom right of the VOP. For this reason, VP is determined according to this order.

  The MPEG-4 standard defines the following method for inverse AC / DC prediction (in this specification, simply referred to as “reverse prediction”) processing. The inverse prediction process is performed in units of blocks.

  FIG. 11 is an explanatory diagram of conventional inverse DC prediction, and FIG. 12 is an explanatory diagram of conventional inverse AC prediction.

  First, the inverse DC prediction will be described with reference to FIG. In FIG. 10, block X is a block to be predicted (hereinafter referred to as the target block), block A is a block adjacent to the left side of block X, block B is a block adjacent to diagonally above block X, and block C is above block X. Adjacent block. Blocks A, B, and C are all blocks that have been subjected to inverse prediction processing before the target block.

  Fa [0] [0] is the DC component after inverse quantization of block A, Fb [0] [0] is the DC component after inverse quantization of block B, and Fc [0] [0] is the block C It is assumed that it is a DC component after inverse quantization.

  In inverse DC prediction, first, a reference block is determined by the gradient between adjacent blocks around the block of interest X.

  if,

If so, block C is a reference block, otherwise block A is a reference block.

  If any of the adjacent blocks A, B, and C corresponds to the outside of the VOP, the outside of the VP, or the inter macro block, Fa [0] [0], Fa [0] [0], Fc [0] [0] is assumed to be 2 ^ (N + 2) (N is the number of bits of the pixel), and the gradient is obtained.

  Next, the DC coefficient (QFx [0] [0]) of the block X is inversely predicted using the DC component of the reference block determined in this way.

  When block C is a reference block

  When block A is a predicted reference block

  Here, QFx [0] [0] is the DC component of the inverse prediction result of block X, PQFx [0] [0] is the DC component of the target block X in the variable-length decoded data, and dc_scaler is the block This is the quantization scale when X is quantized. The operator “//” means “divide and round toward 0”.

  The above process is repeated until all 6 blocks of DC coefficients (4 luminance blocks, 2 color difference blocks) are obtained.

  Next, inverse AC prediction will be described with reference to FIG. In the inverse AC prediction process, the reference block determined during the inverse DC prediction process is referred to.

  In FIG. 12, blocks X, A, B, and C are the same as in FIG. QFa [0] [i] (i is 1, 2, 3, 4, 5, 6, 7) is an AC coefficient of block A, and QFc [j] [0] (j is 1, 2, 3, 4, 5, 6, 7) are AC coefficients of block C.

  When the prediction reference block is block A, the AC coefficient (QFx [0] [i]) of block X is inversely predicted as follows.

  When the prediction reference block is block C, the AC coefficient (QFx [j] [0]) of block X is inversely predicted as follows.

  QFx [j] [i] is the AC component of the inverse prediction result of block X, PQFx [j] [i] is the AC component of variable-length decoded target block X, and QPx is a quantized block X Quantization scale, QPa is a quantization scale when block A is quantized, and QPc is a quantization scale when block C is quantized. The operator “//” means “divide and round toward 0”.

  When any of the adjacent block A and block C corresponds to the outside of the VOP, the outside of the VP, or the inter macro block, QFa [0] [i] and QFc [j] [0] are assumed to be zero. And reverse prediction is performed.

  As described above, the inverse prediction process defined by MPEG-4 is very complicated.

  Reference 1 (Japanese Patent Laid-Open No. 2002-118853) discloses an implementation method for reverse prediction processing. This mounting method will be described with reference to FIGS.

  FIG. 13 shows a conventional image decoding apparatus for decoding an intra macroblock. This image decoding apparatus includes an attention macroblock setting unit 100, an attention macroblock extraction unit 101, an inverse prediction unit 102, an inverse quantization unit 103, and an inverse DCT unit 104.

  The target macroblock setting unit 100 sets a target macroblock (for example, by specifying a block number) in the VOP.

  The target macroblock extraction unit 101 performs variable length decoding processing on the encoded image signal to extract pixel data of the target macroblock.

  The inverse prediction unit 102 performs inverse prediction processing on the extracted pixel data of the target macroblock, and outputs a quantized coefficient after inverse prediction.

  The inverse quantization means 103 performs an inverse quantization process on the quantization coefficient of the macro block of interest and outputs a DCT coefficient.

  The inverse DCT unit 104 performs inverse DCT transform on the DCT coefficient of the target macroblock, and outputs a reproduced image of the target macroblock.

  The image decoding apparatus in FIG. 13 repeats the above processing for the number of times of all macroblocks included in the VOP. As a result, a playback image of one VOP is stored in the image memory 105.

  FIG. 14 is a block diagram of conventional inverse prediction means. As shown in FIG. 14, the inverse prediction means 102 includes a prediction control means 111, a luminance reference value storage means 112, a color difference reference value storage means 113, a luminance prediction value storage means 114, a color difference prediction value storage means 115, and an inverse prediction calculation means. 116.

  The luminance reference value storage unit 112 holds a reference value for the luminance block of the macro block of interest. The color difference reference value storage unit 113 holds a reference value for the color difference block of the macro block of interest.

  The reverse prediction calculation unit 116 performs reverse prediction calculation based on the reference values held in the luminance reference value storage unit 112 and the color difference reference value storage unit 113.

  The luminance predicted value storage unit 114 holds the inverse prediction calculation result of the luminance block among the outputs of the inverse prediction calculation unit 116.

  The color difference predicted value storage unit 115 holds the result of the inverse prediction calculation of the color difference block among the outputs of the inverse prediction calculation unit 116.

  The prediction control unit 111 transfers data necessary for the inverse prediction calculation of the target macroblock from the luminance prediction value storage unit 114 and the color difference prediction value storage unit 115 to the luminance reference value storage unit 112 and the color difference reference value storage unit 113, respectively. And copy.

  Next, a detailed configuration of each unit will be described with reference to FIG.

  As shown in FIG. 15, the luminance reference value storage means 112 includes a line portion 121 that holds a DC component and an AC component for one line of VOP size, a corner portion 122 that holds one DC component, and a set of And a left side portion 123 that holds a DC component and an AC component.

  The DC component and the AC component held by the luminance reference value storage unit 112 are reference data necessary for performing prediction processing for luminance.

