JP2003179931A - Device, method and program for encoding moving image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for encoding a moving image in which cost of a moving image processing system is reduced, and a recording medium in which a program for encoding a moving image is stored and which can be read out with a computer. <P>SOLUTION: This device for encoding a moving image uses an I picture, a B picture and a P picture as well as an MPEG2. When the I picture, the B picture, and the P picture are encoded, processing such as inter-frame prediction coding in forward direction, in backward direction, and intra-coding are done. And further, a macro block which is already encoded in the same frame is used as a reference image for prediction coding. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding method and a computer readable recording medium storing a moving picture coding program.
The present invention relates to a moving picture coding method such as MPEG (Moving Picture Experts Group) 2 that can reduce the amount of coded information, and a computer-readable recording medium that stores a moving picture coding program that implements the method.

[0002]

2. Description of the Related Art MPEG2, which is widely used at present, is a coding system specification which utilizes the correlation existing between consecutive pictures and reduces the amount of coded information by a technique called "motion compensation". In the “motion estimation” necessary for the “complementary motion value”, an image area in a certain picture is used as a reference value, and a reference value that minimizes the difference between the reference value and the image area in the picture to be encoded is used. The spatial relative position (or "motion vector") of the region is determined.

In MPEG2, pictures in a video sequence are classified into the following three types and encoded. That is, an intra picture (Intra-coded picture, I
Picture), predictive-coded picture (Predictive-coded p
icture, P picture), bidirectional predictive coding picture (Bi
directionally predictive-Coded picture, B picture). As shown in FIG. 9, at least one I picture, P picture, and B picture are combined to form one picture group (that is, GOP). Then, a plurality of picture groups are combined to form a series of encoded information, that is, a bit stream.

The I picture is always placed at the head of the picture group. The P picture is always arranged at the end of the picture group, but may be arranged at positions other than the beginning and end of the picture group as necessary. In the example of FIG. 9, P pictures are arranged at the end and the center of the picture group (GOP), respectively. B picture is
At least one I picture is arranged between the I picture and the P picture adjacent to the I picture. In the example of FIG. 9, two B pictures are arranged between the I picture and the central P picture, and two more B pictures are arranged between the central P picture and the last P picture.

An I picture is another picture (ie P
Coding using only the information that the picture itself does not use (picture and B picture) information (intra-frame coding)
It is a picture to be displayed. The P picture is a picture that uses forward I pictures or future P pictures as a “reference picture (reference image)” and is subjected to forward predictive coding (interframe forward predictive coding) on the time axis. . That is, a B picture is a reference picture that is a past or future picture, that is, either an I picture or a P picture, or both an I picture and a P picture.
, And is a picture for which bidirectional predictive coding of forward and backward (interframe backward predictive coding) is performed on the time axis.

[0006]

According to the above-mentioned MPEG2, it is known that the coded information amount can be reduced at a considerably high level, but in recent years, the information amount of moving images to be processed tends to increase. Therefore, it is desired to further improve the coding efficiency. This is because the cost of the moving image processing system can be reduced by narrowing the transmission band and reducing the storage area. As multimedia is advancing, in order to enable many users to use large-capacity communication such as image data with a limited communication capacity, image coding with higher compression efficiency along with faster communication speed Is required.

The present invention has been made under such a background, and provides a moving picture coding method capable of reducing the cost of a moving picture processing system and a computer-readable recording medium storing a moving picture coding program. To provide.

[0008]

The encoding device of the present invention is
A moving picture coding apparatus using an I picture for intra coding, a B picture for interframe forward predictive coding and interframe backward predictive coding, and a P picture for interframe forward predictive coding. When encoding a picture, a B picture, and a P picture, in addition to the inter-frame forward predictive encoding, the inter-frame backward predictive encoding, and the intra-encoding process, the encoded macro block in the same frame is completed. Is used as a reference image for predictive coding.

The encoding method of the present invention is an I picture for intra coding, a B picture for interframe forward predictive coding and interframe backward predictive coding, and a P picture for interframe forward predictive coding. In the process of encoding I-pictures, B-pictures and P-pictures, in addition to the processes of inter-frame forward prediction coding, inter-frame backward prediction coding, and intra-coding, A coding method, comprising: using a coded macroblock in the same frame as a reference image for predictive coding.

The encoding program of the present invention is an I picture for intra encoding, a B picture for performing interframe forward predictive encoding and an interframe backward predictive encoding, and a P picture for performing interframe forward predictive encoding. To use,
A moving picture coding program for operating the coding apparatus according to claim 1, wherein in a process of coding an I picture, a B picture and a P picture, interframe forward prediction coding and interframe backward prediction coding are performed. In addition to the intra-coding process, the computer is caused to execute a coding process including a process of using the coded macroblock in the same frame as a reference image for predictive coding.

