JP2004007788A - Device and method for encoding animation, and device and method for decoding the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce an encoding amount when a quantizing conversion coefficient obtained after orthogonal conversion and quantization is encoded by using a variable length encoding table. <P>SOLUTION: The quantizing conversion coefficient 17 is obtained by performing orthogonal conversion by block unit, and quantization. When the coefficient 17 is encoded, a VLC table group selecting part 4c6 selects a variable length encoding table group from a VLC table storage part 4C1 in accordance with a condition in quantization at every area after dividing a block into the plurality of areas. A VLC table selecting part 4c8 selects the variable length encoding table to be used for encoding the quantizing conversion coefficient in the area from the selected variable length encoding table group. An area data encoding part 4c10 encodes the quantizing conversion coefficient by using the selected variable length encoding table, and outputs it as quantizing conversion coefficient data 17. <P>COPYRIGHT: (C)2004,JPO

Description

The present invention converts and quantizes an input image signal in units of blocks to generate quantized transform coefficients, and codes the quantized transform coefficients by performing variable-length coding using a variable-length coding table for each predetermined area in the block. Video encoding apparatus and a video encoding method for generating an encoded bit stream, a video decoding apparatus for decoding an image signal from the encoded bit stream, and restoring an input image signal before encoding, and The present invention relates to a video decoding method.

Conventionally, as a means for efficiently encoding an image signal, for example, see the video encoding of MPEG-4 (Moving Picture Experts Group Phase-4), which is being standardized in ISO / IEC JTC11 / SC29 / WG111. As adopted in the method (Verification Model 8.0, hereafter VM8.0, ISO / IEC JTC11 / SC29 / WG11 / N1796), discrete cosine transform (DCT) is performed for each block, A method of performing variable length coding of transform coefficients has been used.
In MPEG-4, as means for performing variable-length coding on the quantized transform coefficients in a block, the inside of the block is scanned in a predetermined order, and a flag indicating whether a non-zero coefficient is the last in the block (LAST), (LAST, RUN, LEVEL) is variable-length coded by using the same variable-length coding table in a block by combining the number of consecutive zeros (RUN) and the level of non-zero coefficients (LEVEL) that follow. Is used. In the variable length coding table used at this time, a code word having a predetermined code length is assigned to each (LAST, RUN, LEVEL) based on the occurrence probability of (LAST, RUN, LEVEL).
In other words, a codeword with a long code length is assigned to low occurrence probability (LAST, RUN, LEVEL), whereas a short code length codeword is assigned to high occurrence probability (LAST, RUN, LEVEL). Try to assign words.

However, in the conventional method, since the quantized transform coefficients in a block are encoded by one variable-length encoding table, for example, in encoding, in order to reduce the variation in the values of the quantized transform coefficients, (RUN, LEVEL, etc.), when the transform coefficient prediction for encoding the difference data with the transform coefficients of the preceding and succeeding blocks is performed, and the level of the transform coefficient is smaller than when the transform coefficient prediction is not performed. In the case where the occurrence probabilities are different from each other, there is a problem that if the coding is performed using the same variable length coding table, the redundancy at the time of coding the transform coefficient increases.

Also, since the frequency of occurrence of the transform coefficient level varies depending on the quantization step size at the time of the quantization, encoding using the same variable-length encoding table is also necessary when encoding the transform coefficient. However, there is a problem that the redundancy is increased.

Therefore, the present invention has been made to solve the above-described problem, and it has been proposed to determine whether or not to perform transform coefficient prediction, or to adaptively adjust conditions for various types of encoding such as quantization step sizes. It is an object of the present invention to provide a moving picture coding apparatus and a moving picture coding method capable of improving coding efficiency when switching is performed.

In particular, in the present invention, a block which is a coding unit of an input image signal is divided into several regions, and a variable-length coding table used when performing variable-length coding of transform coefficients in each region is adaptively switched. Accordingly, an object of the present invention is to provide a moving picture coding device and a moving picture coding method capable of improving coding efficiency.

Another object of the present invention is to provide a moving picture decoding apparatus and a moving picture decoding method capable of correctly decoding a coded bit stream generated by the moving picture coding apparatus and the moving picture coding method.

In order to achieve the object, according to the present invention, an input image signal is transformed and quantized in units of blocks to generate quantized transform coefficients, and the quantized transform coefficients are subjected to variable-length coding for each predetermined area in the block. In a moving picture coding apparatus that generates a coded bit stream by performing variable length coding using a table, one of a plurality of variable length coding tables grouped in advance and one of the plurality of variable length coding tables is classified. A variable length coding table group selecting means for selecting a variable length coding table group; and a variable length coding of the quantized transform coefficient using the variable length coding table of the selected group, Bit stream generation means for generating an encoded bit stream including information indicating the variable length coding table of It is an feature.

In the following invention, the plurality of variable-length coding table groups are grouped in advance based on coding conditions at the time of quantization. The variable-length coding table group is selected based on the coding conditions at that time.

In the following invention, the variable-length coding table group includes a plurality of variable-length coding tables grouped based on the presence or absence of a transform coefficient prediction for calculating a difference between quantized transform coefficients between blocks. The long coding table group selecting means selects a corresponding variable length coding table group from the plurality of variable length coding tables based on the presence or absence of the transform coefficient prediction, and the coded bit stream generating means An encoded bit stream including information indicating the presence or absence of the transform coefficient prediction as information indicating a variable length coding table of a selected group is generated.

In the following invention, the variable-length coding table group includes a plurality of variable-length coding tables grouped according to a quantization step size. A corresponding variable-length coding table group is selected from the plurality of variable-length coding table groups based on the size, and the coded bitstream generation unit generates information indicating the variable-length coding table of the selected group. It is characterized in that an encoded bit stream including information on the quantization step size is generated.

Further, in the following invention, the variable length coding table is selected from the variable length coding table group selected by the variable length coding table group selecting means based on the position of the quantized transform coefficient in the area. And a variable-length coding table switching means for switching the table.

Further, in the next invention, when the quantized transform coefficient is subjected to variable-length coding, the case where all the quantized transform coefficients in each region are zero is an invalid region, and the case where there is a non-zero coefficient in the region. Is determined as an effective area. The encoded bit stream generating means, when the area data validity / invalidity determination means determines that the area is an effective area, selects the selected variable-length coding While encoding the quantized transform coefficients in the area by the table and generating an encoded bit stream including information indicating the variable length encoding table used at that time, if it is determined that the area is invalid, the An encoded bit stream including information indicating that the area is invalid is generated.

Further, in the following invention, the encoded bit stream generating means, when performing variable length encoding of the quantized transform coefficient, if the corresponding codeword is not in the variable length encoding table, there is no corresponding codeword. The quantized transform coefficient is converted by a predetermined method so as to be a value included in the variable length coding table, and the quantized transform coefficient after the conversion is variable length coded by the variable length table. And generating an encoded bit stream including the indicated information.

Further, in the next invention, a coded bit stream is input, and coded data subjected to variable length coding for each block included in the coded bit stream is subjected to variable length decoding to restore a quantized transform coefficient. In a moving picture decoding apparatus for decoding an image signal by inversely quantizing and inversely transforming the quantized transform coefficient, a plurality of variable length decoding tables grouped into groups and used on the encoding side from the encoded bit stream. Variable length decoding table group selecting means for decoding information indicating the variable length coding table group, and selecting a variable length decoding table group from the variable length decoding table group based on the variable length coding table group information. Using the variable length decoding table of the selected group to perform variable length decoding of the encoded data to restore the quantized transform coefficients. And 換係 number restoring means, characterized in that it comprises a.

Further, in the following invention, the plurality of variable length decoding table groups are grouped in advance based on encoding conditions at the time of quantization. As information indicating the variable-length coding table group used on the encoding side, information indicating the coding condition at the time of quantization is decoded, and the variable-length decoding table group is decoded based on the coding condition at the time of quantization. It is characterized by selection.

In the following invention, the variable-length decoding table group includes a plurality of variable-length decoding tables grouped based on the presence or absence of a transform coefficient prediction for calculating a difference between quantized transform coefficients between blocks. The long decoding table group selecting means decodes, from the coded bit stream, information indicating the presence or absence of the quantization transform coefficient prediction as information indicating the variable length coding table group used on the coding side, and The variable-length decoding table group is selected based on information indicating the presence or absence of a variable-length decoding table group.

In the following invention, the variable-length decoding table group includes a plurality of variable-length decoding tables grouped based on a quantization step size. Decoding the quantization step size as information indicating the variable length coding table group used on the encoding side, and selecting the variable length decoding table group based on the decoded quantization step size. is there.

Further, in the following invention, the quantization transform coefficient restoring means further switches the plurality of variable length decoding tables in the variable length decoding table group selected by the variable length decoding table group selecting means, and And a variable-length decoding table selecting means for switching the variable-length decoding table based on the position of the quantized transform coefficient in the region when decoding.

Further, in the following invention, when the quantized transform coefficients in the area are all zero, the quantized transform coefficient restoration means indicates that the area is invalid, and one non-zero quantized transform coefficient is included in the area. In the above case, the variable-length coding table selection information indicating the variable-length decoding table used to decode each quantized transform coefficient in the area is decoded, and based on the decoded variable-length coding table selection information, Judging means for judging whether or not the area is invalid; and a quantizing transform coefficient zero for setting all quantized transform coefficients in the area to zero when the judging means judges that the area is invalid. And setting means.

Further, in the following invention, further, a judging means for judging whether or not information indicating that the quantized transform coefficient has been converted by the predetermined method on the encoding side from the encoded bit stream has been decoded, and When it is determined that the information has been decoded by the conversion, when the quantized transform coefficient is restored, decoding is performed using the variable-length decoding table, and then a process reverse to the conversion of the predetermined method is performed. Means for restoring the quantized transform coefficients before the transformation.

Further, in the following invention, an input image signal is transformed and quantized in block units to generate a quantized transform coefficient, and the quantized transform coefficient is calculated using a variable length encoding table determined for each predetermined area in the block. In the moving picture coding method of generating a coded bit stream by performing variable length coding, a variable length coding table group is selected from a plurality of grouped variable length coding tables, and the selected group is selected. The variable-length coding table is used to perform variable-length coding on the quantized transform coefficients to generate a coded bit stream including information indicating the selected variable-length coding table.

Further, in the following invention, when selecting a variable length coding table group from a plurality of grouped variable length coding tables, the selection is performed based on a coding condition at the time of quantization. And generating an encoded bit stream including information on encoding conditions at the time of the quantization as information indicating the variable length encoding table.

Further, in the next invention, a coded bit stream is input, and coded data subjected to variable length coding for each block included in the coded bit stream is subjected to variable length decoding to restore a quantized transform coefficient. In the moving picture decoding method for decoding an image signal by inversely quantizing and inversely transforming the quantized transform coefficient, a variable length used on the encoding side from the encoded bit stream when the quantized transform coefficient is restored. The information indicating the encoding table group is decoded, and a variable length decoding table group is selected based on the variable length encoding table group information from among the plurality of variable length decoding table groups divided into groups. The coded data is variable-length decoded using a variable-length decoding table to restore the quantized transform coefficients.

Further, in the following invention, when restoring the quantized transform coefficient, as information indicating the variable length coding table group used on the encoding side, information on encoding conditions at the time of quantization is used from the encoded bit stream. Decoding is performed, and a variable length decoding table group is selected based on information on coding conditions at the time of quantization.