  Two sets (16 pixels) of DC components and AC components are assigned to the line unit 121 for the four luminance components of each macro block of interest. For example, as shown in FIG. 15, in the case of performing reverse prediction on a VOP in which one line is composed of five macroblocks, a storage area of 16 pixels × 5, that is, 80 pixels is required.

  For the four luminance components of the macro block of interest, two sets (16 pixels) of DC components and AC components are assigned to the left side portion 123.

  The luminance predicted value storage unit 114 holds the luminance DC component and AC component for one macro block of interest.

  The DC component and the AC component held by the predicted luminance value storage unit 114 are data of the inverse prediction processing result for luminance.

  Similarly, although not shown in FIG. 14, the color difference reference value storage unit 113 also has a line portion that holds a DC component and an AC component for one line of the VOP size, a corner portion that holds one DC component, and A left side that holds a set of DC and AC components.

  Note that the DC component and the AC component held by the color difference reference value storage unit 113 are prediction reference data necessary for performing prediction processing for color differences.

  Among these, for the two color difference components of each macro block of interest, a DC component and an AC component are assigned to one line for each set (16 pixels in total).

  That is, as shown in FIG. 14, when predicting a VOP size image consisting of five macroblocks per line, a storage area of 16 pixels × 5, that is, 80 data is required.

  Also for the two color difference components of the macro block of interest on the left side, one set of DC components and one AC component (16 pixels in total) are allocated.

  The color difference prediction value storage unit 115 holds the DC component and the AC component of the color difference for one target macroblock.

  Note that the DC component and AC component held by the color difference predicted value storage unit 115 are the result data of the prediction process for the color difference.

  A conventional inverse prediction process will be described with reference to FIG.

  In step 50 of FIG. 16, the inverse prediction calculation means 6 performs an inverse prediction calculation on the upper left luminance block of the target macroblock according to the MPEG-4 system. The obtained prediction result is held in the luminance predicted value storage unit 4.

  In step 51, it is determined whether or not the macro block of interest is at the left end of the VOP.

  In step 52, the prediction control unit 111 initializes the corner and left side of the luminance reference value storage unit 112 and the corner and left side of the color difference reference value storage unit.

  In step 54, the reverse prediction calculation means 116 performs reverse prediction calculation according to the MPEG-4 method for each of the upper right, lower left, and lower right luminance blocks of the macro block of interest. The obtained prediction result is held in the luminance predicted value storage unit 114.

  In step 55, the prediction control means 111 changes the luminance reference value storage means from the luminance prediction value storage means 114 to use the DC component and AC component for the lower left luminance block as prediction reference data at the time of prediction of the next line. Copy to line part 121 of 112.

  In step 56, the prediction control means 111 uses the luminance predicted value storage means 114 in order to use the DC component and AC component for the upper right and lower right luminance blocks as prediction reference data at the time of prediction of the right adjacent macroblock. Are copied to the left side 123 of the luminance reference value storage means 112.

  In step 57, the reverse prediction calculation means 116 performs reverse prediction calculation of the target macroblock for the color differences Cb and Cr according to the MPEG-4 system. The obtained predicted value is held in the color difference predicted value storage unit 115.

  In step 58, the prediction control unit 111 copies the color difference DC component and AC component from the color difference predicted value storage unit 115 to the color difference reference value storage unit 113. The prediction control unit 111 copies the DC component and AC component of the color difference to the left side. The meaning of this copy is the same as in step 55.

  If it is determined in step 51 that the target macroblock is not at the left end of the VOP, the process proceeds to step 53. In step 53, the prediction control unit 111 copies the DC component and the AC component from the luminance predicted value storage unit 114 to the line unit 121 of the luminance reference value storage unit 112 for the lower right luminance block of the target macroblock.

  When the prediction process for the current macroblock of interest for one line is completed, reference data necessary for prediction of the next macroblock for one line is held in the luminance reference value storage unit 112 and the color difference reference value storage unit 114. As a result, the memory area used for the prediction process can be reduced.

  However, even with the above method, the correspondence to VP and the case where the reference macroblock is an inter macroblock are insufficient. Specifically, when the reference block position is outside the VP, or when the reference macroblock is an inter macroblock, the initialization process of the reference macroblock can be incomplete.

  In the above method, since the line portion holds data for one line of the VOP, a trouble is likely to occur when the size of one line of the VOP changes. In reality, it is unavoidable to secure an area for the maximum possible one line. In other words, there is a need for a technique that can flexibly cope with a case where the size of one line can be changed with a smaller memory capacity.

Further, since the above method uses the quantization scale (QPa, QPc) at the time of quantization of the reference block, the area for storing the quantization scale of the adjacent block is the luminance reference value storage means or the color difference reference. Apart from the value storage means, one line of the image size is required.
JP 2002-188553 A

  Therefore, an object of the present invention is to provide an inverse prediction device that is low in cost and highly applicable.

  A reverse prediction apparatus according to a first aspect of the present invention is an inverse prediction apparatus that performs reverse prediction processing on data of a target macroblock that has been subjected to variable length decoding, and holds a reference value necessary for reverse prediction processing of the target macroblock A value storage unit and an inverse prediction calculation unit that performs an inverse prediction calculation based on a reference value held by the reference value storage unit and generates an inverse prediction result, and the reference value storage unit is adjacent to the upper side of the target macroblock. The upper part that holds the AC component and DC component of the macroblock immediately above, the corner part that holds the DC component of the macroblock located diagonally to the left of the target macroblock, and the immediately preceding adjacent to the left side of the target macroblock It consists of the left side that holds the AC and DC components of the macroblock.

  According to this configuration, the reference value storage means for one line is necessary in the prior art, and it is sufficient to provide the reference value storage means for one macroblock at the minimum. That is, a constant configuration can always be taken regardless of the VOP size. Moreover, the reverse prediction process can be performed with a small storage capacity.

  In the inverse prediction apparatus according to the second invention, the AC component and DC component held by the upper part are composed of the DC values and AC values of two luminance blocks and two color difference blocks, and the DC component held by the corner part. Is composed of DC values of one luminance block and two color difference blocks, and the AC component and DC component held on the left side are composed of DC values and AC values of two luminance blocks and two color difference blocks. .