A computer-readable recording medium according to the present invention has an I picture for intra coding, a B picture for interframe forward predictive coding and an interframe backward predictive coding, and an interframe forward predictive coding. A recording medium in which a moving picture coding program for operating the coding apparatus according to claim 1, which uses P pictures to be carried out, is recorded, and in a process of coding I pictures, B pictures and P pictures A code including processing of using forward-predictive coding, inter-frame backward predictive coding, and intra-coding, as well as processing of using a coded macroblock in the same frame as a reference image for predictive coding A moving image encoding program that causes a computer to execute the encoding process is recorded.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. 1 and 8 are functional block diagrams showing configuration examples of an encoder and a decoder used for video encoding according to an embodiment of the present invention. This configuration itself is
This is the same as that used for moving image processing according to MPEG2. <Encoder> First, the configuration of the encoder of FIG. 1 will be described. The encoder of the present invention is characterized in that a macroblock, which has been encoded and is present in the same frame, is used as a reference image for predictive encoding. The picture rearrangement unit 11 inputs an uncompressed digital input image signal from an external device and generates a plurality of pictures from this digital input image. Here, the picture is
It means a frame or field that can be treated as one still image. Each picture is composed of a combination of data of one luminance signal and two color difference signals.

The picture rearrangement unit 11 divides the picture into three types of I picture, B picture, and P picture in order to perform predictive coding similarly to MPEG2. Also, the picture rearrangement unit 11 rearranges the generated I picture, B picture, and P picture in an order suitable for the predictive coding process and outputs the rearranged pictures. That is, the macroblock type of the I picture is composed of intra-coded (intra-frame coded) frame data without performing inter-frame prediction, and is output first for use in prediction of other P and B frames. . The macroblock has a structure in which a picture is decomposed into 16 × 16 dot areas,
It is composed of four blocks each consisting of a 4 × 4 dot area.

The macroblock type of a P picture is I
Like a picture, it is composed of intra-coded frame data and inter-frame (forward frame prediction screen) data reproduced by prediction from an I picture (used as a reference image). The macroblock type of a B picture is an interframe (forward frame prediction) reproduced by predicting from an intra-coded frame data similar to an I picture and an I picture (used as a reference image) similar to a P picture. Screen) and backward frame (reverse frame prediction screen) data reproduced by prediction from a P picture (used as a reference image). Therefore, the picture rearrangement unit 11 rearranges the image signals rearranged in the order of pictures into I-pictures in the order necessary for encoding.
The P picture and the B picture are output to the motion estimation unit 12 in this order.

The motion estimation unit 12 detects the type of picture, selects a coding mode corresponding to each type, and outputs this coding mode to the motion prediction unit 20 in association with each picture. That is, if the input image signal (picture) to be encoded is an I picture, the motion estimation unit 12 has already completed encoding in the same frame together with an identification signal indicating that the image is an I picture. When a macro block has a macro block with a high similarity, this macro block is regarded as a similar macro block, and the mode signal of the coding mode indicating that intra-frame coding has been performed is used as the macro block of the macro block to be coded. It outputs to the motion prediction unit 20 together with the block address and the motion vector which is the difference in coordinate value between the encoded macro block and the similar macro block. On the other hand, the motion estimation unit 12
Outputs a mode signal of a coding mode indicating that the macro block is used as it is to the motion prediction unit 20 when there is no macro block having high similarity in the same frame, and outputs to the subtraction unit 13 as it is. .

Further, the motion estimation unit 12 determines, in the GOP, a similar macroblock having a high similarity to the macroblocks to be formed if the input image to be encoded is either a P picture or a B picture. The macroblock address of the macroblock to be coded is extracted from other nearby pictures (I picture for P picture, I and P picture for B picture) and similar to the coded macroblock. A motion vector, which is the difference in coordinate value from the macroblock, and a mode signal of a coding mode indicating which of the neighboring pictures are extracted (either interframe forward prediction coding or interframe backward prediction coding). Are output to the motion prediction unit 20.

Furthermore, if the input image to be encoded is either a P picture or a B picture, the motion estimator 12 is similar to a macroblock in the same frame that has already been encoded. When there is a high macroblock, the identification signal indicating the type of picture and the mode signal of the coding mode (intraframe predictive coding) indicating that the same intraframe coding has been performed are used as the macro to be coded. The result is output to the motion prediction unit 20 together with the motion vector which is the difference between the macroblock address of the block and the coordinate value of the encoded macroblock and the similar macroblock. Here, which of the intra-frame predictive coding, the inter-frame forward predictive coding, and the inter-frame backward predictive coding is performed by the motion estimation unit 12 has the highest similarity (correlation)? And the coding mode with the highest similarity is used.