As described above, in the present invention, a block is divided into a plurality of regions, and for each region, a VLC table group is selected according to the position of the region in the block and the encoding conditions at the time of quantization, Furthermore, since the VLC table can be selected from the selected VLC table group and the transform coefficient can be subjected to variable-length coding, the transform coefficient can be coded with a smaller code amount. In addition, by selecting a VLC table group to be switched in the region based on the position of the region and the encoding conditions at the time of quantization, the amount of code required to encode information indicating the VLC table to be switched is reduced. be able to.

Further, in the next invention, when a VLC table is selected from the selected VLC table group, information indicating the VLC table to be switched is added as additional information by selecting the VLC table according to the scan position in the area. Therefore, there is an effect that the amount of code of additional information necessary for switching the VLC table can be reduced.

Further, in the next invention, it is determined whether or not the quantized transform coefficient in the area is zero for each area, and information that the area is invalid is encoded for an area in which all the coefficients in the area are zero. Conventionally, for a region having a non-zero coefficient in a region, information indicating a VLC table used for coding the coefficient is coded, so that a bit stream structure different depending on a coding mode is conventionally shared. This makes it possible to simplify the circuit for syntax analysis in the decoding device.

Further, in the following invention, when performing variable length coding of AC coefficient data, if data corresponding to the AC coefficient data to be coded is not in the VLC table, the AC coefficient data is converted, and the converted AC coefficient data is converted. By searching the VLC table again based on the VLC and performing variable length coding, an effect of increasing the compression efficiency can be obtained.

Embodiment 1 FIG.
In the first embodiment, a moving picture coding apparatus using the coding apparatus according to claim 1 will be described.
FIG. 1 is a block diagram showing a configuration of the moving picture coding apparatus according to the first embodiment.
In the figure, 1 is input shape data, 2 is shape coding unit, 3 is shape coding data, 4 is variable length coding unit, 5 is multiplexing unit, 6 is local decoding shape data, 7 is input texture data. , 8 is an intra / inter determination unit, 9 is a mode instruction flag, 10 is a motion detection unit, 11 is a motion vector, 12 is a motion compensation unit, 13 is a predicted image signal, 14 is a prediction error signal, and 15 is a coded image. Signal, 16 is a texture encoding unit, 17 is quantized transform coefficient data, 18 is a texture local decoding unit, 19 is a local decoding prediction error signal, 20 is a locally decoded image signal, 21 is a switching unit, 22 is a reference image signal, 23 is a memory, 24 is a shape data codeword, 25 is an additional information codeword, 26 is a block data codeword, 27 is a motion vector codeword, 28 is a coded bit stream, 29 is a subtractor, and 30 is an adder , 31 are transform coefficient prediction determination information which is one of the conditions for texture encoding, and 32 is a quantization step size which is another condition for texture encoding.

FIG. 2 is a diagram showing an internal configuration of the variable length coding unit 4 according to the first embodiment shown in FIG.
In the drawing, reference numeral 4a denotes a shape data variable length coding unit, 4b denotes an additional information variable length coding unit, 4c denotes a transform coefficient variable length coding unit, and 4d denotes a motion vector variable length coding unit.

Next, the operation of the entire apparatus will be described with reference to FIGS.
This apparatus is an apparatus that separates an input digital image sequence into image objects in MPEG-4, that is, partial regions having the same characteristics such as motion and picture, and encodes the image objects in units. However, if one frame is defined as one image object, encoding can be performed in image frame units. The image object is composed of texture data representing a color density level in a circumscribed rectangle surrounding the image object, and shape data representing an effective pixel in the circumscribed rectangle. The texture data is composed of a pair of a luminance signal component and a color difference signal component, and corresponds to an input image signal. The shape data is binary matrix data in which the inside of the object is 1 and the outside of the object is 0.

In operation, first, the input shape data 1 is passed to the shape encoding unit 2, where encoding and local decoding are performed, and output as shape encoded data 3 and local decoded shape data 6, respectively.
The shape-encoded data 3 is subjected to variable-length encoding by a variable-length encoding unit 4 and then multiplexed to an encoded bit stream 28 by a multiplexing unit 5.
The local decoding shape data 6 includes a texture coding unit 16, a texture local decoding unit 18, a motion compensation unit, and the like, which are used to perform texture data coding, motion compensation, and the like for only effective pixels. 12, sent to the motion detection unit 10.

On the other hand, the input texture data 7 is cut out for each block called a macroblock, and the following operation is performed for each macroblock.
That is, first, the intra / inter determination unit 8 determines whether to encode the texture data 7 in the image object or between the image objects. Here, when it is determined to perform intra (within image object) encoding, the intra / inter determining unit 8 passes the input texture data 7 as it is to the texture encoding unit 16 as the encoded image signal 15.

On the other hand, when the encoding is determined to be inter (between image objects), the intra / inter determining unit 8 sends the input texture data 7 to the motion detecting unit 10 in order to perform motion compensation prediction.
The motion detection unit 10 searches for a region most similar to the luminance signal component of the input texture data 7 within a predetermined range of the reference image signal 22 stored in the memory 23, The change in the position up to the position of the similar area is output as a motion vector 11 to the motion compensation unit 12 and the variable length coding unit 4.

After the redundancy is reduced by the motion vector variable length coding unit 4d of the variable length coding unit 4, the motion vector 11 output to the variable length coding unit 4 is converted into a coded bit stream 28 by the multiplexing unit 5. Multiplexed.

{Circle around (2)} The motion compensation unit 12 extracts the predicted image signal 13 from the reference image signal 22 stored in the memory 23 based on the motion vector 11 obtained from the motion detection unit 10. Then, the subtractor 29 passes the prediction error signal 14, which is the difference between the predicted image signal 13 and the input texture data 7, to the texture encoding unit 16 as the encoded image signal 15.

The intra / inter mode selected by the intra / inter determining unit 8 is output to the switching unit 21 and is variable-length coded by the variable-length coding unit 4 as a mode instruction flag 9. The data is passed to the multiplexing unit 5 and multiplexed into an encoded bit stream 28.

Then, the texture encoding unit 16 orthogonally transforms and quantizes the coded image signal 15 and outputs it as quantized transform coefficient data 17. The quantized transform coefficient data 17 is subjected to variable length coding with reduced redundancy by the transform coefficient variable length coding unit 4c of the variable length coding unit 4, and then to the coded bit stream 28 by the multiplexing unit 5. Multiplexed.

The quantized transform coefficient data 17 output from the texture encoding unit 16 is also transmitted to the texture local decoding unit 18, where the quantized transform coefficient data 17 is inversely quantized and inverse orthogonally transformed by the texture local decoding unit 18 to generate a local decoding prediction error signal. It becomes 19.
When the mode instruction flag 9 indicates the inter mode, the local decoding prediction error signal 19 is output by the adder 30 since the prediction image signal 13 from the motion compensation unit 12 is output from the switching unit 21. It is added to the predicted image signal 13 and output as a locally decoded image signal 20.

On the other hand, if the mode instruction flag 9 indicates the intra mode, the switching unit 21 selects 0 signal, and the adding unit 30 selects the local decoding prediction error signal 19 and the prediction image signal 13. Is not performed, and is output as the local decoded image signal 20 as it is.
Note that the locally decoded image signal 20 output from the adding unit 30 is written to the memory 23 because it is used as the reference image signal 22 for motion compensation prediction for the next and subsequent frames.

Next, the configurations and operations of the texture encoding unit 16 and the transform coefficient variable length encoding unit 4c, which are the most important elements of the first embodiment, will be described.

FIG. 3 is an internal configuration diagram of the texture encoding unit 16 according to the first embodiment shown in FIG.
In the figure, 16a is a block data extracting unit, 16b is block data, 16c is a DCT unit, 16d is a DCT coefficient, 16e is a quantizing unit, 16f is a quantized transform coefficient, and 16g is a transform coefficient predicting unit.

Next, the operation in the texture encoding unit 16 will be described.
In the texture encoding unit 16, first, a block serving as a unit for performing a discrete cosine transform (hereinafter, referred to as DCT), which is one of the orthogonal transforms, for example, a block of usually 8 pixels × 8 pixels is obtained from the encoded image signal 15. , And is extracted by the block data extraction unit 16a.

The extracted block data 16b is input to the DCT unit 16c, where DCT is performed. The DCT coefficient 16d output from the DCT unit 16c is quantized by the quantization unit 16e based on the quantization step size, and is output to the transform coefficient prediction unit 16g as a quantized transform coefficient 16f.

The conversion coefficient prediction unit 16g performs a conversion coefficient prediction. In the transform coefficient prediction, a predicted value of a quantized transform coefficient of a coding target block is obtained from a quantized transform coefficient of a surrounding block, and a difference value between the quantized transform coefficient and the predicted value is calculated. This difference value is encoded by the transform coefficient variable length encoding unit 4c (see FIG. 2).

Here, when the quantized transform coefficient of the block to be coded has a correlation with the quantized transform coefficient of the surrounding block, the difference between the quantized transform coefficient of the surrounding block and the quantized transform coefficient of the surrounding block is calculated. The value of the transform coefficient is smaller than that of the original quantized transform coefficient. For example, assume that the values of the quantized transform coefficients are distributed in the range of 0 to 255 with the same appearance frequency. The entropy for such a quantized transform coefficient is 8 bits, and the code length of the code word corresponding to each level value is 8 bits for any level.

In addition, when the quantized transform coefficient of the block to be coded has a strong correlation with the quantized transform coefficient of the neighboring block, and particularly when the quantized transform coefficient of The frequency of appearance of the difference quantized transform coefficient, which is the difference value from the quantized transform coefficient, increases near zero. Since the entropy of a signal having such a biased appearance frequency is smaller than the entropy of the original quantized transform coefficient, if the code length is assigned according to the appearance frequency, the average code of the differential quantized transform coefficient is obtained. The length can be smaller than 8 bits. In this way, by performing the transform coefficient prediction and encoding the differentially quantized transform coefficients, it is possible to reduce the amount of code generated when the transform coefficient variable-length encoding unit encodes.

Next, an example of the operation of the transform coefficient prediction unit 16g will be described.
FIG. 4 is an explanatory diagram of transform coefficient prediction by the transform coefficient predictor 16g.
The operation described below corresponds to intra AC / DC prediction disclosed in ISO / IEC JTC1 / SC29 / WG11 MPEG-4 Video VM8.0.
The transform coefficient prediction is performed particularly on the DC coefficient of the intra mode in which the variation of the value becomes large, or the quantized transform coefficient (hereinafter, abbreviated as a transform coefficient) of the low frequency component located in the hatched portion in FIG.

Assuming that the block to be coded is X as shown in FIG. 4, the predicted value of the transform coefficient is the transform coefficient of a block around block X, for example, block B or block C. A difference between the calculated transform coefficient prediction value and the quantized transform coefficient is obtained, and a difference quantized transform coefficient is output.
However, when there is no correlation with the surrounding blocks and the difference quantized transform coefficient becomes larger than the original quantized transform coefficient, the original quantized transform coefficient is output without performing the transform coefficient prediction.

The transform coefficient predictor 16g outputs the difference quantized transform coefficient or the quantized transform coefficient obtained as described above as the quantized transform coefficient data 17. The quantized transform coefficient data 17 is variable-length coded by the transform coefficient variable-length coding unit 4c.
The conversion coefficient prediction unit 16g outputs information indicating whether or not the conversion coefficient prediction has been performed as the conversion coefficient prediction determination information 31.