  With this configuration, it is sufficient that the upper side can store the AC component and DC component for one macroblock, so that the reverse prediction process can be performed with a small storage capacity without depending on the number of macroblocks for one line of VOP. Can be implemented. Thereby, the cost of a reverse prediction apparatus can be reduced and diversion property can be improved.

  The inverse prediction apparatus according to the third aspect of the present invention further includes reference value initialization means for initializing the reference value of the corresponding block among the five luminance blocks and the six color difference blocks based on a certain standard.

  With this configuration, the reference value initialization process can be appropriately performed.

  In the inverse prediction apparatus according to the fourth aspect of the invention, the constant criterion is defined based on the position of the target macroblock on the video object plane and the position of the video packet to which the target macroblock belongs.

  In the inverse prediction apparatus according to the fifth aspect, the position of the macro block of interest and the position of the video packet are evaluated by two-dimensional coordinates on the video object plane.

  With these configurations, a block to be initialized can be easily and appropriately determined using two-dimensional coordinates. That is, it is not necessary to store the block information (VP number, macroblock type, quantization scale) of the reference block until the prediction processing of the target macroblock of the next line.

  In the inverse prediction apparatus according to the sixth aspect of the invention, the reference value initialization means includes a first pattern in which the values of all of the five luminance blocks and the six chrominance blocks are initialized according to a certain standard. Of the five luminance blocks and the six color difference blocks, the second pattern in which the values of the seven blocks in contact with the upper side of the macro block of interest are initialized, the five luminance blocks, and the six color difference blocks Among the blocks, among the third pattern in which the values of the seven blocks in contact with the left side of the target macroblock are initialized, the five luminance blocks, and the six color difference blocks, the upper left diagonally of the target macroblock The process branches to the fourth pattern in which the values of the three blocks located in the area are initialized, and the fifth pattern in which the values of any of the five luminance blocks and the six color difference blocks are not initialized.

  With this configuration, it is possible to branch to five patterns and clearly determine the block to be initialized.

  The inverse prediction apparatus according to a seventh aspect of the present invention further comprises reference value initialization means for initializing a reference value held by the reference value storage means, wherein the reference value initialization means is configured such that the macroblock of interest is an intra macroblock / intermacro. It is determined which one of the blocks, and when the target macroblock is an inter macroblock, at least a part of the reference value held by the reference value storage unit is updated.

  In the prior art, since an inter macroblock is not subjected to AC / DC prediction processing, it has been difficult to cope with a case where a reference block is an inter macroblock. However, with this configuration, it is possible to perform reverse prediction processing using the reference value storage means even when the macro block of interest becomes a reference block in the future.

  In the inverse prediction apparatus according to the eighth aspect, when the target macroblock is an intra macroblock, the reference value initialization unit multiplies the inverse prediction result generated by the inverse prediction calculation unit by the quantization scale to obtain the multiplication result. The reference value held by the reference value storage means is updated so that the multiplication result is the reference value of the macroblock immediately to the right of the target macroblock and the reference value of the macroblock immediately below the target macroblock.

  With this configuration, the quantization scale can be reflected in the reference value.

  In the inverse prediction apparatus according to the ninth aspect, when the macro block of interest is an inter macro block, the reference value initialization means sets the DC value to 2 ^ (N + 2) (N: the number of bits of the pixel) and the AC value to zero. The reference value held by the reference value storage unit is updated so that the reference value becomes the reference value of the macroblock immediately adjacent to the macroblock of interest and the reference value of the macroblock immediately below the macroblock of interest.

  With this configuration, initialization conforming to the standard can be performed.

  According to the present invention, the mounting cost of the data storage area necessary for prediction can be suppressed. Also, even if the image size to be decoded or encoded is changed, the capacity of the data storage area necessary for prediction is constant, so that it is easy to cope without changing the design and the applicability is high.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram of an inverse prediction apparatus according to Embodiment 1 of the present invention. The decoding device of this embodiment basically has the configuration shown in FIG. Therefore, detailed description of components of the decoding apparatus that are not the inverse prediction means 102 is omitted.

  However, as will be described below, the content of the inverse prediction means 102 of this embodiment is significantly different from that of FIG.

  The inverse prediction means 102 inputs data (decoding information 1) obtained by variable-length decoding for the target macroblock and target macroblock information 2. The inverse prediction unit 102 includes a luminance / color difference reference value storage unit 3, a reference value initialization unit 4, an inverse prediction calculation unit 5, a luminance / color difference prediction value storage unit 6, and a next reference value update unit 7. More specifically, the decoding information 1 is input to the inverse prediction calculation unit 5 of the inverse prediction unit 102, and the target macroblock information 2 is input to the reference value initialization unit 4 of the inverse prediction unit 102.

  The luminance / color difference reference value storage means 3 outputs the original reference data 8 to the reference value initialization means 4.

  The reference value initialization unit 4 outputs the initialized reference data 9 to the inverse prediction calculation unit 5. The reference value initialization unit 4 outputs the inter macroblock instruction flag 10 to the next reference value update unit 7.

  The inverse prediction calculation unit 5 outputs the luminance / color difference prediction value 11 to the luminance / color difference prediction value storage unit 6 and outputs the luminance / color difference prediction value 11 to the inverse quantization unit 103 as an inverse prediction calculation result.

  The luminance / color difference predicted value storage unit 6 outputs the luminance / color difference predicted value to the inverse prediction calculation unit 5 and the next reference value update unit 7.

  The next reference value update unit 7 outputs the update data 14 to the luminance / color difference reference value storage unit 3.

  The luminance / color difference reference value storage means 3 holds a reference value necessary for inverse prediction processing of the target macroblock. However, the reference value held by the luminance / color difference reference value storage unit 3 in this embodiment is obtained by multiplying the inverse prediction result (QF) by the quantization scale (QP) of the macro block of interest.

  The reference value initialization unit 4 determines whether or not the reference block is outside the VOP or VP based on the target macroblock information 2, and when it is outside, the luminance / color difference reference value storage unit 3 holds it. Initialize output. The attention macroblock information 2 is generated by the attention macroblock setting unit 100, for example.

