JP2004007778A - Motion compensating and decoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method of decoding a code series of picture information coded at a high efficiency especially after a motion compensating process of the code series with variably coded motion vector information. <P>SOLUTION: A multiplexing separator 22 separates an incoming motion-compensated predictively coded code series from a code input terminal 21 into predictive residual codes, MV codes and code table select information. The predictive residual codes are given to a variable length decoder 23. The MV codes and the code table select information are given to an MV decoder 25 which reversely processes to the MV coding according to the MV codes and the code table select information and gives obtained MV values to a MV buffer 26. A code table for decoding the MVs is selected and used among a plurality of code tables every MV group according to the code table select information. The MV buffer 26 holds the MVs and gives them when a motion compensating predictor 10 needs. <P>COPYRIGHT: (C)2004,JPO

Description

The present invention is directed to decoding a highly efficient coded code sequence that converts image information into a digital signal with a smaller code amount in order to efficiently transmit, store, and display an image. The present invention relates to a method for performing motion compensation decoding on a code string in which processing is performed and motion vector information at that time is variably encoded.

<Motion-compensated video coding>
In moving picture coding, motion-compensated inter-picture prediction is generally performed as represented by MPEG or the like. At this time, the information of the motion vector (MV) used in the motion compensation is variable-length coded, multiplexed with the prediction residual code, and transmitted.

MV coding takes a difference from the MV of an adjacent block, and codes each of the horizontal difference value and the vertical difference value with a variable length code. This variable length code is formed of a Huffman code or the like in accordance with the frequency of occurrence of the difference value.

<Motion Compensation Coding Device of Conventional Example>
An example of the configuration of the conventional motion compensation coding apparatus will be described below with reference to FIG.

動 The subtraction unit 2 subtracts the prediction signal from the motion compensation prediction unit 10 from the moving image signal coming from the image input terminal 1 and supplies the result to the DCT 3 as a prediction residual. The DCT 3 performs a DCT (Discrete Cosine Transform) transform process on the prediction residual, and supplies the obtained coefficient to the quantizer 4. The quantizer 4 quantizes the coefficient with a predetermined step width, and supplies the fixed-length code to the variable-length encoder 5 and the inverse quantizer 7. The variable-length encoder 5 compresses the fixed-length prediction residual with a variable-length code, and the resulting code is provided to the multiplexer 61.

On the other hand, the inverse quantizer 7 and the inverse DCT 13 perform inverse processing of the quantizer 4 and the DCT 3 to reproduce the prediction residual. The obtained prediction prediction residual is added to a prediction signal by an adder 12 to form a reproduction image, which is provided to an image memory 11. The reproduced image stored in the image memory 11 is provided to the motion compensation predictor 10.

The motion compensation predictor 10 performs motion compensation according to the MV provided from the MV detector 6, and supplies the obtained prediction signal to the subtractor 2 and the adder 12. The MV detector 6 obtains the spatial movement amount of the reference frame with respect to the frame to be coded for each 16 × 16 pixel or 8 × 8 pixel block, and supplies it to the motion compensation predictor 10 and the MV encoder 62 as MV. The MV encoder 62 obtains a difference from the immediately preceding MV, and encodes the difference value using a predetermined variable length code. The MV code is multiplexed with the code of the prediction residual by the multiplexer 61 and output from the code output terminal 9.

<Conventional Motion Compensation Decoding Device>
FIG. 7 shows a configuration of a conventional example of a motion compensation decoding device corresponding to the motion compensation encoding device of FIG.

The code string of the motion compensated prediction coded coming from the code input terminal 21 is separated from the code of the prediction residual and the code of the MV by the demultiplexer 71, and the code of the prediction residual is changed by the variable length decoder 23. To the MV decoder 72. The variable length decoder 23 converts the variable length code back to a fixed length code, and is provided to the inverse quantizer 7.