However, the motion estimator 12 determines whether the input coded image is a P picture or a B picture for the constituent macroblocks.
When a similar macro block having similarity within a predetermined range is not extracted, an identification signal indicating the type of picture, a mode signal of a coding mode indicating that the motion prediction unit 20 is used as it is, and a macro of this macro. The block identification signal is output, and this macroblock is output to the subtraction circuit 13 in the same state as the I picture.

The motion prediction unit 20 stores the identification signals of each picture and each macroblock in GOP units, and, based on the coding mode of each macroblock, requests the subtraction unit 13 (if the similar blocks A subtraction command is sent to only a certain macroblock, and image data (reference image) of a similar macroblock used for subtraction is extracted from the picture storage unit 21 and supplied. The subtractor 13 is
If there is no subtraction command (there is no similar macro block to the input macro block), the macro block is passed through as it is, and if there is a subtraction command (the input macro block does not have a similar macro block), this similarity The difference data from the macroblock, that is, the residual prediction signal is output.

The DCT unit 14 divides a macro block of the state as it is without subtraction processing or a macro block of the residual prediction signal subjected to subtraction into each block, and each block is similar to MPEG2. DCT
(Discrete cosine transform) processing is performed. The quantizer 15 divides the DCT coefficient of each block obtained by the DCT processing by a predetermined divisor (quantization step), rounds the remainder, and performs coding (quantization) (code similar to MPEG2). Process).

The inverse quantizing unit 17 multiplies the quantized DCT coefficient of each block output from the quantizing unit 15 by the predetermined divisor (quantization step) to perform the inverse quantization process. The IDCT (Inverse Discrete Cosine Transform) unit 18 uses the MP
Inverse DCT processing similar to that of EG2 is performed, a macroblock is reconstructed from the decoded block, and this macroblock is output to the adder 19.

Here, the motion predicting section 20 determines, as necessary, to the adder 19 based on the coding mode of each macroblock (only when the reconstructed macroblock is a residual prediction signal). While sending the addition command, the image data of the similar macroblock used for the addition is extracted from the picture storage unit 21 and supplied. If there is no addition command (the input macroblock is not the residual prediction signal and is in the state as it is), the adder 19 passes the macroblock as it is, and the addition command is issued (the input macroblock is input). Is a residual prediction signal), the decoded residual prediction signal and the supplied similar macroblock are added to reproduce the macroblock.

The picture accumulating section 21 composes pictures from the reproduced macroblocks output from the adder 19 and accumulates these pictures. Here, there are two types of pictures accumulated in the picture accumulating unit 21, that is, an I picture and a P picture (the reason for the I picture and the P picture is explained in the operation section).

The variable-length coding unit 16 performs variable-length coding on the quantized block input from the quantizing means 15 in the same manner as MPEG2, and the binary-coded block is macroblock unit. , To the multiplexing unit 22.
The multiplexing means 22 is sent from the variable length coding unit 16 2
The binary-coded macroblock, the motion vector data, and the coded signal indicating the coding mode are multiplexed and output to the buffer 23.

The buffer 23 temporarily stores the multiplexed binary coded / quantized DCT output signal, the motion vector signal, and the coding mode signal, shapes the bit stream in a predetermined format, and outputs the bit stream. This bit stream is transmitted to a decoder (for example, the decoder 3 in FIG. 8) via a predetermined transmission path.
1) is transmitted. The buffer 23 controls the bit rate of the bits output from the quantizer 15 in accordance with the bit rate output to the transmission path.

Next, an operation example of the encoder 1 according to the embodiment of the present invention will be described with reference to FIGS. Here, FIG. 2 is a conceptual diagram showing a configuration example of a GOP (Group of Pictures) for explaining the moving picture coding process executed in the above-mentioned encoder 1.
3 to 6 are conceptual diagrams showing the structure of a picture for explaining the extraction processing of similar macroblocks to the macroblock of each picture. Picture sorting section 11
Is a picture input in time series, for example, I picture {n}, B picture {n + in GOP units shown in FIG.
1, n + 2}, P picture {n + 3}, B picture {n
+4, n + 5}, P picture {n + 6},
Let these pictures be {n, n + 3, n + 1, n + 2,
n + 6, n + 4, n + 5} are rearranged in this order and output. 3 to 5, MBI indicates an I picture,
MBP indicates a P picture and MBB indicates a B picture.

The motion estimation unit 12 encodes each input picture according to the flowchart shown in FIG. In step S1, the motion estimation unit 12 causes the GOP
It is determined whether the picture is an I picture, a B picture, or a P picture in the unit, and these pictures are set as the target pictures in the input order.