Next, the detailed configuration and operation of the transform coefficient variable length coding unit 4c will be described.
FIG. 5 is an internal configuration diagram of the transform coefficient variable length coding unit 4c shown in FIG. 2, and FIG. 6 is a flowchart showing the operation thereof.
In FIG. 5, 4c1 is a VLC table storage unit, 4c2 is a switching unit, 4c3 is a DC coefficient data encoding unit, 4c4 is AC coefficient data, 4c5 is a DC coefficient codeword, 4c6 is a VLC table group selection unit, and 4c7 is a VLC table. Group information, 4c8 is a VLC table selection unit, 4c9 is VLC table selection information, 4c10 is an area data encoding unit, and 4c11 is an AC coefficient codeword.

The transform coefficient variable length coding unit 4c according to the first embodiment configured as described above divides the block into a plurality of regions, and for each region, based on the statistical characteristics of the transform coefficients in the region. Set a plurality of variable length coding tables. When coding the transform coefficients in the area, the variable length coding table to be used is adaptively switched from among the plurality of set variable length coding tables. When a plurality of variable length coding tables are set for each area, the variable length coding tables are set based on whether or not the transform coefficient prediction has been performed on the position of the area and the transform coefficient in the area.

Hereinafter, the operation will be described in detail with reference to FIGS.
(1) DC Coefficient Data Encoding First, the switching unit 4c2 determines the mode instruction flag 9 (step S1). When the mode is the intra mode, the quantized transform coefficient data 17 is converted to the DC coefficient data encoding unit 4c3. And encodes the DC coefficient data (step S2), and outputs a DC coefficient codeword 4c5. On the other hand, in the case of the inter mode, since there is no DC coefficient, switching to the DC coefficient data encoding unit 4c3 is not performed.

(2) Encoding of AC coefficient data Next, encoding of the AC coefficient data 4c4 of the quantized transform coefficient data 17 will be described. The coding of the AC coefficient data 4c4 is performed for each area by dividing the block into a plurality of areas. When DCT is performed for each block, the probability of occurrence of the value of the transform coefficient differs depending on the position in the block. Therefore, the inside of the block is divided into a plurality of regions based on the statistical characteristics of the image data.

{For example, as shown in FIG. 7, the AC coefficient area in the block is divided into four areas 1 to 4. The area 1 is an area including a low-frequency component in a vertical direction parallel to the scanning direction, and the area 2 is an area including a low-frequency component in a horizontal direction perpendicular to the scanning direction. For this reason, a high value AC coefficient is likely to occur in the region 1 and the region 2. On the other hand, the region 3 or the region 4 is a region containing a high frequency component, and an AC coefficient of zero or a small value is easily generated.

The encoding of the AC coefficient data in the regions 1 to 4 is performed by scanning the AC coefficient data in each of the regions 1 to 4 in a predetermined order from the low frequency component to the high frequency component, and obtaining a plurality of obtained (RUN, LEVEL). The data is subjected to variable length coding (hereinafter abbreviated as VLC (Variable Length Coding)).

(4) Then, the areas 1 to 4 are first scanned in the order determined for each of the areas 1 to 4, and (RUN, LEVEL) data is obtained (step S3). Here, RUN is the number of consecutive zeros when scanning, and LEVEL is the absolute value of the following non-zero coefficient.

For example, when the area 2 is scanned in the direction of arrow A shown in FIG. 7, the first coefficient scanned is -8, so the first (RUN, LEVEL) is (0, -8). Become. Next, since -1 continues after one 0 continues, the next (RUN, LEVEL) data is (1, -1). The above processing is repeated until a non-zero coefficient in the area reaches the end.

The (RUN, LEVEL) data obtained by scanning is subjected to variable length coding by a VLC table set based on the occurrence probability of (RUN, LEVEL) data in each area.
The VLC table assigns a short code length to a symbol having a high probability of occurrence and a long code length to a symbol having a low probability of occurrence according to the probability of occurrence of the symbol, and pairs the symbol with the corresponding code word. It is a table.
The symbol of the VLC table used when encoding the AC coefficient data corresponds to (RUN, LEVEL) data. Since the probability of occurrence of (RUN, LEVEL) differs for each region, by setting a different VLC table for each region, the code amount of the AC coefficient data can be reduced.

The value of the AC coefficient also differs depending on various coding conditions when DCT and quantization are performed. For example, when transform coefficient prediction that takes a difference from the AC coefficient of a surrounding macroblock is performed, the value of the AC coefficient is smaller than when the transform coefficient prediction is not performed. Further, it differs depending on the size of the quantization step size at the time of quantization.
Therefore, by setting the VLC table in consideration of these conditions, the code amount of the AC coefficient data can be further reduced.

From the above viewpoint, in the first embodiment, a VLC table group is selected for each of the areas 1 to 4 according to the statistical characteristics of the (RUN, LEVEL) data in the areas 1 to 4, and the selected VLC table group is selected. The VLC table used for encoding each (RUN, LEVEL) data is switched in the table group.

Hereinafter, a method of selecting a VLC table for each of the regions 1 to 4 will be described.
(2.1) VLC table group selection Since the appearance probabilities of (RUN, LEVEL) differ depending on the region 1 to 4 or the scanned position in the region 1 to 4, a plurality of VLCs different for the regions 1 to 4 are set. Set up the table.
Also, the appearance probability differs depending on the encoding conditions when the quantized transform coefficient data is obtained.
Therefore, in the first embodiment, the VLC table is used as a condition for texture encoding based on transform coefficient prediction determination information 31 (see FIG. 3) obtained from transform coefficient predictor 16g of texture encoder 16. Select a group.

First, a VLC table group, which is a combination of VLC tables that can be selected for encoding each region in a block, is selected by the VLC table group selection unit 4c6 (step S4).
For example, in the case of division as shown in FIG. 7, a VLC table group is selected for each of the areas 1 to 4. At this time, when scanning from the low frequency component to the high frequency component, the selection is made based on the characteristic that the higher the frequency component, the higher the appearance probability of the data of the higher LEVEL, and the higher the frequency component, the more the LEVEL or zero increases.
Here, a method of selecting a VLC table group based on the transform coefficient prediction determination information 31 for each of the regions 1 to 4 will be described with reference to FIG.

FIG. 8 shows a VLC table group in which a set of (RUN, LEVEL) stored in the VLC table storage unit 4C1 and a code word to which a code length according to the occurrence probability is assigned.
Since the appearance probability of (RUN, LEVEL) differs depending on whether or not the transform coefficient prediction has been performed, or the position where the transform coefficient has been scanned in the regions 1 to 4, as shown in FIG. A VLC table group having a plurality of VLC tables corresponding to each case is prepared for each of the areas 1 to 4 as to whether or not the processing has been performed.

For example, as illustrated in FIG. 8, when the conversion coefficient prediction is not performed on the area 1 and the area 2, the VLC table group selection unit 4c6 sets the conversion coefficient prediction determination information 31 and (RUN, LEVEL) And a table group T1 having a VLC table 1A including a symbol of a small LEVEL, a VLC table 2A, 3A including a symbol of a short RUN and a symbol of a short RUN, and selecting the selected table group T1. The indicated VLC table group information 4c7 is output to the VLC table selection unit 4c8.

On the other hand, when the conversion coefficient prediction is performed, the effect of reducing the variation of the coefficient level is obtained. Therefore, compared with the case where the conversion coefficient prediction is not performed, the VLC table including the higher LEVEL is required. The degree is small.
Therefore, the VLC table group selection unit 4c6, based on the transform coefficient prediction determination information 31 and the appearance probability of (RUN, LEVEL) when the transform coefficient prediction is performed, generates a VLC table 1B including a symbol of a small LEVEL, In the LEVEL, a table group T2 having a VLC table 2B or the like including a short RUN symbol is selected, and VLC table group information 4c7 indicating the selected table group T2 is output to the VLC table selection unit 4c8.

In addition, since the appearance probability of the high LEVEL symbol is small for the region 3 and the region 4, the table group T3 having the VLC tables 4 and 5 including the long RUN symbol is selected instead, and the selected table is selected. The VLC table group information 4c7 indicating the group T3 is output to the VLC table selection unit 4c8.

It should be noted that the VLC table group, on the decoding side corresponding to the first embodiment, corresponds to the transform coefficient prediction determination information 31 for each of the regions 1 to 4 in the block, similarly to the encoding side of the first embodiment. Therefore, it is not necessary to multiplex the VLC table group information 4c7 as additional information into the bit stream 28.
In other words, as shown in FIGS. 1 and 2, the transform coefficient prediction determination information 31 includes the additional information codeword 25 in the bit stream 28 transmitted to the decoding side in order to determine whether to perform the transform coefficient prediction. Since it is multiplexed as one, this may be used.
Therefore, since it is not necessary to multiplex the VLC table group information 4c7 on the decoding side into the bit stream 28, the VLC table group used for each area can be selected without newly increasing the code amount. .

(2.2) Region Data Coding Next, the variable length coding of the AC coefficient data 4c4 in each of the regions 1 to 4 will be described.
The AC coefficient data 4c4 is represented by a combination of a plurality of (RUN, LEVEL) data obtained by scanning each of the areas 1 to 4 in a predetermined order. Certain AC coefficient data 4c4 is output to the VLC table selection unit 4c8 and the area data encoding unit 4c10.

The VLC table selection unit 4c8 uses the VLC table used for variable-length coding of the AC coefficient data 4c4, which is the input (RUN, LEVEL) data, based on the VLC table group information 4c7 from the VLC table group selection unit 4c6. Is adaptively selected from the VLC table groups selected by the VLC table group selection unit 4c6 (step S5). VLC table selection information 4c9 indicating the selected VLC table is output to region data encoding section 4c10.
The criterion for selecting the VLC table is determined, for example, by the magnitude of the LEVEL of the AC coefficient data 4c4, which is (RUN, LEVEL) data.

Next, the operation of the area data encoding unit 4c10 will be described.
The region data encoding unit 4c10 selects a VLC table stored in the VLC table storage unit 4c1 based on the VLC table selection information 4c9 output from the VLC table selection unit 4c8, and changes (RUN, LEVEL) data. Long coding is performed (step S6).

In variable length coding, for example, first, a codeword corresponding to (RUN, LEVEL) is searched from the selected VLC table. If a corresponding codeword is found, it is output as an AC coefficient codeword 4c11. If a corresponding codeword is not found, a fixed-length codeword called an escape code is output. When the escape code is output, RUN and LEVEL are each fixed-length encoded, and the escape code followed by the fixed-length code words of RUN and LEVEL, respectively, is output as an AC coefficient codeword 4c11.

Steps S5 to S6 are repeated until all the (RUN, LEVEL) data in the area has been encoded in this way ("No" in step S7).

Further, when all the (RUN, LEVEL) data in a certain area is variable-length coded ("Yes" in step S7), all the (RUN, LEVEL) data in the other areas 1 to 4 in the block are variable. It is determined whether the data has been long-coded (step S8), and if it is determined that the variable-length coding has not been performed on all the (RUN, LEVEL) data in the other areas 1 to 4 in the block (step S8 "No"). ), The processing from step S3 to step S7 is repeated (step S8).

On the other hand, if it is determined that all the (RUN, LEVEL) data in each of the areas 1 to 4 in the block have been subjected to variable length coding (“Yes” in step S8), the transform coefficient variable length coding unit 4c sets After the variable-length coding process is completed, the DC coefficient codeword 4c5 and the AC coefficient codeword 4c11 are passed to the multiplexing unit 5 as the block data codeword 26, and the multiplexing unit 5 multiplexes the coded bitstream 28. I do.

Here, in the first embodiment, the selection of the VLC table from the selected VLC table group is performed using the LEVEL value of each (RUN, LEVEL) data. On the side, it is necessary to select the VLC table before decoding the LEVEL. For this reason, the VLC table selection information 4c9 is sent to the additional information variable length coding unit 4b to be coded to be a VLC table selection codeword, and is multiplexed as one of the additional information codewords 25 into the bit stream 28. .