The fixed-length code becomes a reproduced DCT coefficient value of the prediction residual in the inverse quantizer 7 and is given to the inverse DCT 13. The inverse DCT 13 converts the 8 × 8 coefficients into a reproduction prediction residual signal and supplies the signal to the adder 12. The adder 12 adds the prediction signal to the reproduction prediction residual signal to form a reproduction image. The reproduced image signal thus obtained is output from the image output terminal 24 and supplied to the image memory 11. The motion compensation predictor 10 performs motion compensation on the image stored in the image memory 11 based on the MV to form a prediction signal. The obtained prediction signal is provided to the adder 12. On the other hand, the MV decoder 72 performs the reverse process of the MV encoder 62 of FIG. 6 to obtain an MV value from the MV code. The obtained MV is provided to the motion compensation predictor 10.

The conventional motion compensation coding apparatus uses one type of code table for variable length coding of the motion vector (MV). The code table used there is created according to the distribution of the MV difference (the probability of occurrence of each difference value), but the MV difference varies greatly depending on the motion situation of the image, and the distribution matches the code table depending on the image part. In many cases, the number of generated codes was large. In particular, when the motion compensation block is made finer and the accuracy is increased, the ratio of the MV information amount increases, so that the influence on the total code amount becomes remarkable.

The present invention has been made in view of the above points, and has a plurality of types of variable length code tables for MV coding, integrates a predetermined number of MVs, and converts one image (one frame or field) into a plurality. A motion compensation code capable of reducing the MV code amount by performing MV encoding using a code table selected in divided MV group units, transmitting the code table selection information, and using a common code table for decoding. It is an object of the present invention to provide a motion compensation decoding method corresponding to encoding.

Therefore, in order to solve the above problems, the present invention provides the following method.

The moving image is a moving image code sequence that has been motion-compensated and predicted coded by a motion vector for each motion compensation block, and a code table is switched from among a plurality of variable-length coding tables so that the motion vector is variable-length coded. A motion compensation decoding method for decoding a moving image code string in which information of a motion vector is multiplexed,
The incoming video code sequence is information on the motion vector, code table selection information on a variable length code table specified for each unit obtained by integrating a plurality of the motion compensation blocks and dividing one image into a plurality, A prediction image obtained by performing motion-compensated prediction coding on the moving image using the motion vector is multiplexed with information of a prediction residual,
From the incoming video code sequence, the motion vector information, the code table selection information, and the prediction residual information are separated and obtained, and a variable length code table corresponding to the code table selection information is used. Variable length decoding of the information of the motion vector, holding the decoded motion vector, supplying the decoded motion vector corresponding to each of the motion compensation blocks, and calculating the information of the prediction residual. A motion compensated predictive decoding method.

(Operation)
The present invention accurately performs motion compensation decoding on a code sequence coded by the following motion compensation coding. In motion compensation video coding, a plurality of variable length code tables for motion vector (MV) coding are provided, a predetermined number of MVs are integrated, and one image (one frame or field) is divided into a plurality. MV coding is performed using a code table selected in units of MV group, the code table selection information is transmitted, and a common code table is used for decoding, so that an appropriate code table can be selected from a plurality of code tables. Is used. Since a plurality of code tables are prepared according to different MV difference occurrence distributions, even if the MV difference distribution in the MV group is significantly different from the average distribution, a code table suitable for the MV group is selected. The selected code table is composed of variable length codes adapted to the distribution of the MV difference of the MV group, and the code amount of the MV group is smaller than that of the average code table.

符号 A code table suitable for each MV group is selected, and the MV code amount is reduced overall. On the other hand, since the code table selection information is added, the code amount increases accordingly, but since it is 1 to 2 bits in the MV group, it can be only slightly increased compared to the reduction effect.

According to the present invention, a code string coded by the following motion compensation coding can be accurately motion compensated decoded.

In motion-compensated video coding, a plurality of types of variable length code tables for motion vector (MV) coding are integrated, a predetermined number of MVs are integrated, and an MV obtained by dividing one image (one frame or field) into a plurality MV coding is performed using a code table selected in units of groups, and the code table selection information is transmitted and a common code table is used for decoding, so that an appropriate code table can be used from among a plurality of code tables. A plurality of code tables are prepared in accordance with different MV difference occurrence distributions, and a code table that is suitable for the MV difference distribution of the MV group and that is suitable for the partial state of the image is selected. Is reduced.