Next, in step S2, the motion estimation section 12 divides the target picture into 16 × 16 dot macroblocks. Here, the motion estimation unit 12
As shown in FIGS. 3 to 6, each picture is vertically
Divided into a matrix of 6 rows and 6 columns in the horizontal direction, that is, 48
Divide into macroblocks. Hereinafter, each macroblock corresponds to an I picture, a P picture, and a B picture, respectively, and MI (m (row number of matrix), n (column number of matrix)), MP (m, n), and MB (m , N) (m =
It represents as an integer of 1 to 6, and n = an integer of 1 to 8).

Then, in step S3, the motion estimation section 12 determines whether or not the divided target picture is an I picture. Here, the motion estimation unit 12
If it is determined that this target picture is an I picture, the encoding process proceeds to step S6.

Next, in step S6, the motion estimation unit 12 performs the encoding process on the target picture T for all macroblocks MI, for example, in the order of arrows, as shown in FIG. That is, the motion estimation unit 12 starts the encoding process from the macro block MI (1,1) in the upper left corner (first row, first column) and proceeds to the right, and then moves to the upper right corner (first row, eighth column). ) Macro block MI
It ends once with (1, 8). This macroblock MI
(1,1) to macroblock MI (1,8) are macroblock addresses for identifying macroblocks in each picture. Then, the motion estimation unit 12 immediately moves to the second row, and the macroblock MI in the second row and the first column.
Macro block MI from (2, 1) to the second row and eighth column
Proceed to (2,8). Hereinafter, similarly, the motion estimation unit 1
2 repeats the same movement for the third row to the sixth row, and ends at the macroblock MI (6,8) in the sixth row and eighth column.

The motion estimator 12 refers to the coding of the I picture, for example, by referring to the macroblock that has been coded in the I picture (currently coding itself) as a reference macroblock and newly codes the code. Similar macroblocks that are similar to the macroblock to be converted are detected. That is, the motion estimation unit 12 uses the macroblock MI
Although (1, 1) is encoded, it is an I picture, and there is no macro block that can be referred to. Since intra encoding is performed, the macro block is output to the subtractor 13 as it is and The mode signal of the coding mode indicating that there is is output to the motion prediction unit 20. At this time,
The motion estimation unit 12 does not output the motion vector data to the motion prediction unit 20 because there is no similar macroblock.

Then, the motion predicting section 20 includes the motion estimating section 1
Since a mode signal indicating intra coding is input from 2, the macro block MI corresponding to this mode signal is input.
When (1, 1) is input to the subtractor 13, the subtractor 13
The subtraction command is not output to. At this time, the subtractor 13
Does not receive a subtraction command from the motion prediction unit 20, and therefore outputs the input macroblock MI (1,1) as it is. Next, the DCT unit 14 divides this macroblock MI (1,1) into blocks of 4 × 4 dots,
DCT processing of each block is performed, and the processing result is output to the quantizer 15 as a macroblock MID (1,1).

The quantizer 15 quantizes the input macroblock MID (1,1) and outputs the processing result as the macroblock MIQ (1,1). As a result, the variable length coding 16 determines that the macroblock MIQ
(1,1) is variable length coded to generate macroblock MIV
Output as (1,1). Next, the multiplexing unit 22 encodes the macroblock MIV (1,1) into a mode signal of a coding mode indicating that the macroblock MIV (1,1) is to be intra-coded (a picture signal indicating that an image to be coded is an I picture. (Including an identification signal indicating the type) and output to the buffer 23.

The inverse quantizer 17 inversely quantizes the macroblock MIQ (1,1) output from the quantizer 15 to reproduce the macroblock MID '(1,1), and the IDC
Output to the T section 18. Hereafter, the decoded block contains
Since the decoding result is irreversible, "'" is attached to the name of the macroblock in order to distinguish it from the original macroblock. Then, the IDCT unit 18 performs inverse discrete cosine transform on the input macroblock MID ′ (1,1), reproduces the macroblock MI ′ (1,1), and outputs it to the adder 19. .

At this time, since the macroblock MI '(1,1) corresponds to the intra-coded macroblock MI (1,1), the adder 19 issues an addition command from the motion prediction unit 20. Is not output, no data is added to the macroblock MI (1,1) and is stored in the area corresponding to the identification number of the I picture in the picture storage unit 21. That is, the picture storage unit 21 stores (stores) a reproduced image similar to the image of the I picture transmitted from the buffer 23 as a bit stream and reproduced at the reception destination.