Attention macroblock information 2 is
Whether the target macroblock is an inter macroblock or an intra macroblock,
Xy coordinate values (cMBx, cMBy) of the target macroblock and xy coordinate values (cVPx, cVPy) of the first macroblock of the VP to which the target macroblock belongs
Is sufficient.

  The reference value initialization unit 4 refers to the target macroblock information 2, and instructs the next reference value update unit 7 to indicate whether the target macroblock is an inter macroblock or an intra macroblock. 10 is output.

  The inverse prediction calculation means 5 performs an inverse prediction calculation on the decoded information 1 using the output of the reference value initialization means 4 as a reference value, and the inverse prediction result (luminance / color difference prediction value 11) is used as the luminance / color difference prediction value storage means 6. And output to the inverse quantization means 104.

  The luminance / color difference predicted value storage unit 6 holds the reverse prediction result output by the reverse prediction calculation unit 5. The inverse prediction result is returned to the inverse prediction calculation unit 5 as the reference value of the luminance block in the target macroblock, and is output to the next reference value update unit 7 as the reference value of the subsequent target macroblock.

  The next reference value updating means 7 calculates the reference value used in the next left macroblock and the macroblock immediately below based on the inter macroblock instruction flag 10 and the luminance / color difference reference value storage means 3. Update the pixel position at. This process will be described in more detail later with reference to FIG.

  Next, the relationship between the macro block of interest and the reference block will be described with reference to FIG. The macro block is a set of four luminance blocks and two color difference blocks. Each of these six blocks consists of a rectangular area of 8 × 8 pixels.

  That is, as shown on the left side of FIG. 3, four luminance blocks Y0 to Y3 and two color difference blocks Cb and Cr belong to one macroblock.

  For the inverse prediction processing of the four luminance blocks Y0 to Y3, five reference blocks A to E that are in contact with the upper side, the left side, and the upper left side of the four luminance blocks Y0 to Y3 are necessary.

  In order to perform inverse prediction processing of one color difference block Cb, three reference blocks L to N that are in contact with the upper side, the left side, and the left upper side of one color difference block Cb are necessary.

  In order to perform reverse prediction processing of one color difference block Cr, three reference blocks X to Z that are in contact with the upper side, left side, and left upper side of one color difference block Cr are necessary.

  Eventually, 11 reference blocks A to E, L to N, and X to Z are necessary for inverse prediction of one macro block.

  The reference value initialization process by the reference value initialization unit 4 is performed by applying a DC component to a reference block located outside the VOP or outside the VP among the 11 reference blocks A to E, L to N, and X to Z. This is a process of setting 2 ^ (N + 2) (N is the number of bits of the pixel) and zeroing the AC component.

  Next, details of each component will be described. The luminance / color difference reference value storage means 3 includes a line unit 121 that holds DC components and AC components for one line in one VOP, a corner portion 122 that holds one DC component, a set of DC components and AC. And a left side 123 for holding the components.

  Two sets (16 pixels) of a DC component and an AC component are assigned to the line unit 121 for the four luminance components of each target macroblock. For each of the two color difference components, two sets (16 pixels) of a DC component and an AC component are allocated.

  As shown in FIG. 15, when one line of one VOP is composed of five macroblocks, the line part 121 is an area of (16 pixels + 16 pixels) × 5, that is, 160 pixels.

  On the left side 123, two sets (16 pixels) of the DC component and the AC component are assigned to the four luminance components, and two sets (16 pixels) of the DC component and the AC component are assigned to the color difference component. Therefore, the left side portion 123 is an area of 16 pixels + 16 pixels, that is, 32 pixels.

  Since the corner 122 is assigned one set of DC components for the four luminance components and one set of DC components for each of the two color difference components, the corner 122 is an area for three pixels.

  Here, the DC component and the AC component held by the luminance / color difference reference value storage unit 3 are not values obtained by copying the inverse prediction result as a reference value as in the prior art, but the inverse prediction result (QF) It is a value obtained by multiplying the quantization scale (QP) of the macro block of interest.

  The luminance / color difference predicted value storage means 6 is the same as that of the prior art shown in FIG. Note that the DC component and the AC component held by the luminance / color difference predicted value storage unit 6 are the result data of the reverse prediction process.

  Of course, the luminance / color difference reference value storage unit 3 and the luminance / color difference predicted value storage unit 6 may be configured as separate storage media, or may be provided as separate areas of one storage medium (for example, a memory). .

  Next, the inverse prediction calculation in the present embodiment will be described with reference to FIGS. In order to perform prediction calculation for one block of interest X, three blocks on the left side (block A), diagonally above (block B), and upper side (block C) of the block of interest X are referred to.

  That is, in step 1 of FIG. 2, the inverse prediction calculation means 5 refers to the reference value initialization means 4 and the luminance / color difference prediction value storage means 5, and data Fa [0] [0], Fb [0] [ 0], Fc [0] [0]. Here, the subscript “a” in “Fa [0] [0]” indicates the block A.

  In step 3, the inverse prediction calculation means 5 obtains two absolute values abs1 and abs2, and in step 3, the inverse prediction calculation means 5 performs determination of (Equation 1), and processing is performed if the result is true. Advances to step 4, and if false, the process advances to step 5.

  In step 4, the inverse prediction calculation means 5 sets the reference block as block C, and substitutes Fc [0] [0] for the reference value DC component F [0] [0] in order to perform the subsequent inverse prediction calculation. .

  In step 5, the inverse prediction calculation means 5 sets the reference block to block A, and substitutes Fa [0] [0] for the reference value DC component F [0] [0] in order to perform the subsequent inverse prediction calculation. .

  In step 6, the inverse prediction calculation means 5 calculates the DC prediction result (QFx [0] [0]) of the block of interest X. The calculation formula is as follows, and this formula is equivalent to (Equation 2) and (Equation 3).

  PQFx [0] [0] is DC data of the block of interest X subjected to variable length decoding processing, given by decoding information 1 (see FIG. 1), and dc_scaler is given by decoding information 1 (see FIG. 1). The quantization scale of the block of interest X subjected to variable length decoding.

  In step 7, when the reference block is block C, the process proceeds to step 8, and when the reference block is block A, the process proceeds to step 9.

  In step 8, the inverse prediction calculation means 5 obtains the AC prediction result (QFx [j] [0], j = 1, 2,..., 7) of the block of interest using the following equation.