符号 Since the code amount of the prediction residual does not change, the total generated code amount is reduced. In particular, when the motion compensation block is made finer and the accuracy is increased, the ratio of the MV code amount is relatively increased. Therefore, the effect of reducing the MV code amount directly contributes to the reduction of the total code amount.

<First Example of Motion Compensation Coding Apparatus Corresponding to the Present Invention>
A first example of a motion compensation coding apparatus corresponding to the motion compensation decoding method of the present invention (generating a code string to be accurately decoded in the motion compensation decoding method of the present invention) will be described below.

FIG. 1 shows the configuration, and the same components as those in the conventional example of FIG. 6 are denoted by the same reference numerals. FIG. 1 is different from FIG. 6 in that an MV integrator 15, a temporary MV encoder 16, a code amount calculator 17, a code table determiner 18, and a delay unit 19 are added. The operations of the MV encoder 14 and the multiplexer 8 are different from those of the MV encoder 62 and the multiplexer 61 in FIG.

に お い て In this example, a significant difference from the conventional example is the method of coding a motion vector (MV). Inter-picture prediction processing and encoding of prediction residuals are basically the same.

動 The subtraction unit 2 subtracts the prediction signal from the motion compensation prediction unit 10 from the moving image signal coming from the image input terminal 1 and supplies the result to the DCT 3 as a prediction residual. The DCT 3 performs a DCT (Discrete Cosine Transform) transform process on the prediction residual, and supplies the obtained coefficient to the quantizer 4. The quantizer 4 quantizes the coefficient with a predetermined step width, and supplies the fixed-length code to the variable-length encoder 5 and the inverse quantizer 7. The variable-length encoder 5 compresses the fixed-length prediction residual with a variable-length code, and the compressed code sequence is supplied to the multiplexer 8.

On the other hand, in the inverse quantizer 7 and the inverse DCT 13, the inverse processing of the DCT 3 and the quantizer 4 is performed, and the prediction residual is reproduced. The obtained prediction prediction residual is added to a prediction signal by an adder 12 to form a reproduction image, which is provided to an image memory 11.

再生 The reproduced image stored in the image memory 11 is given to the motion compensation predictor 10. The motion compensation predictor 10 generates a prediction signal according to the MV supplied from the MV detector 6 and supplies the prediction signal to the subtractor 2 and the adder 12.

The MV detector 6 obtains the spatial movement amount of the reference frame with respect to the frame to be coded for each 16 × 16 pixel or 8 × 8 pixel block, and supplies it to the motion compensation predictor 10 and the MV integrator 15 as MV.

<MV encoding>
One embodiment of the coding portion of the MV that contributes to the features of the present invention is described below.

で は The MV integrator 15 shown in FIG. 1 integrates about 16 to 64 MVs to form an MV group in which one frame is divided into a plurality. The MV group is preferably a two-dimensional one such as 4 × 4 in FIG. 4 from the viewpoint of encoding characteristics, but may be a one-dimensional one such as 16 in a row. In the two-dimensional case, the array conversion is performed by having a memory for the MV value.

Each MV of the MV group from the MV integrator 15 is provided to the temporary MV encoder 16 and the delay unit 19. The provisional MV encoder 16 performs provisional encoding for all types of code tables included in the MV encoder 14 and provides the obtained code sequence for each code table to the next code amount calculator 17. .

Here, specific examples of the code table are shown in Tables 1 and 2, respectively.

第 The first variable length code table in Table 1 is suitable when the change in MV is relatively small, and the second variable length code table in Table 2 is suitable when the change in MV is relatively large.

FIG. 5 shows examples of the MV when the change is large and when it is small. In the table, s is a sign indicating the polarity (+/-) of the MV. The variable length codes in the table are applied to each of the horizontal and vertical components of the MV difference. When the MV has half-pixel accuracy, the value is applied to a value that is doubled and converted to an integer.