Then, as shown in FIG. 3, the motion estimation unit 12 advances the coding, and when the macroblock MI (4, 4) is coded, the motion estimation section 12 defines the macroblock MI (4, 4),
Macroblock MI ′ read from the picture storage unit 21
The difference data with respect to each of (1, 1) to MI '(4, 3) is calculated, and the macroblock MI' (4, 1) is converted into macroblocks MI '(1, 1) to MI' (4, 3). Among them, when it is determined that the difference data with the macro block MI (4, 4) is the smallest and exceeds the predetermined similarity (the difference data is less than or equal to the set value), the macro block MI ′ (4, 4). 1) is extracted as a similar macroblock.

Therefore, the motion estimation unit 12 encodes the macro block MI (4,4), but since the macro block MI '(4,1) is a similar macro block, the same intra-frame prediction encoding ( A mode signal of a coding mode indicating that macroblocks in the same picture are encoded as similar blocks) is output to the motion prediction unit 20 and the macroblocks MI (4,4) are output to the subtractor 13. . At this time, the motion estimation unit 12 has the mask block M corresponding to the position of the mask block MI (4,4).
The motion vector indicating the relative position of I ′ (4,1) is output to the motion prediction unit 20 and the multiplexing unit 22.

Then, the motion predicting section 20 includes the motion estimating section 1
Since a mode signal indicating the same intraframe predictive coding is input from 2, the subtraction is performed on the subtractor 13 at the time when the macroblock MI (4, 4) corresponding to this mode signal is input to the subtractor 13. A macroblock MI ′ used for subtraction from the picture storage unit 21 while outputting a command.
(4, 1) is read and output to the subtractor 13. At this time, the subtractor 13 receives the subtraction command and the data of the macroblock MI ′ (4,1) used for the subtraction from the motion prediction unit 20, and thus the input macroblock MI (4,4).
4), the macroblock MI '(4, 1) is subtracted, and the subtraction result is used as residual prediction data MS (4, 4) to obtain D
Output to the CT unit 14.

Next, the DCT unit 14 outputs the residual prediction data MS (4,4) of this macroblock MI (4,4) to 4
It is divided into blocks of × 4 dots, DCT processing of each block is performed, and the processing result is defined as macroblock MSD (4,
4) to the quantizer 15. Then, the quantizer 15 quantizes the input macroblock MSD (4, 4) and outputs the processing result as the macroblock MSQ (4, 4). As a result, the variable length coding 16 performs variable length coding on the macroblock MSQ (4,4) and outputs it as the macroblock MSV (4,4). Next, the multiplexing unit 22 encodes the macroblock MSV (4, 4) into a mode signal of an encoding mode indicating that the macroblock MSV (4, 4) is forward prediction encoding between frames (the image to be encoded is an I picture. (Including an identification signal indicating the type of picture) shown in FIG.

The inverse quantizer 17 inversely quantizes the macroblock MSQ (4,4) output from the quantizer 15 to reproduce the macroblock MSD '(4,4), and IDC
Output to the T section 18. Then, the IDCT unit 18 performs an inverse discrete cosine transform on the input macroblock MSD '(4, 4), reproduces the residual prediction signal MS' (4, 4), and sends it to the adder 19. Output.

At this time, since the prediction residual signal MS '(4, 4) corresponds to the macroblock MI (4, 4) that has been intraframe-prediction coded, the adder 19 performs the motion prediction. The addition command is output from the unit 20, and the motion prediction unit 20
The macroblock M read from the picture storage unit 21
I ′ (4,1) and the prediction difference signal MS ′ (4,4) are added to decode the macroblock MI ′ (4,4). Then, the adder 19 stores the decoded macroblock MI ′ (4,4) in the area corresponding to the identification number of the I picture in the picture storage unit 21.

On the other hand, the motion estimation unit 12 similarly
When encoding (4,7), MI (4,7) and M
Difference data from each of I '(1,1) to MI' (4,6) is calculated, and macro blocks MI '(2,6) out of MI' (1,1) to MI '(4,6) are calculated. 4) is the macroblock M (4,
It is determined that the difference data with 7) is the smallest, but if it is determined that the similarity does not exceed the predetermined similarity (the difference data exceeds the set value), the macroblock MI ′ (2,
It is assumed that 4) is not used as a similar macroblock, and that intra-coding is performed in the same manner as the macroblock MI ′ (1,1). Therefore, the motion estimation unit 12 outputs to the motion prediction unit 20 a mode signal of a coding mode indicating that the same intraframe predictive coding is to be performed. The subsequent macroblock MI (4, 7) processing is performed by the macroblock MI (1, 1).
Since it is the same as, the description will be omitted.

Then, when the coding of the macroblock MI (6,8) in the I picture is completed, the motion estimation unit 12 determines that the coding of all the macroblocks is completed and returns the process to step S1. At this point, the picture storage unit 21 stores the I-picture image data that has been encoded (by predictive encoding, DCT processing, quantization, etc.) and then decoded.