  The value Fc [j] [0] in (Expression 7) is obtained by multiplying the inverse prediction result QFc [j] [0] of the block C by the quantization scale of the block C during the inverse prediction process of the block C. It is held in the luminance / color difference reference value storage means 3 (see FIG. 1). Therefore, (Equation 7) is equivalent to (Equation 4).

  PQFx [j] [0] is AC data of the target block X subjected to variable length decoding processing obtained from decoding information 1 (see FIG. 1), and QPx is obtained from decoding information 1 (see FIG. 1). The quantization scale of the block of interest X subjected to variable length decoding.

  In step 8, the inverse prediction calculation means 5 obtains the AC prediction result (QFx [0] [i], i = 1, 2,..., 7) of the block of interest using the following equation.

  Fa [0] [i] in (Equation 8) is obtained by multiplying the inverse prediction result QFa [0] [i] of the block A by the quantization scale of the block A during the inverse prediction processing of the block A. It is held in the luminance / color difference reference value storage means 3 (see FIG. 1). Therefore, (Equation 8) is equivalent to (Equation 5).

  PQFx [0] [i] is AC data of the block of interest X obtained from the decoding information 1 (see FIG. 1) and subjected to variable length decoding processing.

  As described above, through steps 1 to 9, the inverse prediction calculation unit 5 outputs the prediction result (the luminance / color difference prediction value 11) of the block of interest X.

  Next, the processing of the reference value initialization unit 4 will be described. First, the four two-dimensional coordinate values cMBx, cMBy, cVPx, and cVPy used in this embodiment will be described with reference to FIG. As described above, the four two-dimensional coordinate values cMBx, cMBy, cVPx, and cVPy can be extracted from the target macroblock information 2.

  As shown in FIG. 10, VOP1 is composed of VP1, VP2,. VP1 is a set of a plurality of macroblocks m1 to m17, and VP2 is a set of a plurality of macroblocks m18 to m33.

  The upper left corner point of VOP1 (upper left corner point of macroblock m1) is defined as the origin O (0, 0), and the horizontal axis (X axis) and the vertical axis (Y axis) are defined. As a result, the XY plane exists on the plane formed by VOP1.

  The upper left corner point of the macro block of interest is defined as (X, Y) = (cMBx, cMBy), and the upper left corner point of the first macro block of VP is defined as (X, Y) = (cVPx, cVPy). More specifically, these four two-dimensional coordinate values may be macroblock numbers or their function values.

  (CVPx, cVPy) of VP1 is the upper left corner point of the macroblock m1, and (cVPx, cVPy) of VP2 is the upper left corner point of the macroblock m18. This is because, as shown in the lower part of FIG. 10, the processing direction of the macroblock follows the order described in the section of the prior art (see the arrow).

  As mentioned above, the process of determining the block to be initialized is very complex. However, in conclusion, according to the present embodiment, by using these four two-dimensional coordinate values, the state can be divided into five patterns and a block to be initialized can be determined at high speed by a simple calculation. . And since it does not depend on the size of VOP or VP, the versatility and applicability of the inverse prediction device can be improved.

  FIG. 4 is a flowchart of the reference value initialization process in the first embodiment of the present invention.

  In step 10, the reference value initialization unit 4 extracts four coordinate values cMBx, cMBy, cVPx, and cVPy from the target macroblock information 2.

In step 11, by using these coordinate values cMBx, cMBy, cVPx, cVPy,
(Condition 1) Whether the target macroblock position is the head of the VOP,
(Condition 2) Whether the target macroblock is the head of the VP,
(Condition 3) Whether the target macroblock is the left end of VOP and the upper end of VP,
Evaluation is performed by calculating (Equation 9), (Equation 10), and (Equation 11).

  Here, the formula for evaluating (Condition 1) is (Formula 9), the formula for evaluating (Condition 2) is (Formula 10), and the formula for evaluating (Condition 3) is (Formula 11). .

  In step 12, the reference value initialization unit 4 calculates the logical sum of (Equation 9), (Equation 10), and (Equation 11). When the logical sum is “true”, the process proceeds to Step 12; When the logical sum is “false”, the process proceeds to step 13.

  When the processing proceeds to step 13, it is found that all the reference blocks A to E, L to N, and X to Z in FIG. 3 are outside the VOP or VP. Therefore, at this time, the reference value initialization means 4 initializes the DC components of all the reference blocks A to E, L to N, and X to Z to 2 ^ (N + 2) (N is the number of bits of the pixel), and all The AC components of the reference blocks A to E, L to N, and X to Z are initialized to zero.

In step 14, the reference value initialization means 4
(Condition 4) Whether the target macroblock position is the upper end of the VOP,
(Condition 5) Whether the target macroblock position is the upper end of VP,
(Condition 6) Whether or not the position of the target macroblock coincides with the upper end of the first macroblock of the VP,
Evaluation is performed by calculating (Equation 12), (Equation 13), and (Equation 14).

  Here, the expression for evaluating (Condition 4) is (Expression 12), the expression for evaluating (Condition 5) is (Expression 13), and the expression for evaluating (Condition 6) is (Expression 14). .

  In step 13, the reference value initializing means 4 obtains the logical sum of (Equation 12), (Equation 13) and (Equation 14), and when the logical sum is “true”, the process proceeds to step 14; When the logical sum is “false”, the process proceeds to step 15.

  When the process proceeds to step 14, it is found that all the reference blocks C to E, M to N, and Y to Z in FIG. 3 are outside the VOP or VP. Therefore, at this time, the reference value initialization means 4 initializes the DC components of these reference blocks C to E, M to N, and Y to Z to 2 ^ (N + 2) (N is the number of bits of the pixel). The AC components of the reference blocks C to E, M to N, and Y to Z are initialized to zero.

In step 15, the reference value initialization means 4
(Condition 7) Whether or not the target macroblock position is the left end of the VOP is evaluated by (Equation 15). When the result is “true”, the process proceeds to step 16, and when the result is “false”, the process is Proceed to step 17.