　符号量算出器１７は、それぞれの符号表の符号列毎に符号量が計数され、ＭＶ群単位で符号量を符号表判定器１８に与える。なお、ここで最終的に必要なのは符号量だけであるので、必ずしも正式な符号列を得る必要はなく、仮ＭＶ符号化器１６は各ＭＶ差分の符号長を出力し、符号量算出器１７でそれをＭＶ群単位に積算しても良い。

The code amount calculator 17 counts the code amount for each code sequence in each code table, and provides the code amount to the code table determiner 18 in MV group units. Here, since only the code amount is finally required, it is not always necessary to obtain a formal code string. The temporary MV encoder 16 outputs the code length of each MV difference, and the code amount calculator 17 It may be integrated for each MV group.

The code table determiner 18 compares the code amounts of the respective code tables, and determines and selects the code table having the smallest code amount from, for example, either Table 1 or Table 2. The result of the determination is given to the MV encoder 14 and the multiplexer 8 as code table selection information.

On the other hand, each MV in the MV group is delayed by a delay unit 19 until code table selection information is output, and is then provided to the MV encoder 14.

The basic coding method of the MV encoder 14 is the same as that of the conventional example shown in FIG. 7, but has a plurality of code tables, one of which is selected according to the code table selection information, and used for coding. Is done. The obtained code is provided to the multiplexer 8.

The multiplexer 8 multiplexes the code of the prediction residual given from the variable length encoder 5 and the code of the MV given from the MV encoder 14, and a code table given from the code table determiner 18. The selection information is inserted into the header of the MV code.

符号 The multiplexed code string is output from the code output terminal 9 here.

<Second example of motion compensation coding apparatus according to the present invention>
A second example of the motion compensation coding apparatus corresponding to the motion compensation decoding method of the present invention (generating a code string to be accurately decoded in the motion compensation decoding method of the present invention) will be described below.

FIG. 3 shows the configuration, and the same components as those in the first example of FIG. 1 are denoted by the same reference numerals. 3 includes an activity detector 31 instead of the provisional MV encoder 16 and the code amount calculator 17 as compared with FIG. The operation of the code table determiner 32 is different from that of the code table determiner 18 in FIG.

に お い て In the second example, the difference from the first example is the method of determining the code table, and the other processing methods are basically the same, and only different parts will be described below.

For each MV in the MV group provided from the MV integrator 15, the activity of the MV is obtained by the activity detector 31, and is provided to the code table determiner 32. The activity is obtained for each MV group by the sum of the absolute values of the MV difference values. Here, if the absolute value of the MV difference value is logarithmically converted and then added, the correlation with the MV code amount increases, which is convenient.

The code table determiner 32 determines an appropriate code table from the relationship between the activity and a preset threshold.

な ら If the code table is of two types, Table 1 and Table 2, medium activity is used as the threshold value. If the activity is smaller than that, Table 1 is used. If the activity is larger than that, Table 2 is used. The activity may be a multidimensional activity that shows the distribution status more accurately than the above. Specifically, the frequency of a relatively small MV difference other than 0 (absolute value is 1 to 3) and the frequency of a relatively large MV difference (absolute value is 4 or more) are obtained, and two-dimensional A threshold value.

<Third example of motion compensation coding apparatus according to the present invention>
A third example of the motion compensation coding apparatus corresponding to the motion compensation decoding method of the present invention (generating a code string to be accurately decoded in the motion compensation decoding method of the present invention) will be described below.

３The third example is a combination of the first and second examples, and the configuration is similar to that of FIG. 1 and is not particularly shown.

に お い て In the third example, the difference from the other examples is the method of determining the code table, and the other processing is basically the same. Therefore, only different parts will be described below based on FIG.

(1) The temporary MV encoder 16 and the code amount calculator 17 in FIG. 1 obtain the code amount for each MV group for one type of code table, and supply the code amount to the code table determiner 18. The code table used at this time is the most average among the plurality.

The operation of the code table determiner 18 is similar to the operation of the code table determiner 32 in FIG. 3. The code table determiner 18 determines an appropriate code table from the relationship between the code amount and a preset threshold value. Output. When the code amount is small, a code table suitable for the MV difference with little change as shown in Table 1 is selected, and when the code amount is large, a code table suitable for a MV difference with rapid change as shown in Table 2 is selected. I do.