Then, in step S1, the picture rearranging section 11 inputs the P picture (N +
3) is set as the target picture. Next, in step S2, the motion estimation unit 12 divides the target picture into macroblocks of 16 × 16 dots.
Here, as shown in FIGS. 3 to 6, the motion estimation unit 12 divides the P picture into a matrix of 8 rows in the vertical direction and 6 columns in the horizontal direction, that is, 48 macro blocks (macro block MP (1,1) to MP (6,8)).

Then, in step S3, the motion estimation section 12 determines whether or not the divided target picture is an I picture. Here, the motion estimation unit 12
It is determined that this target picture is not an I picture,
The process proceeds to step S4. Next, in step S4, the motion estimation unit 12 determines whether or not the divided target picture is a P picture. Here, when the motion estimation unit 12 determines that the target picture is a P picture, the motion estimation unit 12 advances the encoding process to step S7.

Next, in step S7, as shown in FIG. 6, the motion estimation unit 12 performs the coding process on the target picture T for all macroblocks MP as in the case of the I picture already described. For example,
Follow the order of the arrows. That is, the motion estimation unit 12
Macro block MP (1, in the upper left corner (first row, first column))
After starting the encoding process from 1), the process proceeds to the right, and ends at the macro block MP (1,8) at the upper right corner (first row, eighth column). Then, the motion estimation unit 12 immediately outputs the second
Moving to the row, the macroblock MP (2, row 1, column 1)
1) to macro block MP (2,8) in the second row and eighth column
Proceed to. Thereafter, in the same manner, the motion estimation unit 12 determines the third
The same movement is repeated for the rows to the sixth row, and ends at the macroblock MP (6,8) in the sixth row and eighth column.

The motion estimator 12 encodes P-pictures, for example, macros of I-pictures stored in the picture accumulator 21 and P-pictures that have been encoded (currently encoding themselves). The block is referred to as a reference macroblock (and a reference image), and a similar macroblock that is similar to the macroblock to be newly encoded is detected. That is, the motion estimation unit 12 encodes the macroblock MP (1,1) using the I-picture macroblock stored in the picture storage unit 21 as a reference macroblock.

Then, as shown in FIG. 4, the motion estimator 12 starts the coding from the macroblock MP (1,1), the coding advances, and the coding of the macroblock MP (1,6) starts. When performing, the macroblock MP (1,6), the macroblocks MI ′ (1,1) to MI ′ (6,8) of the I-picture stored in the picture storage unit 21, and the macroblock MP ′ of the P-picture. (1,1) to MP '(1,5) Difference data with each of them is calculated to obtain the macro block MI'.
(5, 6) is the macroblock MI '(1, 1) to MI'
(6, 8), macroblocks MP '(1, 1) to MP'
Of the (1, 5), the difference data with the macro block MP (1, 6) is the smallest, and if it is determined that the similarity exceeds a predetermined similarity (the difference data is less than or equal to a set value), the macro block MI ′ (5,6) is extracted as a similar macroblock.

Therefore, the motion estimation unit 12 encodes the macro block MP (1,6), but since the macro block MI '(5,6) is a similar macro block, the inter-frame forward prediction encoding is performed. A mode signal of a coding mode indicating that (previous picture, that is, a macro block in an I picture is encoded as a similar block) is output to the motion prediction unit 20, and the macro block MI is also output.
(1,6) is output to the subtractor 13. At this time, the motion estimation unit 12 uses the motion vector indicating the relative position of the mask block MI ′ (5,6) corresponding to the position of the mask block MI (1,6) as the motion prediction unit 20 and the multiplexing unit 22. Output to.

Then, the motion prediction section 20 is connected to the motion estimation section 1
Since the mode signal indicating the inter-frame forward predictive coding is input from 2, the macro block MP (1,6) corresponding to this mode signal is input to the subtractor 13,
The subtraction command is output to the subtractor 13, and the macro block MI ′ used for subtraction from the picture storage unit 21.
(5, 6) is read and output to the subtractor 13. At this time, the subtractor 13 receives the subtraction command and the data of the macroblock MI '(5, 6) used for the subtraction from the motion prediction unit 20, and thus the input macroblock MP (1,
6), the macroblock MI '(5, 6) is subtracted, and the subtraction result is used as residual prediction data MS (1, 6) to obtain D
Output to the CT unit 14.