Here, as a method of encoding the VLC table selection information 4c9, the VLC table selection information 4c9 (NEXT) used to encode the current (RUN, LEVEL) data and the next (RUN, LEVEL) data. Is a variable length coding method. The variable length coding table used at this time is a table in which a set of (NEXT, RUN, LEVEL) stores a code word whose code length is determined according to the appearance probability.

In this method, the VLC table selection information 4c9 used to encode the first (RUN, LEVEL) in the area needs to be separately encoded as VLC table initial switching information. The VLC table selection information 4c9 used for encoding the second and subsequent (RUN, LEVEL) is encoded as a set with the immediately preceding (RUN, LEVEL) data. When the LEVEL that is not zero is the last in the area, NEXT is information indicating that the (RUN, LEVEL) paired with the LEVEL is the last (RUN, LEVEL) in the area.

切 り 替 え When switching a plurality of VLC tables to encode each area, it is necessary to encode the VLC table selection information 4c9 as additional information as described above. For example, as shown in FIG. 8, when switching the VLC table group to be used between when the conversion coefficient prediction is performed and when the conversion coefficient prediction is not performed, the number of VLC tables to be switched when the conversion coefficient prediction is performed is as follows. Since there are two tables 1B and 2B of T2, the code length required when the VLC table selection information 4c9 is coded with a fixed length is 1 bit.

However, if the VLC table group is not switched between when the transform coefficient prediction is performed and when it is not performed, it is necessary to switch from the five types of VLC tables, and the VLC table selection information 4c9 is encoded with a fixed length. In this case, the required code length becomes 3 bits.
As described above, when the VLC table selection information 4c9 is combined with (RUN, LEVEL) and subjected to variable length encoding, if the number of NEXTs increases, the number of combinations of NEXT, RUN, and LEVEL increases. NEXT, RUN, LEVEL) also increases the code amount of the code word corresponding to the code word. As described in the first embodiment, a VLC table group is selected for each area according to the statistical characteristics of the (RUN, LEVEL) data in the area, and among the selected VLC table groups, By switching the VLC table used when encoding (RUN, LEVEL) data, the amount of code required to encode the VLC table selection information 4c9 can be reduced.

As described above, in the first embodiment, a block is divided into a plurality of regions, and a VLC table group is selected for each region in accordance with the position of each region in the block and the transform coefficient prediction determination information. Further, since the VLC table is selected from the selected VLC table group and the transform coefficient is subjected to variable-length coding, the transform coefficient can be coded with a smaller code amount.

In the first embodiment, the type of the VLC table to be switched is encoded by preparing a VLC table group to be switched in each region in advance based on the position of each region and the transform coefficient prediction determination information 31. Can reduce the amount of code required.

Embodiment 2 FIG.
The second embodiment is characterized in that the configuration of the transform coefficient variable length coding unit 4c is changed. Therefore, the configuration and operation of the transform coefficient variable length coding unit 4c according to the second embodiment will be mainly described.

FIG. 9 is an internal configuration diagram of the transform coefficient variable length coding unit 4c according to the second embodiment. In the figure, reference numeral 32 denotes a quantization step size at the time of quantization output from the quantization unit 16e of the texture encoding unit 16 shown in FIG.

That is, in the transform coefficient variable length coding unit 4c according to the second embodiment, instead of the transform coefficient prediction determination information 31 input to the VLC table group selection unit 4c6 as a condition for texture coding, a quantization step size 32 Is input, and other operations are the same as those of the transform coefficient variable length coding unit 4c in the first embodiment.

Next, the operation of the VLC table group selection unit 4c6 according to the second embodiment will be described.

The VLC table group selection unit 4c6 is based on the position in the block of the region recognized in advance by the VLC table group selection unit 4c6, and the quantization step size 32 input as a condition for texture encoding from outside. , Select a VLC table group that can be selected to encode each region in the block.

{For example, in a region quantized by the large quantization step size 32, the number of small LEVELs or zeros increases, and the appearance probability of data of a large LEVEL decreases. Since the appearance probability of (RUN, LEVEL) data differs according to the quantization step size 32, the VLC table group selection unit 4c6 selects the VLC table group according to the quantization step size 32.

As shown in FIG. 1 and FIG. 2, the quantization step size 32 is a variable length encoded by the additional information variable length encoding unit 4 b of the variable length encoding unit 4 as information when performing inverse quantization on the decoding side. After being coded, it is multiplexed into the bit stream 28 by the multiplexing unit 5 as one of the additional information codes 25, so that the decoding side uses this to code the variable length decoding table group used for decoding. It can be selected in the same way as the chemical side.

Therefore, in the second embodiment, it is not necessary to multiplex new additional information by selecting a VLC table group into the bit stream 28 as in the first embodiment. However, as in the case of the first embodiment, when the selection of the VLC table from the selected VLC table group is performed using the LEVEL value of each (RUN, LEVEL) data, the decoding side , LEVEL, before decoding, the VLC table selection information 4c9 is sent to the additional information variable length coding unit 4b to be coded to be a VLC table selection codeword, and the additional information codeword 25 is selected. Is multiplexed into the bit stream 28.

Therefore, according to the second embodiment, a block is divided into a plurality of regions, and a VLC table group is selected for each region according to the position of the region in the block and the quantization step size. Since the VLC table is selected from the selected VLC table group and the AC coefficient data in each area is subjected to variable-length encoding, the transform coefficient can be encoded with a smaller code amount. The effect of increasing the compression efficiency is obtained.

Further, in the second embodiment, when the VLC table group is selected, the quantization step size 32 is used. However, both the transform coefficient prediction determination information 31 and the quantization step size 32 described in the first embodiment are used. Thus, the VLC table group may be selected.

Embodiment 3 FIG.
In the third embodiment, the selection of the VLC table used to encode each (RUN, LEVEL) data in the area from the VLC table group selected by the VLC table group selection unit is performed in the area. It is characterized in that it is performed based on the statistical characteristics of the LEVEL value such that the appearance probability of a LEVEL having a value of zero or a smaller value increases as going from a low frequency component to a high frequency component.

For example, in a low-frequency component, a large value level is likely to occur. Therefore, when (RUN, LEVEL) data to be encoded is data scanned with a low-frequency component, a VLC table including a large value LEVEL is generated. select.

On the other hand, in the case of data scanned with a high frequency component, a VLC table including a small value level and long RUN data and assigning a short code length to them is selected.
Hereinafter, a description will be given of the transform coefficient variable length coding unit 4c including the VLC table selection unit that selects the VLC table based on the scanned position in the area.

FIG. 10 shows the internal configuration of the transform coefficient variable length coding unit 4c according to the third embodiment.

The operations of the switching unit 4c2, the DC coefficient data encoding unit 4c3, and the VLC table group selection unit 4c6 are the same as those in the first or second embodiment, and differ only in the operation of the VLC table selection unit 4c8. is there.

That is, the VLC table group selection unit 4c6, like in the first and second embodiments, sets the coding conditions for texture coding such as the quantization step size 32 or the transform coefficient prediction determination information 31 for each region in the block. And selects a selectable VLC table group based on the VLC table group, and outputs the selected VLC table group to the VLC table selection unit 4c8 as VLC table group selection information 4c7.

The VLC table selection unit 4c8 selects a VLC table used to encode the AC coefficient data 4c4, which is the (RUN, LEVEL) data, from the VLC table groups selected by the VLC table group selection unit 4c6.

Here, the appearance probability of (RUN, LEVEL) differs depending on the scanned position in the area. Even in the same region, the level of zero or a small value increases as going from the low frequency component to the high frequency component. That is, the appearance probability of the LEVEL value differs depending on the position where the LEVEL is scanned.

Therefore, unlike the first and second embodiments, the VLC table selection unit 4c8 according to the third embodiment scans an area without inputting (RUN, LEVEL) data as the AC coefficient data 4c4. The selection of the VLC table is performed according to the position of. That is, each of the regions 1 to 4 is further divided into a plurality of regions, and the VLC table is selected according to which region the (RUN, LEVEL) data to be encoded has been scanned.

Since the (RUN, LEVEL) data is encoded in the scanning order in a predetermined order in each of the areas 1 to 4, the decoding side determines the scan position of the (RUN, LEVEL) to be decoded on the decoding side. The scan position on the encoding side can be matched based on the already decoded (RUN, LEVEL). For this reason, the decoding side can select the variable length decoding table to be used at the time of decoding based on the scan position. Therefore, the encoding side of the third embodiment uses the VLC table selection information 4c9 as additional information in the bit stream. There is no need for multiplexing.

As a result, in the first and second embodiments, it was necessary to multiplex the VLC table selection information 4c9 into the coded bit stream 28, but in the third embodiment, the VLC table selection information 4c9 was added to the bit stream 28. Since there is no need for multiplexing, the code amount of the coded bit stream 28 can be reduced as compared with the first and second embodiments.

Then, after the VLC table selection information 4c9 is output from the VLC table selection unit 4c8 to the region data encoding unit 4c10, the region data encoding unit 4c10 is switched to the VLC table selection unit 4c8 as in the first and second embodiments. The VLC table is switched and selected on the basis of the VLC table selection information 4c9 output from, and (RUN, LEVEL) data, which is the AC coefficient data 4c4, is subjected to variable-length encoding.
The operation of the area data encoding unit 4c10 is the same as in the first and second embodiments, and the above operation is repeated until all (RUN, LEVEL) data in the area is encoded.

Therefore, in the third embodiment, the VLC table is selected according to the scan position in the area. Therefore, it is not necessary to multiplex the VLC table selection information 4c9 as additional information into the bit stream, and the The effect is obtained that the code amount of the additional information required for switching can be reduced.

Embodiment 4 FIG.
The fourth embodiment is characterized in that the configuration of the transform coefficient variable length coding unit 4c is changed.

FIG. 11 shows an internal configuration diagram of transform coefficient variable length coding section 4c according to the fourth embodiment.
In the figure, reference numeral 4c12 denotes an area data valid / invalid determination unit, 4c13 denotes area valid / invalid information, 4c14 denotes an area valid / invalid information coding unit, and 4c15 denotes an area valid / invalid information codeword.

Next, the operation will be described. In FIG. 11, the operations differ from each other in an area data valid / invalid determining unit 4c12 and an area valid / invalid information encoding unit 4c14. Or, it is completely the same as the second embodiment.
Therefore, the operation of the area data valid / invalid determination unit 4c12 and the corresponding operation of the area valid / invalid information encoding unit 4c14 will be described below.

First, in the transform coefficient variable length coding unit 4c according to the fourth embodiment, the area data valid / invalid determination unit 4c12 determines whether or not all the AC coefficients in the area are zero for each area. The determination result is output as area valid / invalid information 4c13. The region valid / invalid information encoding unit 4c14 encodes the region valid / invalid information 4c13 and outputs it as a region valid / invalid information codeword 4c15. As an encoding method of the area valid / invalid information 4c13, for example, variable-length encoding may be performed on the area valid / invalid information of all the areas in the block.

If the AC coefficient data 4c4 in the area is all zero, the area is regarded as an invalid area and the encoding of the next area is skipped without encoding the AC coefficient data 4c4 in the area. . Therefore, when the AC coefficient data 4c4 in the area is all zero, only the area valid / invalid information 4c13 needs to be encoded, and the code amount of the AC coefficient data 4c4 can be reduced.