<Motion Compensation Decoding Apparatus to which Embodiment is Applied>
A motion compensation decoding apparatus to which the embodiment of the motion compensation decoding method according to the present invention corresponding to the motion compensation coding apparatus of each of the above examples is applied will be described below with reference to the drawings.

FIG. 2 shows an example of the configuration of the motion compensation decoding apparatus.

The code string of the motion compensated prediction coded coming from the code input terminal 21 is separated into the code of the prediction residual, the code of the MV and the code table selection information by the demultiplexer 22, and the code of the prediction residual is variable. The MV code and code table selection information are supplied to the long decoder 23 and the MV decoder 25. The variable length decoder 23 converts the variable length code back to a fixed length code, and is provided to the inverse quantizer 7.

予 測 The prediction residual of the fixed-length code becomes a coefficient value in the inverse quantizer 7 and is given to the inverse DCT 13. The inverse DCT 13 converts the 8 × 8 coefficients into a reproduction prediction residual signal and supplies the signal to the adder 12. The adder 12 adds the prediction signal to the reproduction prediction residual signal to form a reproduction image. The reproduced image signal thus obtained is output from the image output terminal 24 and supplied to the image memory 11.

The motion compensation predictor 10 performs motion compensation on the image stored in the image memory 11 based on the MV to form a prediction signal. The obtained prediction signal is provided to the adder 12.

On the other hand, the MV decoder 25 performs the reverse process of the MV encoder 14 in FIG. 1 based on the MV code and the code table selection information, and supplies the obtained MV value to the MV buffer 26.

符号 One of a plurality of types of code tables used in MV decoding is selected and used according to code table selection information in MV group units.

The MV buffer 26 holds the MV and gives it when needed by the motion compensation predictor 10. This is because when the MV groups are integrated in two dimensions, the block order of the main decoding process is different from the unit of the MV group.

FIG. 3 is a diagram illustrating a configuration of a first example of a motion compensation encoding device corresponding to the motion compensation decoding method of the present invention. 1 is a diagram illustrating a configuration of an example of a motion compensation decoding apparatus to which an embodiment of the motion compensation decoding method according to the present invention is applied. It is a figure which shows the structure of the 2nd example of the motion compensation encoding apparatus corresponding to the motion compensation decoding method of this invention. It is a figure showing a situation of MV group by MV integration. It is a figure showing a situation of MV corresponding to a degree of motion change. FIG. 11 is a diagram illustrating a configuration example of a conventional motion compensation encoding device. FIG. 11 is a diagram illustrating a configuration example of a conventional motion compensation decoding device.

Explanation of reference numerals

1 image input terminal 2 subtractor 3 DCT
Reference Signs List 4 Quantizer 5 Variable length encoder 6 MV detector 7 Dequantizer 8, 61 Multiplexer 9 Code string output terminal 10 Dynamic compensation predictor 11 Image memory 12 Adder 13 Inverse DCT
14, 62 MV encoder 15 MV integrator 16 Temporary MV encoder 17 Code amount calculator 18, 32 Code table determiner 19 Delayer 21 Code string input terminal 22, 71 Demultiplexer 23 Variable length decoder 24 Image output terminal 25, 72 MV decoder 26 MV buffer 31 Activity detector

Claims (1)

The moving image is a moving image code sequence that has been motion-compensated and predicted coded by a motion vector for each motion compensation block, and a code table is switched from among a plurality of variable-length coding tables so that the motion vector is variable-length coded. A motion compensation decoding method for decoding a moving image code string in which information of a motion vector is multiplexed,
The incoming video code sequence is information on the motion vector, code table selection information on a variable length code table specified for each unit obtained by integrating a plurality of the motion compensation blocks and dividing one image into a plurality, A prediction image obtained by performing motion-compensated prediction coding on the moving image using the motion vector is multiplexed with information of a prediction residual,
From the incoming video code sequence, the motion vector information, the code table selection information, and the prediction residual information are separated and obtained, and a variable length code table corresponding to the code table selection information is used. Variable length decoding of the information of the motion vector, holding the decoded motion vector, supplying the decoded motion vector corresponding to each of the motion compensation blocks, and calculating the information of the prediction residual. A motion compensated predictive decoding method.

Images