  When the process proceeds to step 16, it is found that all of the reference blocks A to C, L to M, and X to Y in FIG. 3 are outside the VOP or VP. Accordingly, at this time, the reference value initialization means 4 initializes the DC components of these reference blocks A to C, L to M, and X to Y to 2 ^ (N + 2) (N is the number of bits of the pixel). The AC components of the reference blocks A to C, L to M, and X to Y are initialized to zero.

In step 17, the reference value initialization means 4
(Condition 8) Whether the current macroblock is directly below the position of the first macroblock of the VP to which the macroblock belongs

  When the evaluation is performed according to (Expression 16) and the result is “true”, the process proceeds to Step 18.

  When the process proceeds to step 18, it is found that all the reference blocks C, M, and Y in FIG. 3 are outside the VOP or VP. Therefore, at this time, the reference value initializing means 4 initializes the DC components of these reference blocks C, M, Y to 2 ^ (N + 2) (N is the number of bits of the pixel), and these reference blocks C, M, , Y AC component is initialized to zero.

  When the result of step 17 is “false”, the reference value initialization unit 4 does not perform any initialization and ends the process.

  Next, the process of the next reference value update unit 7 will be described with reference to FIG. FIGS. 5A to 5C show the positional relationship between the target macroblock and the reference block necessary for the prediction process of the target macroblock, and the relationship between the update source and the update destination.

  The next reference value update unit 7 calculates a reference value necessary for the reverse prediction process of the next target macroblock after the reverse prediction process of the target macroblock, and updates the luminance / color difference reference value storage unit 3.

  Specifically, the luminance block is updated as follows. As indicated by an arrow N <b> 1 in FIG. 5A, among the reference values held in the line unit 121 of the luminance / color difference reference value storage unit 3, the DC component of the block E is referred to the luminance block of the corner unit 122. Copy to the value storage area 41.

  Next, as indicated by an arrow N2 in FIG. 5A, the quantization of the macroblock of interest is performed on the block Y2 held in the luminance / color difference prediction value storage unit 6 and the prediction result data 50 above the block Y3. The scaler (QP) is multiplied and the multiplication result is copied to the reference value storage area 40 of the block D and the block E in the line part 121 of the luminance / color difference reference value storage means 3.

  However, when the macro block of interest is an inter macro block, this update is not performed, and the DC component of the reference value storage area 40 of the block D and the block E in the line portion 121 of the luminance / color difference reference value storage unit 3 is 2. ^ (N + 2) (N is the number of bits of the pixel), and the AC component is initialized to zero.

  Next, as indicated by an arrow N3 in FIG. 5A, the quantization of the target macroblock is performed on the block Y1 held in the luminance / color difference prediction value storage unit 6 and the prediction result data 51 on the left side of the block Y3. The scaler (QP) is multiplied, and the multiplication result is copied to the reference value storage areas 42 of the blocks A and B in the left side 123 of the luminance / color difference reference value storage means 3.

  However, when the target macroblock is an inter macroblock, this update is not performed, and the DC component of the reference value storage area 42 of the block A and the block B in the left side 123 of the luminance / color difference reference value storage unit 3 is 2 ^ (N + 2) (N is the number of bits of the pixel), and the AC component is initialized to zero.

  Of the color difference blocks, the block Cb is updated as follows.

  As indicated by the arrow N4 in FIG. 5B, the DC component of the block N of the reference value of the target macroblock held in the line unit 121 of the luminance / color difference reference value storage unit 3 is The color difference Cb block is copied to the reference value storage area 44.

  Subsequently, as indicated by an arrow N5 in FIG. 5B, the quantization scaler (QP) of the target macroblock is added to the prediction result data 52 on the upper side of the block Cb held in the luminance / color difference predicted value storage unit 6. The result is copied to the reference value storage area 43 of the block N in the line portion 121 of the luminance / color difference reference value storage means 3.

  However, when the target macroblock is an inter macroblock, this update is not performed, and the DC component of the reference value storage area 43 of the block N in the line unit 121 of the luminance / color difference reference value storage unit 3 is 2 ^ (N + 2). ) (N is the number of bits of the pixel), and the AC component is initialized to zero.

  Furthermore, as indicated by an arrow N6 in FIG. 5B, the quantization scaler (QP) of the macro block of interest is shown in the prediction result data 53 on the left side of the block Cb held in the luminance / color difference predicted value storage unit 6. The result is multiplied and copied to the reference value storage area 45 of the block L in the left side 123 of the luminance / color difference reference value storage means 3.

  However, when the target macroblock is an inter macroblock, this update is not performed, and the DC component of the reference value storage area 45 of the block L in the left side 123 of the luminance / color difference reference value storage means 3 is 2 ^ (N + 2). ) (N is the number of bits of the pixel), and the AC component is initialized to zero.

  Of the color difference blocks, the Cr block is updated as follows.

  As indicated by an arrow N 7 in FIG. 5C, the DC component of the block Z of the reference value of the target macroblock held in the line unit 121 of the luminance / color difference reference value storage unit 3 is It is copied to the reference value storage area 47 of the color difference Cr block.

  Subsequently, as indicated by an arrow N8 in FIG. 5, the prediction result data 54 above the block Cr held in the luminance / color difference predicted value storage unit 6 is multiplied by the quantization scaler (QP) of the target macroblock. The result is copied to the reference value storage area 46 of the block Z (36) in the line portion 121 of the luminance / color difference reference value storage means 3.

  However, when the target macroblock is an inter macroblock, this update is not performed, and the DC component of the reference value storage area 46 of the block Z in the line portion 121 of the luminance / color difference reference value storage means 3 is 2 ^ (N + 2). ) (N is the number of bits of the pixel), and the AC component is initialized to zero.

  Further, as indicated by an arrow N9 in FIG. 5C, the quantization scaler (QP) of the macro block of interest is shown in the prediction result data 55 on the left side of the block Cr held in the luminance / color difference predicted value storage unit 6. The result is multiplied and copied to the reference value storage area 48 of the block X in the left side portion 123 of the luminance / color difference reference value storage means 3.

  However, when the macro block of interest is an inter macro block, this update is not performed, and the DC component of the reference value storage area 48 of the block X in the left side 123 of the luminance / color difference reference value storage means 3 is 2 ^ (N + 2). ) (N is the number of bits of the pixel), and the AC component is initialized to zero.