In the case of the transform coefficient variable length coding unit 4c described in Embodiments 1 and 2, VLC table selection information 4c9 indicating a VLC table used for coding each (RUN, LEVEL) data is added to the additional information variable. The encoding is performed by the long encoding unit 4b.

Therefore, in the fourth embodiment, the region valid / invalid information 4c13 is used to select the first VLC table indicating the VLC table used to encode the first (RUN, LEVEL) data in each of the regions 1 to 4. The information may be encoded into the information 4c9 or the like.
For example, the additional information variable length coding unit 4b sets the first VLC table selection information 4c9 to "0" when the AC coefficients in the area are all zero and the area is not to be coded, and n types of If the information indicating the VLC table when the VLC table is switched and encoded is “1” to “n”, the area valid / invalid information 4c13 can be substituted by the VLC table selection information 4c9. By doing so, the area valid / invalid information encoding unit 4c14 itself becomes unnecessary.

Next, the variable-length coding unit 4c described in the fourth embodiment shown in FIG. 4 has an advantageous effect as compared with the related art obtained when the variable-length coding unit 4c is applied to coding of macroblock data in MPEG-4, namely, In the related art, the effect that the structure of different bit streams can be made common depending on the encoding mode and the circuit for syntax analysis in the decoding device can be simplified will be described.

FIG. 12 shows a structure of an encoded bit stream when encoding macroblock data in the conventional MPEG-4.
Conventionally, in MPEG-4, different bit streams are used for each VOP when performing intra / inter coding (I-VOP / P-VOP) and performing bidirectional prediction coding (B-VOP). It has the structure of Intra coding is a mode in which coding is performed using only information in a VOP. The inter coding is a mode in which motion compensation prediction is performed using the immediately preceding VOP. In addition, bidirectional prediction encoding is a mode in which motion compensation prediction is performed using one or more preceding and subsequent VOPs.

FIG. 12A shows the structure of a bit stream B1 in the case of a conventional I-VOP / P-VOP.
In the figure, shape attribute information D1 is shape attribute information such as a motion vector of shape data.
The valid macroblock information D2 is information indicating the validity / invalidity of the macroblock. If the transform coefficient data of the macroblock is all zero and the motion vector is zero, the macroblock becomes an invalid macroblock, and the subsequent information Are not multiplexed into the bitstream.
The effective chrominance block information / macroblock type D3 is information in which the effective chrominance block information and the macroblock type are combined, and the effective chrominance block information indicates the validity / invalidity of the chrominance block in the macroblock. When the conversion coefficient data in the chrominance block is all zero, the chrominance block is invalid.
The AC prediction instruction information D4 is information indicating whether or not transform coefficient prediction has been performed on the AC coefficient in the macroblock, and is the same as the transform coefficient prediction determination information 31.
The effective luminance block information D5 indicates whether the luminance block in the macroblock is valid or invalid. When all the conversion coefficient data in the luminance block is zero, the luminance block becomes invalid.
The difference quantization step size D6 indicates a difference from the quantization step size of the immediately preceding macroblock data when the quantization step size is different from the immediately preceding macroblock data.
The interlace information D7 is information necessary for an interlaced image.
The motion data D8 is data relating to motion existing when motion compensation is performed.
The shape data D9 is information necessary for encoding a VOP having an arbitrary shape.
The block data D10 exists only when the effective color difference block information or the effective luminance block information D5 indicates that the block is an effective block.

FIG. 12B shows the structure of a bit stream B2 in a conventional B-VOP. Only the multiplexing method of the macroblock type D10 and the effective block information D12 is different.
Specifically, in the case of the I-VOP / P-VOP shown in FIG. 12A, the chrominance block and the luminance block are separately multiplexed for the effective block information, and the effective information of the chrominance block is , A macroblock type, and variable-length encoding as effective color difference block information / macroblock type D3. The luminance block information is obtained by combining valid / invalid information of four luminance blocks and performing variable length coding as effective luminance block information D5.

On the other hand, in the case of the B-VOP shown in FIG. 12B, the macroblock type D11 and the effective chrominance block information are separately multiplexed, and information indicating the valid / invalid of each of the luminance block and the chrominance block is provided. It is fixed-length coded as effective block information D12.

As described above, in the conventional I-VOP / P-VOP and B-VOP, a method of multiplexing effective block information indicating whether or not all quantized transform coefficients (AC coefficients) in a block are zero in a region And the structure of the bit stream is different.

However, according to the variable length coding unit 4c described in the fourth embodiment, for each region in the block, the region data validity / invalidity determination unit 4c12 determines whether the AC coefficient of the quantized transform coefficient is zero. Then, the region valid / invalid information encoding unit 4c14 encodes the region valid / invalid information 4c13 and multiplexes it as the region valid / invalid information encoding unit 4c14 on the bit stream 28. It is not necessary to determine whether the quantized transform coefficient AC coefficient is zero or not and to multiplex it as effective block information into the bit stream 28.

FIG. 13 shows a bit stream B3 obtained when the variable length coding unit 4c of the fourth embodiment is applied to coding of macroblock data in MPEG-4.
Since the effective block information is not multiplexed into the bit stream D3, the common structure of the bit stream B3 shown in FIG. 13 is obtained in any of the I-VOP, P-VOP, and B-VOP coding modes.
In the fourth embodiment, the area valid / invalid information D14 is set for each AC coefficient data D13 of each area of the block data D10.

As described above, when the variable length encoding means described in Embodiment 4 is applied to encoding of macroblock data in MPEG-4, regardless of the encoding mode of intra, inter, and bidirectional prediction, The structure of the common bit stream D3 is as shown in FIG.

Therefore, according to the moving picture coding apparatus of the fourth embodiment, the decoding side can use a common means for analyzing the syntax, which is the structure of the bit stream, regardless of the coding mode. This has the effect of simplifying the circuit configuration for tax analysis.

Embodiment 5 FIG.
The fifth embodiment is characterized by the configuration of the area data encoding unit 4c10 of the transform coefficient variable length encoding unit 4c.

FIG. 14 is a detailed internal configuration diagram of the area data encoding unit 4c10 of the transform coefficient variable length encoding unit 4c in the present embodiment. In the figure, reference numeral 4c101 denotes a VLC table search unit, 4c102 denotes an escape code, 4c103 denotes a conversion coefficient conversion unit, 4c104 denotes a conversion AC coefficient data, 4c105 denotes a conversion flag, and 4c106 denotes a conversion coefficient fixed length decoding unit.

Next, the operation will be described. The VLC table search unit 4c101 searches the VLC table of the VLC table storage unit 4c1 for a codeword corresponding to the AC coefficient data (RUN, LEVEL) 4c4 that is the input conversion coefficient data. When a codeword corresponding to the AC coefficient data (RUN, LEVEL) 4c4 is found in the VLC table, the codeword is output as an AC coefficient codeword 4c11 as in the first to fourth embodiments.

Here, the operation in the case where the information is not found in the VLC table, which is a feature of the fifth embodiment, will be described in detail.
First, when a codeword corresponding to the input AC coefficient data 4c4 is not present in the VLC table, the VLC table search unit 4c101 converts a fixed-length codeword called an escape code 4c102 and the AC coefficient data 4c4 into a conversion coefficient conversion unit. 4c103.

The conversion coefficient conversion unit 4c103 converts the RUN or LEVEL value of the AC coefficient data 4c4.
The first conversion is performed on LEVEL among the values of RUN or LEVEL of the AC coefficient data 4c4. The converted LEVEL * is obtained by subtracting the maximum value (LMAX) of the LEVEL of the AC coefficient data included in the VLC table. Specifically, it is shown in the following equation 1.

LEVEL * = sign (LEVEL) × [abs (LEVEL−LMAX)] (Equation 1)
However, sign (LEVEL) is 1 if LEVEL is 0 or more, and -1 otherwise. Abs (LEVEL) is the absolute value of LEVEL.

Then, the VLC table search unit 4c101 searches the VLC table again using the converted AC coefficient data 4c104 (RUN, LEVEL *). Here, if a codeword corresponding to the converted AC coefficient data 4c104 (RUN, LEVEL *) is found in the VLC table, a code indicating that the value of LEVEL has been converted after the escape code and a VLC table Attached with the detected code word, it is output as AC coefficient code word 4c11.

On the other hand, if the codeword corresponding to the converted AC coefficient data 4c104 (RUN, LEVEL *) on which the first conversion has been performed is not found in the VLC table, the conversion coefficient conversion unit 4c103 performs the second conversion. .
The second conversion is performed on RUN. The converted RUN * is obtained by subtracting the RUN maximum value (RMAX) +1 of the AC coefficient data contained in the VLC table. Specifically, it is shown in the following equation 2.

{RUN * = RUN- (RMAX + 1)} ... (Equation 2)

{Circle around (4)} Then, the VLC table search unit 4c101 searches the VLC table again using the new converted AC coefficient data 4c105 (RUN *, LEVEL *). Here, when a code word corresponding to the new converted AC coefficient data 4c105 (RUN *, LEVEL *) is found in the VLC table, a code indicating that the value of RUN has been converted is added after the escape code. , And a codeword detected from the VLC table, and output as an AC coefficient codeword 4c11.

On the other hand, if the codeword corresponding to the converted AC coefficient data 4c105 (RUN *, LEVEL *) subjected to the second conversion is not found in the VLC table, RUN and LEVEL are respectively fixed-length coded, After the escape code, a code indicating that the AC coefficient data has been fixed-length coded and a fixed-length codeword of the AC coefficient data are added, and output as an AC coefficient codeword 4c11.

As described in the first embodiment, (RUN, LEVEL) and VLC table selection information 4c9 (NEXT) used for encoding the next (RUN, LEVEL) are set as a set to perform variable-length encoding. The same processing can be performed in the case of performing.

Therefore, according to the fifth embodiment, when performing variable length coding of AC coefficient data, if there is no data corresponding to the AC coefficient data to be coded in the VLC table, the AC coefficient data is converted. Since the VLC table is searched again based on the obtained AC coefficient data and variable length coding is performed, an effect of further improving the compression efficiency can be obtained.

Embodiment 6 FIG.
In the sixth embodiment, a moving image decoding device that decodes an encoded bit stream generated by the moving image encoding device described in the first or second embodiment will be described.

FIG. 15 is a block diagram illustrating a configuration of a video decoding device according to Embodiment 6.
In the figure, 101 is an encoded bit stream, 102 is a syntax analysis unit, 103 is a shape data codeword, 104 is a block data codeword, 105 is an additional information codeword, 106 is a motion vector codeword, and 107 is a variable length. Decoder 108, shape encoded data, 109, quantized transform coefficient data, 110, additional information, 111, motion vector, 112, shape decoder, 113, texture decoder, 114, motion compensator, 115, decoded shape Data, 116 is a decoded prediction error signal, 117 is a memory, 118 is a reference image, 119 is position information in a decoded macroblock screen, 120 is a predicted image, 121 is a decoding and adding unit, 122 is a decoded image, 123 is a switching unit, It is. Reference numeral 110a is a mode instruction flag included in the additional information 110.

FIG. 16 is a diagram showing an internal configuration of the variable length decoding unit 107 in FIG.
In the figure, 107a is a shape data variable length decoding unit, 107b is a transform coefficient variable length decoding unit, 107c is an additional information variable length decoding unit, and 107d is a motion vector variable length decoding unit. Note that reference numeral 110 b denotes an encoding condition for texture encoding that is the transform coefficient prediction determination information 31 or the quantization step size 32 on the encoding side included in the additional information 110.