Next, the DCT unit 14 outputs the residual prediction data MS (1,6) of this macroblock MP (1,6) to 4
It is divided into blocks of × 4 dots, DCT processing of each block is performed, and the processing result is defined as macroblock MSD (1,
6) to the quantizer 15. Then, the quantization unit 15 quantizes the input macroblock MSD (1,6) and outputs the processing result as the macroblock MSQ (1,6). As a result, the variable length coding 16 performs variable length coding on the macroblock MSQ (1,6) and outputs it as the macroblock MSV (1,6). Next, the multiplexing unit 22 encodes the macroblock MSV (1,6) into a mode signal of an encoding mode indicating that the macroblock MSV (1,6) is forward prediction encoding (the image to be encoded is a P picture. , Including the identification signal indicating the type of picture)
Mask block MI (1,6) and mask block MI '
It is multiplexed with the motion vector indicating the relative position of (5, 6) and output to the buffer 23.

Then, the buffer 23 outputs a bit stream in which the macro block MSV (1,6), the mode signal and the motion vector are multiplexed. In addition, the inverse quantizer 1
7 is a macro block MSQ output from the quantizer 15.
Inverse quantization of (1,6) is performed and macroblock MSD '
(1,6) is reproduced and output to the IDCT unit 18. Then, the IDCT section 18 receives the input macroblock MS
Inverse discrete cosine transform is performed on D ′ (1,6) to reproduce the residual prediction signal MS ′ (1,6), which is output to the adder 19.

At this time, since the prediction residual signal MS '(1,6) corresponds to the intra-frame predictive-coded macroblock MP (1,6), the adder 19 calculates the motion prediction unit. 20 outputs an addition command, and the motion prediction unit 20 reads the macroblock MI ′ from the picture storage unit 21.
(5, 6) and the prediction difference signal MS '(1, 6) are added to decode the macroblock MP' (1, 6).
Then, the adder 19 stores the decoded macroblock MP '(1, 6) in the area corresponding to the identification number of the P picture in the picture storage unit 21.

Then, the motion estimation unit 12 is further stored in the picture storage unit 21 and the macro block MP (5, 6) when the coding proceeds and the macro block MP (5, 6) is coded. The difference data between each of the I picture macroblocks MI '(1,1) to MI' (6,8) and the P picture macroblocks MP '(1,1) to MP' (5,5). Macro block MP '(2,3)
Among the macro blocks MI '(1,1) to MI' (6,8) and the macro blocks MP '(1,1) to MP' (5,5),
When it is determined that the difference data with the macro block MP (5, 6) is the smallest and exceeds the predetermined similarity (the difference data is less than or equal to the set value), the macro block MP '
(2, 3) is extracted as a similar macroblock.

Therefore, the motion estimation unit 12 encodes the macroblock MP (5, 6), but since the macroblock MP '(2, 3) is a similar macroblock, the same intraframe prediction encoding ( A mode signal of a coding mode indicating that a picture currently being coded, that is, a macro block in a P picture is coded as a similar block is output to the motion prediction unit 20, and
The macro block MP (5, 6) is output to the subtractor 13. At this time, the motion estimation unit 12 determines that the mask block MP
Mask block MP 'corresponding to the position of (5, 6)
The motion vector indicating the relative position of (2, 3) is output to the motion prediction unit 20 and the multiplexing unit 22.

Then, the motion predicting section 20 includes the motion estimating section 1
Since the mode signal indicating the same intra prediction coding is input from 2, the macro block MP corresponding to this mode signal is input.
When (5, 6) is input to the subtractor 13, the subtracter 1
3, a subtraction command is output, and a macro block MP '(2, 3) used for subtraction from the picture storage unit 21.
Is read out and output to the subtracter 13. At this time, the subtractor 13 receives the subtraction command and the data of the macroblock MP '(5, 6) used for the subtraction from the motion prediction unit 20, so that the input macroblock MP (5, 6) to the macroblock MP (5, 6) MP '(2, 3) is subtracted, and the subtraction result is output to the DCT unit 14 as residual prediction data MS (5, 6).

Next, the DCT unit 14 outputs the residual prediction data MS (5,6) of this macroblock MP (5,6) to 4
It is divided into blocks of × 4 dots, DCT processing is performed on each block, and the processing result is set to macroblock MD (5,
6) to the quantizer 15. Then, the quantizer 15 quantizes the input macroblock MSD (5, 6) and outputs the processing result as the macroblock MSQ (5, 6). As a result, the variable length coding 16 performs variable length coding on the macroblock MSQ (5, 6) and outputs it as the macroblock MSV (5, 6). Next, the multiplexing unit 22 encodes the macroblock MSV (5, 6) into a mode signal in a coding mode indicating that the macroblock MSV (5, 6) is in the same frame predictive coding (indicating that the image to be coded is a P picture. , And a mask block MP (5, 6) and a mask block MP ′.
It is multiplexed with the motion vector indicating the relative position of (2, 3) and output to the buffer 23.