  Each value held in the corner portion 122 and the left side portion 123 of the updated luminance / color difference reference value storage unit 3 is used as a reference value when the macroblock adjacent to the left of the macroblock of interest is subjected to inverse prediction processing. used. Usually, since the decoding process of VOP is performed in raster order, after the reverse prediction process related to the current macroblock being processed is completed, the macroblock on the left side of the current macroblock is identified as a new macroblock. Become.

  The value held in the line unit 121 of the updated luminance / color difference reference value storage unit 3 is used as a reference value when the macroblock immediately below the target macroblock is subjected to reverse prediction processing.

  When decoding in raster order, the macroblock immediately below becomes the target macroblock one line after the target macrobook currently being processed.

  Note that different values can be used for the quantization scaler (QP) of the target macroblock depending on the luminance block, the color difference block, the DC component, and the AC component.

  Next, the process of the reverse prediction means 102 is demonstrated, referring FIG.

  In step 30 of FIG. 6, the reference value initialization unit 4 refers to the target macroblock signal 2 and determines whether the target macroblock is an intra macroblock or an inter macroblock. If so, the process proceeds to step 31; if it is an inter macroblock, the process proceeds to step 34.

  In step 31, the reference value initialization unit 4 initializes the reference value stored in advance in the luminance / color difference reference value storage unit 3 (FIG. 4).

  In the subsequent step 32, the inverse prediction calculation means 5 performs an inverse prediction calculation on the four luminance blocks using the reference value determined in step 31 and the decoding information 1 (FIG. 2), The luminance / color difference predicted value storage means 6 stores the result.

  In step 33, the inverse prediction calculation means 5 performs an inverse prediction calculation on the two color difference blocks using the reference value determined in step 31 and the decoding information 1 (FIG. 2). The color difference predicted value storage means 6 stores the result.

  In step 34, when the target macroblock is an intra macroblock, the next reference value update unit 7 uses the inverse prediction result held in the luminance / color difference predicted value storage unit 6 and uses the luminance / color difference reference value storage unit. 3 is updated. When the target macroblock is an inter macroblock, the next reference value update unit 7 initializes the luminance / color difference reference value storage unit 3.

  According to this embodiment, it is possible to update the reference value to be initialized simply and accurately using four two-dimensional coordinate values. And since it can cope with any VOP size and VP size, the diversion of hardware can be improved, and the initialization target block can be managed by simple calculations such as addition, subtraction, AND, and OR. Hardware cost can be reduced.

  Further, when inverse prediction processing is performed for the next and subsequent macroblocks, processing can be performed without any problem even if the quantization scaler and intra macroblock / intermacroblock type regarding the reference block are unknown. In other words, it is possible to omit the holding of information related to these, thereby saving the storage area and reducing the hardware cost.

(Embodiment 2)
FIG. 7 is a block diagram of the inverse prediction means in Embodiment 2 of the present invention. Constituent elements similar to those of the first embodiment are denoted by the same reference numerals, and redundant description is omitted.

  In the inverse prediction unit 102 in the second embodiment, unlike the first embodiment, the luminance / color difference reference value storage unit 60 has a communication function with the large-capacity memory 61.

  As shown in FIG. 8, the luminance / color difference reference value storage means 60 includes an upper portion 70 that holds the DC component and AC component on the upper side of one macroblock, a corner portion 122 that holds one DC component, and one macro. And a left side portion 123 that holds a DC component and an AC component on the left side of the block. This point is significantly different from the first embodiment.

  Two sets (16 pixels) of the DC component and the AC component are assigned to the upper portion 70 for the four luminance components of each target macroblock, and two sets of the DC component and the AC component (16 pixels) are assigned for the color difference component. Pixel). Therefore, the area required for the upper portion 70 is 16 pixels + 16 pixels, which is a total of 32 pixels.

  As in the first embodiment, the DC component and the AC component held by the luminance / color difference reference value storage unit 60 are values obtained by multiplying the inverse prediction result (QF) by the quantization scale (QP) of the macroblock of interest. is there.

  The upper part 70 in the luminance / color difference reference value storage means 60 of this embodiment is controlled by an external control means such as a processor (not shown), and the stored data is a large-capacity memory space 61 outside the inverse prediction means 102. It has a function to communicate with. This is based on the premise that general image decoding processing for VOP is performed in raster order with macroblocks as a unit.

  At the end of the inverse prediction process for the target macroblock, the reference values stored in the corner 122 and the left part 123 are used for the reverse prediction process for the immediately subsequent target macroblock.

  However, the reference value stored in the upper part 70 is used when the macroblock immediately below the target macroblock is subjected to reverse prediction processing. The immediately following macroblock is processed one line after the current target macroblock.

  In the prior art, a line unit 121 that stores data for one line is provided in order to hold data during this period. However, in this embodiment, since a technique that does not hinder even if data for one line is not stored can be established, only the reference data necessary for the processing of the target macroblock is held in the inverse prediction means 102.

  FIG. 9 is a flowchart showing a prediction process in the second embodiment of the present invention. In FIG. 9, steps 40 and 41 are different from FIG.

  In step 40, the large-capacity memory 61 transfers the upper reference value necessary for the inverse prediction process of the target macroblock to the upper part 70 of the luminance / color difference reference value storage means 60. The transferred data is already stored in the large-capacity memory 61 after the macro block processing immediately above the target macro block.

  Note that if there is no macroblock immediately above the target macroblock, that is, if a macroblock located at the upper end of the VOP is processed, step 40 may be omitted.

  In step 41, the reference value that is stored in the upper part 70 of the luminance / chrominance reference value storage unit 60 and is required when the immediately below macroblock is subjected to reverse prediction processing is stored in the large-capacity memory 61.

  According to the present embodiment, since the conventional technique requires the luminance / color difference reference value storage means for storing data for one line, it can be configured with only the luminance / color difference reference value storage means for one macroblock. It is possible to reduce the storage capacity to be mounted on the prediction calculation means. In addition, since it can always take a constant configuration regardless of the VOP size, the applicability at the time of mounting is high.

  The inverse prediction processing apparatus according to the present invention can be suitably used in a technical field for decoding an image signal encoded by the MPEG-4 standard or a related field.