Next, the operation of the entire apparatus will be described with reference to FIGS.
First, the coded bit stream 101 is separated into a shape data codeword 103, a block data codeword 104, an additional information codeword 105, and a motion vector codeword 106 by a syntax analysis unit 102, and each of them is subjected to shape data variable length decoding. After being subjected to variable-length decoding by the unit 107a, the transform coefficient variable-length decoding unit 107b, the additional information variable-length decoding unit 107c, and the motion vector variable-length decoding unit 107d, the shape encoded data 108 is quantized by the shape decoding unit 112. The transform coefficient data 109 and various types of additional information 110 are sent to the texture decoding unit 113, and the motion vector 111 is sent to the motion compensation unit 114. It should be noted that the motion compensator 114 inputs the decoded macroblock in-screen position information 119 from the variable length decoder 107.

The shape decoding unit 112 decodes the shape encoded data 108 and outputs it as decoded shape data 115. The decoded shape data 115 is used by the texture decoding unit 113 and the motion compensation unit 114 for decoding texture data and performing motion compensation only for valid pixels.

The texture decoding unit 113 inversely transforms the quantized transform coefficient data 109 using the quantization step size included in the additional information 110, and then performs inverse orthogonal transform to obtain a decoded prediction error signal 116. The decoded prediction error signal 116 is sent to the decoding adder 121.

The motion compensating unit 114 extracts the predicted image 120 from the reference image 118 in the memory 117 according to the position information 119 in the decoded macroblock screen and the motion vector 111 decoded from the encoded bit stream 101, and sends the predicted image 120 to the switching unit 123. Output.

If the switching unit 123 indicates intra-encoding based on the mode instruction flag 110a indicating the encoding mode on the encoding side included in the decoded additional information 110, the prediction unit 120 Is switched to the output 0 side, while in the case of inter-coding, since the predicted image 120 is required for generating the decoded image, it is switched to the output side of the predicted image 120.

Based on the mode instruction flag 110a included in the additional information 110, the decoding addition unit 121 outputs the decoded prediction error signal 116 output from the texture decoding unit 113 as a decoded image 122 in the case of an intra-coded block. On the other hand, in the case of an inter-coded block, a predicted image 120 is added to a decoded prediction error signal 116 which is an output of the texture decoding unit 113, and the resultant is output as a decoded image 122. The decoded image 122 is written to the memory 117 so as to be used as the reference image 118 in the subsequent VOP decoding processing.

Next, operations of the transform coefficient variable length decoding unit 107b and the texture decoding unit 113, which are the most important elements of the sixth embodiment, will be described.

FIG. 17 is an internal configuration diagram of the transform coefficient variable length decoding unit 107b of FIG.
17, 107b1 is a switching unit, 107b2 is a DC coefficient data decoding unit, 107b3 is a decoded DC coefficient, 107b4 is an AC coefficient data codeword, 107b5 is a VLD (Variable Length decoding) table storage unit, 107b6 is A VLD table group selection unit, 107b7 is VLD table group information, 107b8 is an area data decoding unit, 107b9 is VLC table selection information, 107b10 is a VLD table selection unit, 107b11 is a VLD table, and 107b12 is a decoding AC coefficient.
The VLD table storage unit 107b5 stores a VLD table group and a VLD table having the same contents, for example, corresponding to the VLC table group and the VLC table of the VLC table storage unit 4C1 (see FIG. 5 and the like) on the encoding side. Have been.

FIG. 18 is a flowchart illustrating the operation of the transform coefficient variable length decoding unit 107b.
The switching unit 107b1 determines whether the encoding mode of the encoded block to be decoded is the intra encoding or the inter encoding based on the mode instruction flag 110a included in the additional information 110 (step S101). ), In the case of an intra-coded block (“intra” in step S101), the block data codeword 104 is passed to the DC coefficient data decoding unit 107b2, and the DC coefficient data included in the block data codeword 104 is decoded ( Step S102), output as the decoded DC coefficient 107b3, and then decode the AC coefficient data.

On the other hand, in the case of the inter mode (“inter” in step S101), since the decoding processing of the DC coefficient data in step S102 is not performed, only the decoding of the AC coefficient data is performed. Since the AC coefficient data is encoded for each of a plurality of divided regions in the block, decoding is performed for each region.

That is, first, a variable length decoding (hereinafter abbreviated as VLD (Variable Length Decoding)) table group used for decoding the AC coefficient data in the area is selected for each area (step S103).
The selection of the VLD table group is performed by the VLD table group selection unit 107b6, and the position of the region in the block and the transform coefficient prediction determination information 31 or the quantization step used in the transform coefficient variable length coding unit 4c on the encoding side are used. This is performed based on the encoding condition 110b at the time of encoding the texture of size 32.

Since the 予 測 transform coefficient prediction determination information 31 or the quantization step size 32 is necessary at the time of decoding the texture, it is multiplexed as additional information in the bit stream 101 for each block. The operation of VLD table group selection section 107b6 is the same as the operation of VLC table group selection section 4c6 described in the first or second embodiment.

{Circle around (4)} On the encoding side, AC coefficient data is encoded by performing variable length encoding on (RUN, LEVEL) data obtained by scanning the area in a predetermined order. At this time, the variable length coding table used for coding each (RUN, LEVEL) data is selected from the VLC table groups selected by the VLC table group selection unit 4c6. In the transform coefficient variable length coding unit 4c described in the first or second embodiment, the type of the VLC table used to code each (RUN, LEVEL) data is coded as VLC table selection information 4c9. ing.

Therefore, in transform coefficient variable length decoding section 107b described in the sixth embodiment, region data decoding section 107b8 uses VLC used to encode each (RUN, LEVEL) data included in block data codeword 104. It decodes table selection information 107b9 and outputs this VLC table selection information 107b9 to VLD table selection section 107b10.

The VLD table selection unit 107b10 selects a VLD table 107b11 used for decoding (RUN, LEVEL) data from the VLD table storage unit 107b5 based on the VLC table group information 107b9, and outputs it to the area data decoding unit 107b8. (Step S104).

The area data decoding unit 107b8 decodes (RUN, LEVEL) data in the area using the VLD table 107b11 from the VLD table selection unit 107b10.

Specifically, first, the AC coefficient data codeword 107b4 included in the block data codeword 104 is searched in the VLD table 107b11, and if the corresponding (RUN, LEVEL) data is found, it is decoded. It is output as AC coefficient data 107b12.

On the other hand, when the corresponding (RUN, LEVEL) data is not found and the escape code is detected, the AC coefficient data codeword 107b4, which is the next fixed-length codeword of RUN and LEVEL, is fixed-length decoded. Then, (RUN, LEVEL) data can be obtained, and this is output as decoded AC coefficient data 107b12. The above processing of steps S104 to S105 is repeated until all (RUN, LEVEL) data is decoded in the area (step S106 "No").

Then, when decoding of all (RUN, LEVEL) data in a certain area is completed (“Yes” in step S106), it is determined whether the processing in steps S103 to S106 has been performed on all the areas in the block. It is determined whether or not the process has been performed (step S107). If the process has not been performed (step S107 “No”), the processes of steps S103 to S106 are performed, whereas if the process has been performed (step S107 “Yes”), Terminate the processing.

Then, as described in the first embodiment, the (RUN, LEVEL) data and the VLC table selection information 107b9 (NEXT) used to encode the following (RUN, LEVEL) data are grouped as (NEXT). , RUN, LEVEL) data, the (RUN, LEVEL) data and the VLC table selection information 107b9 (NEXT) used for decoding the next (RUN, LEVEL) data are variable-length decoded together. The basic operation other than that is the same.

The above decoding process is performed on all the regions in the block, and the reverse process to the scan on the encoding side is performed based on the obtained decoded DC coefficient 107b3 and (RUN, LEVEL) of the decoded AC coefficient data 107b12. Is performed, the quantized transform coefficient data 109 is obtained. Note that the quantized transform coefficient data 109 is output to the texture decoding unit 113.

Next, the detailed configuration and operation of the texture decoding unit 113 will be described.
FIG. 19 is an internal configuration diagram of the texture decoding unit 113. In the figure, 113a is a switching unit, 110b1 is transform coefficient prediction determination information, 113c is a transform coefficient predictor, 113d is a differential quantized transform coefficient, 113e is an adjacent block quantized transform coefficient, 113f is a quantized transform coefficient, and 113g is a quantized transform coefficient. An inverse quantization unit, 113h is a quantization step, 113i is a DCT coefficient, and 113j is an inverse DCT unit. 110b1 is transform coefficient prediction determination information which is one of the encoding conditions 110b for texture encoding included in the additional information 110.

Next, the operation will be described. First, when the transform coefficient prediction determination information 110b1 included in the decoded additional information 110 indicates that transform coefficient prediction is to be performed, the switch 113a performs transform coefficient prediction by the transform coefficient predictor 113c. Do.

The transform coefficient prediction unit 113c receives the differential quantized transform coefficient 113d, adds the quantized transform coefficient 113e of the neighboring block, and obtains the quantized transform coefficient 113f.
Here, the prediction of the quantized DC coefficient is performed by adding the quantized DC coefficient of the neighboring block as a predicted value to the differential quantized DC coefficient. Selection of a block used for prediction is performed in the same manner as on the encoding side.

The prediction of the quantized AC coefficient is performed by first determining whether or not the transform coefficient prediction has been performed based on the transform coefficient prediction determination information 110b1 included in the additional information 110. Is performed, the quantized AC coefficient of the neighboring block is added to the differential quantized AC coefficient as a prediction value, similarly to the encoding side.

In this way, the transform coefficient prediction unit 113c obtains the quantized DC coefficient and the quantized AC coefficient, and uses the obtained quantized DC coefficient and the quantized AC coefficient as the quantized transform coefficient 113f. Output to the quantization unit 113g.

The inverse quantization unit 113g inversely quantizes the quantized transform coefficient 113f using the quantization step size 113h included in the additional information 110 and outputs the result as a DCT coefficient 113i, and the inverse DCT unit 113j inversely transforms the DCT coefficient 113i. DCT is performed and output as a decoded prediction error signal 116.

Therefore, according to the sixth embodiment, on the encoding side, a block is divided into a plurality of regions, and a VLC table group used to encode AC coefficient data for each region is used as the transform coefficient prediction determination information for each block. Alternatively, even when the bit stream is selected based on the quantization step size, the VLC table is selected from the VLC table group, the AC coefficient data ((RUN, LEVEL) data) is encoded, and the bit stream is transmitted, It can be decoded correctly.

Embodiment 7 FIG.
The seventh embodiment shows another embodiment of the transform coefficient variable length decoding unit described in the sixth embodiment. The transform coefficient variable length encoding unit 4c described in the third embodiment performs variable length coding. The transform coefficient variable length decoding unit that decodes the encoded transform coefficients will be described.

FIG. 20 is an internal configuration diagram of the transform coefficient variable length decoding unit 107b according to the seventh embodiment. In the figure, reference numeral 107b13 denotes position information, and an area data decoding unit 107b8 of the seventh embodiment outputs VLC table selection information 107b9 (see FIG. 17) to a VLD table selection unit 107b10 as in the sixth embodiment. , The position information 107b13 is output.

Next, the operation will be described. The operations of the switching unit 107b1, the DC coefficient data decoding unit 107b2, and the VLD table group selection unit 107b6 are the same as those in the sixth embodiment, and the operations of the VLD table group selection unit 107b6 are the same as those described in the third embodiment. The same processing as performed by the VLC table group selection unit 4c6 on the conversion side may be performed.