The block diagram of the reverse prediction apparatus in Embodiment 1 of this invention Flowchart of inverse prediction apparatus in Embodiment 1 of the present invention Explanatory drawing of the positional relationship between the target macroblock and the reference block in Embodiment 1 of the present invention Flowchart of reference value initialization processing in Embodiment 1 of the present invention (A) Explanatory diagram of the next reference value (luminance) update process in the first embodiment of the present invention (b) Explanatory diagram of the next reference value (color difference) update process in the first embodiment of the present invention (c) Explanatory drawing of the next reference value (color difference) update process in the first embodiment Flowchart of inverse prediction apparatus in Embodiment 1 of the present invention The block diagram of the reverse prediction apparatus in Embodiment 2 of this invention Explanatory drawing of the brightness | luminance and color difference reference value storage means in Embodiment 2 of this invention Flowchart of the inverse prediction apparatus in Embodiment 2 of the present invention Explanatory drawing of the two-dimensional coordinate value in Embodiment 1 of this invention Explanatory drawing of the conventional reverse DC prediction process Explanatory drawing of the conventional inverse AC prediction process Block diagram of a conventional image decoding processing device Block diagram of conventional inverse prediction processing means Explanatory drawing of the conventional brightness | luminance reference value storage means Conventional prediction process flowchart

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Decoding information 2 Macroblock information 3, 60, 112, 113 Luminance / color difference reference value storage means 4 Reference value initialization means 5, 116 Inverse prediction calculation means 6, 114, 115 Luminance / color difference prediction value storage means 7 Next reference value Update means 70 Upper part 122 Corner part 123 Left part 61 Large capacity memory 100 Target macroblock setting means 101 Target macroblock extraction means 102 Inverse prediction means 103 Inverse quantization means 104 Inverse DCT means 105 Image memory 111 Predictive control means

Claims (10)

A reverse prediction device that performs reverse prediction processing on data of a macroblock of interest subjected to variable length decoding,
Reference value storage means for holding a reference value necessary for inverse prediction processing of the target macroblock;
Reverse prediction calculation means for performing reverse prediction calculation based on a reference value held by the reference value storage means and generating a reverse prediction result,
The reference value storage means includes an upper part that holds an AC component and a DC component of a macroblock immediately above the target macroblock,
A corner that holds the DC component of the macroblock located diagonally above and to the left of the macroblock of interest;
An inverse prediction apparatus comprising: a left side portion that holds an AC component and a DC component of a previous macro block adjacent to the left side of a target macro block. The AC component and DC component held by the upper part are composed of DC values and AC values of two luminance blocks and two color difference blocks,
The DC component held by the corner is composed of DC values of one luminance block and two color difference blocks,
The inverse prediction apparatus according to claim 1, wherein the AC component and the DC component held by the left side portion include DC values and AC values of two luminance blocks and two color difference blocks. 3. The inverse prediction apparatus according to claim 2, further comprising reference value initialization means for initializing a reference value of a block corresponding to a certain standard among the five luminance blocks and the six color difference blocks. 4. The inverse prediction apparatus according to claim 3, wherein the fixed criterion is defined based on a position of a target macroblock on a video object plane and a position of a video packet to which the target macroblock belongs. 5. The inverse prediction apparatus according to claim 4, wherein the position of the target macroblock and the position of the video packet are evaluated by two-dimensional coordinates on the video object plane. The reference value initialization means is based on the fixed standard.
A first pattern in which values of all of the five luminance blocks and the six color difference blocks are initialized;
Of the five luminance blocks and the six color difference blocks, a second pattern in which the values of seven blocks in contact with the upper side of the target macroblock are initialized;
A third pattern in which values of seven blocks in contact with the left side of the macro block of interest among the five luminance blocks and the six color difference blocks are initialized;
A fourth pattern in which values of three blocks located obliquely above and to the left of the macro block of interest among the five luminance blocks and the six color difference blocks are initialized;
The inverse prediction device according to claim 4, wherein the process branches to a fifth pattern in which the values of any of the five luminance blocks and the six chrominance blocks are not initialized. Reference value initialization means for initializing a reference value held by the reference value storage means further comprises
The reference value initialization means determines whether the target macroblock is an intra macroblock or an inter macroblock,
The reverse prediction apparatus according to claim 1, wherein when the target macroblock is an inter macroblock, at least a part of the reference value held by the reference value storage unit is updated. When the target macroblock is an intra macroblock, the reference value initialization unit multiplies the inverse prediction result generated by the inverse prediction calculation unit by a quantization scale to generate a multiplication result, and the multiplication result is 8. The inverse of claim 7, wherein the reference value held by the reference value storage means is updated so that the reference value of the macroblock immediately adjacent to the target macroblock and the reference value of the macroblock immediately below the target macroblock are the same. Prediction device. When the target macroblock is an inter macroblock, the reference value initializing means uses a reference value having a DC value of 2 ^ (N + 2) (N: the number of bits of a pixel) and an AC value of zero as the target macroblock. 8. The inverse prediction apparatus according to claim 7, wherein the reference value held by the reference value storage unit is updated so that the reference value of the macroblock immediately adjacent to the block becomes the reference value of the macroblock immediately below the target macroblock. Attention macroblock setting means for setting the attention macroblock in the video object plane;
Attention macroblock extraction means for variable-length decoding the data of the set attention macroblock;
Reverse prediction means for performing reverse prediction processing on the data of the macroblock of interest subjected to variable length decoding;
An inverse quantization means for inversely quantizing the data of the target macroblock subjected to the inverse prediction process;
An inverse DCT means for inverse DCT transforming the inversely quantized data of the macro block of interest,
The inverse prediction means includes
Reference value storage means for holding a reference value necessary for inverse prediction processing of the target macroblock;
Reverse prediction calculation means for performing reverse prediction calculation based on a reference value held by the reference value storage means and generating a reverse prediction result,
The reference value storage means includes an upper part that holds an AC component and a DC component of a macroblock immediately above the target macroblock,
A corner that holds the DC component of the macroblock located diagonally above and to the left of the macroblock of interest;
The decoding apparatus which consists of the left side part which hold | maintains the AC component and DC component of the macroblock immediately before adjacent to the left side of a macroblock of interest.

Images