Next, the VLD table selection unit 107b10 selects a VLD table used for decoding (RUN, LEVEL) data according to the position of the scan in the area. Position information 107b13 indicating the position of the scan in the area is obtained from the area data decoding unit 107b8. The region data decoding unit 107b8 decodes the (RUN, LEVEL) data in the order scanned on the encoding side, so that the next (RUN, LEVEL) data to be decoded is based on the decoded (RUN, LEVEL) data. Can be specified and output as position information 107b13.

The area data decoding unit 107b8 decodes (RUN, LEVEL) data using the VLD table 107b11 selected by the VLD table selection unit 107b8. Based on the decoded (RUN, LEVEL) data, the scan position of the next (RUN, LEVEL) data to be decoded is specified, and is output to the VLD table selection unit 107b10 as position information 107b13.

Therefore, in the seventh embodiment, a block is divided into a plurality of regions, a VLC table group used for encoding AC coefficient data is selected for each region, and encoding is performed from the selected VLC table group. Since the VLD table used to decode the AC coefficient data is selected based on the scan position of the (RUN, LEVEL) data which is the AC coefficient data to be used, as in the encoding side of the third embodiment, Even if a bit stream in which the VLC table is selected according to the scan position in each area and the coded AC coefficient data is multiplexed is transmitted, the decoding can be performed correctly.

Embodiment 8 FIG.
The eighth embodiment is another embodiment of the transform coefficient variable length decoding unit described in the sixth embodiment or the seventh embodiment, and includes a transform coefficient variable length encoding unit 4c described in the fourth embodiment. A description is given of a transform coefficient variable length decoding unit that decodes transform coefficients that have been subjected to variable length coding.

FIG. 21 shows an internal configuration diagram of transform coefficient variable length decoding section 107b in the present embodiment.
In the figure, 107b13 is a region valid / invalid information codeword, 107b14 is a region valid / invalid information decoding unit, 107b15 is region valid / invalid information, 107b16 is a region valid / invalid determination unit, and 107b17 is a region data zero setting unit. is there.

In FIG. 21, the operations are different in area valid / invalid information decoding section 107b14, area valid / invalid determination section 107b16, and area data zero setting section 107b17. Other configurations and operations are the same as those in the seventh embodiment and the like. Therefore, the following mainly describes the operations of the area valid / invalid information decoding unit 107b14, the area valid / invalid determination unit 107b16, and the in-area AC coefficient zero setting unit 107b17.

In the eighth embodiment, first, region valid / invalid information decoding section 107b14 inputs and decodes region valid / invalid information codeword 107b13 included in block data codeword 104, and the AC coefficient data in each region is It is output as area valid / invalid information 107b15 indicating whether or not it has been encoded.
The region validity / invalidity determination unit 107b16 determines whether or not the AC coefficient data in the region is encoded based on the region validity / invalidity information 107b15 decoded by the region validity / invalidity information decoding unit 107b14.

Here, when the area valid / invalid information 107b15 indicates that the AC coefficient data in the area is coded, the area valid / invalid determination unit 107b16 outputs the AC coefficient data code included in the block data codeword 104. The word 107b4 is output to the area data decoding unit 107b8. The region data decoding unit 107b8 uses the VLD table 107b11 selected by the VLD table selection unit 107b10 based on the VLC table selection information 107b9 in the same manner as in the seventh embodiment or the like, and uses the (RUN, LEVEL) data in the region. And outputs it as decoded AC coefficient data 107b12.

On the other hand, when the area valid / invalid information 107b15 indicates that the area is skipped, that is, that the AC coefficient data in the area is not encoded, the area valid / invalid determination unit 107b16 sends the area data zero setting unit 107b17 The command is sent, and the area data zero setting unit 107b17 sets all the AC coefficients in the area to zero, and outputs the result as decoded AC coefficient data 107b12.

Therefore, in the eighth embodiment, in order to reduce the code amount on the encoding side, the AC coefficient data in a certain area is not encoded, and the area valid / invalid information 107b15 indicating the AC coefficient data is multiplexed on the encoded bit stream. In this case, the region valid / invalid information 107b15 in the received coded bit stream is decoded, and it is determined whether the AC coefficient data is zero for each region in the block. When all the coefficients indicate zero, the coding of the AC coefficient data in the area is not performed, so that the area valid / invalid information 107b15 is multiplexed as in the coding side of the fourth embodiment. Even if the coded bit stream is transmitted, it can be correctly decoded.

Embodiment 9 FIG.
The ninth embodiment shows another embodiment of the area data decoding section of the transform coefficient variable length decoding section described in the sixth to eighth embodiments, and the transform coefficient variable length decoding section described in the fifth embodiment. The transform coefficient variable length decoding unit that decodes the transform coefficient that has been subjected to the variable length coding by the coding unit 4c will be described.

FIG. 22 is an internal configuration diagram of the area data decoding unit 107b8 according to the ninth embodiment. In the figure, 107b81 is a VLD table search unit, 107b82 is an escape code, 107b83 is a conversion flag determination unit, 107b84 is a decoded conversion AC coefficient, 107b85 is a conversion coefficient inverse conversion unit, 107b86 is a conversion coefficient fixed length decoding unit, and 107b87 is a conversion. AC coefficient codeword.

Next, the operation will be described.
The area data decoding unit 107b8 decodes (RUN, LEVEL) data in the area using the VLD table 107b11 selected by the VLD table selection unit 107b10 (see FIG. 21).

At this time, in the area data decoding unit 107b8, first, the VLD table search unit 107b81 converts (RUN, LEVEL) data corresponding to the AC coefficient data codeword 107b4 included in the block data codeword 104 from the VLD table selection unit 107b10. When the corresponding (RUN, LEVEL) data is found in the VLD table 107b11, it is output as decoded AC coefficient data 107b12.

On the other hand, when (RUN, LEVEL) corresponding to the AC coefficient data codeword 107b4 is not found and the escape code 107b82 is detected, the VLD table search unit 107b81 of the area data decoding unit 107b8 sets the escape code 107b82 and Is output to the conversion flag determination unit 107b83, and the conversion flag determination unit 107b83 determines the conversion flag following the escape code 107b82.

Here, if the conversion flag indicates that LEVEL or RUN has been converted, the conversion flag determination unit 107b83 returns the conversion AC coefficient codeword 107b87 following the conversion flag to the VLD table search unit 107b81.

The VLD table search unit 107b81 performs variable-length decoding on the converted AC coefficient codeword 107b87 using the VLD table 107b11, and outputs the decoded AC coefficient codeword 107b87 to the conversion coefficient inverse conversion unit 107b85 as a decoded converted AC coefficient 107b84. The transform coefficient inverse transform unit 107b85 performs a process on the decoded transform AC coefficient 107b84, which is the reverse of the process of the transform coefficient transform unit 4c103 (see FIG. 14) of the encoding apparatus described in the fourth embodiment. The decoded AC coefficient data 107b12 is obtained.

If the conversion flag indicates that RUN and LEVEL are each fixed-length coded, the conversion flag determination unit 107b83 outputs the AC coefficient data codeword 107b4 to the conversion coefficient fixed-length decoding unit 107b86. The fixed-length transform coefficient decoding unit 107b86 performs fixed-length decoding on the fixed-length codewords of RUN and LEVEL following the conversion flag to obtain decoded AC coefficient data 107b12. Similarly, the above process is repeated until all (RUN, LEVEL) data in the area is decoded.

The above decoding process is performed on all the regions in the block, and based on the obtained decoded DC coefficient 107b3 and decoded AC coefficient data 107b12, the order of the scan on the encoding side is reversed. By arranging and outputting, the decoded quantized transform coefficient data 109 is obtained. The decoded quantized transform coefficient data 109 is output to the texture decoding unit 113 (see FIG. 15).

As described in the sixth embodiment, (RUN, LEVEL) data and VLC table selection information 107b9 (NEXT) used to encode the following (RUN, LEVEL) data are grouped as (NEXT). , RUN, LEVEL) data is decrypted in the same basic manner.

Therefore, in the ninth embodiment, when the coding side performs variable length coding of AC coefficient data in a region using a VLC table selected for each region, the AC coefficient data to be coded in the VLC table is used. If there is no code word corresponding to, even if the AC coefficient data is converted and multiplexed again into a bit stream including the AC coefficient data encoded by the VLC table, the conversion flag determination unit 107b83 performs the conversion. Since the conversion of the AC coefficient data on the encoding side is determined and the processing reverse to that on the encoding side is performed in the area data decoding unit 107b8, an escape code is used as in the encoding side of the fifth embodiment. Even when an encoded bit stream obtained by converting and multiplexing the AC coefficient data is transmitted, it can be correctly decoded.

FIG. 1 is a block diagram illustrating a configuration of a video encoding device according to Embodiment 1 of the present invention. FIG. 3 is a block diagram showing an internal configuration of a variable length coding unit according to Embodiment 1 of the present invention. FIG. 2 is a block diagram showing an internal configuration of a texture encoding unit according to Embodiment 1 of the present invention. FIG. 3 is an explanatory diagram of variable coefficient prediction according to the first embodiment of the present invention. FIG. 3 is a block diagram showing an internal configuration of a transform coefficient variable length coding unit 4c according to Embodiment 1 of the present invention. 5 is a flowchart showing the operation of the transform coefficient variable length coding unit 4c according to Embodiment 1 of the present invention. FIG. 4 is a diagram for explaining an area serving as a unit for encoding AC coefficient data according to the first embodiment of the present invention. FIG. 4 is a diagram illustrating a VLC table group according to Embodiment 1 of the present invention. FIG. 13 is a block diagram showing an internal configuration of a transform coefficient variable length coding unit 4c according to Embodiment 2 of the present invention. FIG. 13 is a block diagram showing an internal configuration of a transform coefficient variable length coding unit 4c according to Embodiment 3 of the present invention. FIG. 13 is a block diagram showing an internal configuration of a transform coefficient variable length coding unit 4c according to Embodiment 4 of the present invention. FIG. 9 is a diagram illustrating a macroblock data structure in conventional MPEG-4. FIG. 14 is a diagram illustrating a macroblock data structure according to a fourth embodiment of the present invention. FIG. 15 is a block diagram showing an internal configuration of an area data encoding unit according to Embodiment 5 of the present invention. FIG. 15 is a block diagram showing a configuration of a video decoding device according to Embodiment 6 of the present invention. FIG. 21 is a block diagram showing an internal configuration of a variable length decoding unit according to Embodiment 6 of the present invention. FIG. 21 is a block diagram showing an internal configuration of a transform coefficient variable length decoding unit according to Embodiment 6 of the present invention. 15 is a flowchart showing an operation of a transform coefficient variable length decoding unit according to Embodiment 6 of the present invention. FIG. 21 is a block diagram showing an internal configuration of a texture decoding unit according to Embodiment 6 of the present invention. FIG. 21 is a block diagram showing an internal configuration of a transform coefficient variable length decoding unit according to Embodiment 7 of the present invention. FIG. 21 is a block diagram showing an internal configuration of a transform coefficient variable length decoding unit according to Embodiment 8 of the present invention. FIG. 39 is a block diagram showing an internal configuration of an area data decoding unit according to Embodiment 9 of the present invention.

Explanation of reference numerals

1 input shape data, 2 shape coding unit, 3 shape coding data, 4 variable length coding unit, 5 multiplexing unit, 6 local decoded shape data, 7 input texture data, 8 intra / inter determination unit, 9 mode indication Flag, 10 motion detector, 11 motion vector, 12 motion compensator, 13 predicted image signal, 14 prediction error signal, 15 encoded image signal, 16 texture encoding unit, 17 quantized transform coefficient data, 18 texture local decoding Unit, 19 local decoding prediction error signal, 20 local decoding image signal, 21 switching unit, 22 reference image signal, 23 memory, 24 shape data codeword, 25 additional information codeword, 26 block data codeword, 27 motion vector codeword , 28 coded bit stream, 29 shape data variable length coding unit, 30 Additional information variable length coding section 4c {transform coefficient variable length coding section 442} motion vector variable length coding section, 16a block data extraction section, 16b block data, 16c DCT section, 16d DCT coefficient, 16e quantization section, 16f quantization Transform coefficient, 16g transform coefficient predictor, 42 transform coefficient prediction determination information, 43 switcher, 44 DC coefficient data encoder, 45 AC coefficient data, 4c5 DC coefficient codeword, 4c6 VLC table group selector, 4c7 VLC table group Information, 4c8 @ VLC table selection unit, 4c9 @ VLC table selection information 107b9, 4c10} area data encoding unit, 4c11 {AC coefficient codeword, 26} block data codeword, 54 {quantization step size, 55} encoding condition, 4c12} area data valid / Nothing Judgment unit, 4c13 area valid / invalid information, 4c14 area valid / invalid information coding unit, 4c15 area valid / invalid information codeword, 60 VLC table search unit, 4c102 escape code, 4c103 conversion coefficient conversion unit, 4c104 conversion AC coefficient data 4c106 conversion flag, 4c106 conversion coefficient fixed-length decoding unit, 101 encoded bit stream, 102 syntax analysis unit, 103 shape data codeword, 104 block data codeword, 105 additional information codeword, 106 motion vector codeword, 107 Variable length decoding unit, 108 {shape encoded data, 109} quantized transform coefficient data, 110} additional information, 111 # motion vector, 112 # shape decoding unit, 113 # texture decoding unit, 114 # motion compensation unit, 115 # decoded shape data, 1 16} decoded prediction error signal, 117 memory, 118 reference image, 119 decoded macroblock screen position information, 120 predicted image, 121 decoded addition unit, 122 decoded image, 123 switching unit, 110a mode instruction flag, 107a shape data variable length Decoding section, 107b {transform coefficient variable length decoding section, 107c} additional information variable length decoding section, 107d {motion vector variable length decoding section, 107b1} switching section, 107b2 {DC coefficient data decoding section, 107b3} decoded DC coefficient, 107b4 {AC coefficient data codeword, 110b encoding condition, 107b6 VLD table group selection unit, 107b7 VLD table group information, 107b8 region data decoding unit, 107b9 VLC table selection information 107b9, 107b10 VLD table selection unit, 07b11 {VLD table, 107b12} area data decoding unit, 107b12 {decoding AC coefficient, 113a} switching unit, 110b1} conversion coefficient prediction determination information, 113c} conversion coefficient prediction unit, 113d {difference quantization conversion coefficient, 113e} adjacent block quantization conversion coefficient, 113f} quantization Transform coefficient, 113g {inverse quantization unit, 113h} quantization step, 113i {DCT coefficient, 113j} inverse DCT unit, 107b13} position information, 107b {region valid / invalid information codeword, 107b14} region valid / invalid information decoding unit, 107b15} region valid / invalid Information, 107b16 {area valid / invalidity determination section, 107b17} area data zero setting section, 107b81 {VLD table search section, 107b82} escape code, 107b83} conversion flag determination section, 107b 4 decoded transform AC coefficients, 107B85 transform coefficient inverse transform unit, 107B86 transform coefficient fixed length decoder, 107B87 converts AC coefficient codewords.

Claims (18)

An input image signal is transformed and quantized in units of blocks to generate quantized transform coefficients, and the quantized transform coefficients are variable-length coded by a variable-length coding table for each predetermined area in the block to form a coded bit stream. In the moving picture encoding device to generate,
A plurality of variable-length coding tables that are grouped in advance;
A variable-length coding table group selecting unit that selects one variable-length coding table group from the plurality of variable-length coding table groups;
A code for performing variable length coding on the quantized transform coefficients using the variable length coding table of the selected group, and generating a coded bit stream including information indicating the variable length coding table of the selected group. Bit stream generation means;
A moving picture coding apparatus comprising: The plurality of variable length encoding table groups are grouped in advance based on encoding conditions at the time of quantization,
2. The moving picture coding apparatus according to claim 1, wherein the variable length coding table group selecting means selects a variable length coding table group based on a coding condition at the time of quantization. The variable-length coding table group includes a plurality of variable-length coding tables grouped based on the presence or absence of a transform coefficient prediction for calculating a difference between quantized transform coefficients between blocks.
The variable-length coding table group selecting means selects a corresponding variable-length coding table group from the plurality of variable-length coding table groups based on the presence or absence of the transform coefficient prediction,
2. The coded bit stream generating unit generates a coded bit stream including information indicating the presence or absence of the transform coefficient prediction as information indicating a variable length coding table of a selected group. Video encoding device. The variable length coding table group includes a plurality of variable length coding tables grouped according to the quantization step size,
The variable length coding table group selecting means selects a corresponding variable length coding table group from the plurality of variable length coding table groups based on the quantization step size,
2. The moving picture according to claim 1, wherein the coded bit stream generating means generates a coded bit stream including information on the quantization step size as information indicating a variable length coding table of the selected group. Encoding device. A variable-length encoding table for switching the variable-length encoding table from the variable-length encoding table group selected by the variable-length encoding table group selecting means based on the position of the quantized transform coefficient in the area; 5. The moving picture coding apparatus according to claim 1, further comprising a switching unit. Further, when the quantized transform coefficients are subjected to variable-length coding, a case where all the quantized transform coefficients in each region are zero is an invalid region, and a case where a non-zero coefficient is present in the region is an effective region. Data valid / invalid determining means,
When the area data validity / invalidity determination means determines that the area is an effective area, the coded bit stream generation means encodes the quantized transform coefficient in the area using the selected variable-length coding table and uses it. While generating an encoded bit stream including information indicating the variable-length coding table that has been generated, if it is determined to be an invalid area, generating an encoded bit stream including information indicating that the area is invalid The moving picture coding apparatus according to any one of claims 1 to 5, wherein: When performing variable-length encoding on the quantized transform coefficient, if the corresponding codeword is not included in the variable-length encoding table, the encoded bitstream generating unit converts the quantized transform coefficient without the corresponding codeword into a predetermined codeword. A method for converting to a value included in the variable-length coding table according to the method, and performing variable-length coding on the quantized conversion coefficient after the conversion using the variable-length table, and coding bits including information indicating the conversion method. The moving picture coding apparatus according to claim 1, wherein the moving picture coding apparatus generates a stream. An encoded bit stream is input, and variable-length encoded data that is variable-length encoded for each block included in the encoded bit stream is variable-length decoded to restore a quantized transform coefficient. In a video decoding device that decodes an image signal by performing inverse quantization and inverse transformation,
A plurality of variable-length decoding tables grouped into groups,
Decoding information indicating a variable length coding table group used on the coding side from the coded bit stream, and selecting a variable length decoding table group from the variable length decoding table group based on the variable length coding table group information. Variable length decoding table group selecting means for selecting
Quantized transform coefficient restoring means for restoring the quantized transform coefficient by performing variable length decoding of the encoded data using the variable length decoding table of the selected group,
A video decoding device comprising: The plurality of variable length decoding table groups are grouped in advance based on encoding conditions at the time of quantization,
The variable length decoding table group selecting means decodes information indicating an encoding condition at the time of quantization as information indicating the variable length encoding table group used on the encoding side from the encoded bit stream, and 9. The moving picture decoding apparatus according to claim 8, wherein a variable length decoding table group is selected based on encoding conditions at that time. The variable-length decoding table group includes a plurality of variable-length decoding tables grouped based on the presence or absence of a transform coefficient prediction for calculating a difference between quantized transform coefficients between blocks,
The variable length decoding table group selecting means decodes information indicating presence / absence of quantization transform coefficient prediction from the coded bit stream as information indicating a variable length coding table group used on the encoding side, and performs quantization conversion. 9. The moving picture decoding apparatus according to claim 8, wherein a variable length decoding table group is selected based on information indicating presence / absence of coefficient prediction. The variable length decoding table group includes a plurality of variable length decoding tables grouped based on a quantization step size,
The variable length decoding table group selecting means decodes the quantization step size as information indicating the variable length coding table group used on the encoding side from the encoded bit stream, based on the decoded quantization step size, 9. The moving picture decoding device according to claim 8, wherein a variable length decoding table group is selected. The quantized transform coefficient restoring unit is configured to switch the plurality of variable length decoding tables in the variable length decoding table group selected by the variable length decoding table group selecting unit to decode the quantized transform coefficient, A variable length decoding table selecting means for switching the variable length decoding table based on the position of the transform coefficient in the area,
The moving picture decoding apparatus according to any one of claims 8 to 11, further comprising: The quantized transform coefficient restoring means indicates that the region is invalid when the quantized transform coefficients in the region are all zero, and indicates that the region is invalid when there is one or more non-zero quantized transform coefficients. Decodes variable-length coding table selection information indicating a variable-length decoding table used to decode each quantized transform coefficient in, and determines whether the area is invalid based on the decoded variable-length coding table selection information. Determining means for determining;
When the region is determined to be invalid by the determination unit, a quantization conversion coefficient zero setting unit that sets all quantization conversion coefficients in the region to zero,
The moving picture decoding apparatus according to any one of claims 8 to 12, further comprising: Judgment means for judging whether or not information indicating that the quantized transform coefficient has been converted by the predetermined method on the encoding side from the encoded bit stream has been decoded,
When it is determined by the determining means that the information is decoded, the information is decoded by the variable-length decoding table when the quantized transform coefficients are restored, and thereafter, the inverse of the conversion of the predetermined method is performed. Means for performing a process of restoring to the quantized transform coefficient before the conversion,
The moving picture decoding apparatus according to any one of claims 8 to 13, further comprising: The input image signal is transformed and quantized in units of blocks to generate quantized transform coefficients, and the quantized transform coefficients are coded by variable-length coding according to a variable-length coding table defined for each predetermined area in the block. In a moving picture coding method for generating a bit stream,
A variable-length coding table group is selected from a plurality of grouped variable-length coding tables, and the quantized transform coefficients are converted into variable-length codes using the variable-length coding tables of the selected group. And generating a coded bit stream including information indicating the selected variable-length coding table. When selecting a variable-length encoding table group from a plurality of grouped variable-length encoding tables, the variable-length encoding table is selected based on encoding conditions at the time of quantization, and the selected variable-length encoding table is selected. 16. The moving image encoding method according to claim 15, wherein an encoded bit stream including information on encoding conditions at the time of quantization is generated as information indicating. An encoded bit stream is input, and variable-length encoded data that is variable-length encoded for each block included in the encoded bit stream is variable-length decoded to restore a quantized transform coefficient. In a video decoding method for decoding an image signal by inverse quantization and inverse transform,
Upon restoration of the quantized transform coefficients, decodes information indicating the variable length coding table group used on the coding side from the coded bit stream, and from among a plurality of grouped variable length decoding tables, A variable length decoding table group is selected based on the variable length coding table group information, and the coded data is variable length decoded using the variable length decoding table of the selected group to restore the quantized transform coefficient. A moving picture decoding method characterized by the above-mentioned. At the time of restoration of the quantized transform coefficient, as information indicating the variable length coding table group used on the encoding side, decoding of information on encoding conditions at the time of quantization from the encoded bit stream, 18. The moving picture decoding method according to claim 17, wherein a variable length decoding table group is selected based on information on encoding conditions at that time.

Images