Then, the buffer 23 outputs a bit stream in which the macro block MSV (5, 6), the mode signal and the motion vector are multiplexed. In addition, the inverse quantizer 1
7 is a macro block MSQ output from the quantizer 15.
Inverse quantization of (5, 6) is performed, and macroblock MSD '
(5, 6) is reproduced and output to the IDCT unit 18. Then, the IDCT section 18 receives the input macroblock MS
The inverse discrete cosine transform is performed on D ′ (5,6) to reproduce the residual prediction signal MS ′ (5,6) and output to the adder 19.

At this time, since the prediction residual signal MS '(5, 6) corresponds to the same intra-frame predictive-coded macroblock MP (5, 6), the adder 19 performs motion prediction. The addition command is output from the unit 20, and the motion prediction unit 20
The macroblock M read from the picture storage unit 21
The macro block MP '(5, 6) is decoded by adding P' (2, 3) and the prediction difference signal MS '(5, 6). Then, the adder 19 stores the decoded macroblock MP '(5, 6) in the area corresponding to the identification number of the P picture in the picture storage unit 21.

Then, the motion estimation unit 12 is further stored in the picture storage unit 21 and the macro block MP (6, 3) when the coding proceeds and the macro block MP (6, 3) is coded. The difference data between each of the I picture macro blocks MI '(1,1) to MI' (6,8) and the P picture macro blocks MP '(1,1) to MP' (6,2). Macro block MI '(5,6)
Among the macroblocks MI ′ (1,1) to MI ′ (6,8) and macroblocks MP ′ (1,1) to MP ′ (6,2),
It is determined that the difference data with the macroblock MP (6, 3) is the smallest.

Next, when the motion estimation unit 12 determines that the difference data does not exceed the predetermined similarity (the difference data exceeds the set value), a code indicating that intra coding is to be performed. The mode signal of the conversion mode is output to the motion prediction unit 20. The following encoding process is the same as that of the macro block MI (1,1) already described, and thus the description thereof is omitted. Finally, the buffer 23 outputs the bitstream in which the macroblock MSV (6,3) and the mode signal are multiplexed. And the adder 19
The decoded macro block MP '(6, 3) is stored in the area corresponding to the identification number of the P picture in the picture storage unit 21. Next, when the encoding of the macroblock MP (6, 8) in the P picture is completed, the motion estimation unit 12 determines that the encoding of all macroblocks is completed, and returns the process to step S1. At this point, the picture storage unit 21 stores the image data of the I picture and P picture that have been coded once (by predictive coding, DCT processing, quantization, etc.) and decoded.

Returning to FIG. 7, in step S1, the picture rearranging unit 11 inputs the B picture (N
+1) is set as the target picture. Next, in step S2, the motion estimation unit 12 divides the target picture into macroblocks of 16 × 16 dots. Here, as shown in FIGS. 3 to 6, the motion estimation unit 12 divides the B picture (n + 1) into a matrix having 8 rows in the vertical direction and 6 columns in the horizontal direction, that is, 48 macro blocks ( Macro blocks MB (1,1) to MB (6,
8)).

Then, in step S3, the motion estimation section 12 determines whether or not the divided target picture is an I picture. Here, the motion estimation unit 12
It is determined that this target picture is not an I picture,
The process proceeds to step S4. In step S4,
The motion estimation unit 12 determines whether the divided target picture is a P picture. Here, when it is determined that the target picture is not a P picture, the motion estimation unit 12 determines that the target picture is a B picture, and advances the encoding process to step S5.

Next, in step S5, as shown in FIG. 6, the motion estimation unit 12 performs the coding process for the target picture T on all macroblocks MB as in the case of the I and P pictures already described. Is performed in the order of arrows, for example. That is, the motion estimation unit 12
Is the macroblock MB in the upper left corner (first row, first column)
After starting the encoding process from (1, 1), the process proceeds to the right, and the macro block MB in the upper right corner (first row, eighth column)
It ends once with (1, 8). Then, the motion estimation unit 12
Immediately moves to the second row, and the macroblock MB (2,1) in the second row and the first column to the macroblock MB in the second row and the eighth column
Proceed to (2,8). Hereinafter, similarly, the motion estimation unit 1
2 repeats the same movement for the third to sixth rows, and ends at the macroblock MB (6,8) in the sixth row and eighth column.

The motion estimator 12 encodes B pictures, for example, I and P pictures stored in the picture accumulator 21 and B pictures that have been encoded (currently encoding itself. ) Is referred to as a reference macroblock, and a similar macroblock that is similar to the macroblock to be newly encoded is detected. That is, the motion estimator 12 encodes the macroblock MB (1, 1) using the I and B picture macroblocks accumulated in the picture accumulator 21 as reference macroblocks.

Then, as shown in FIG. 5, the motion estimator 12 starts the coding from the macro block MB (1,1), the coding progresses, and the coding of the macro block MB (2,3) is carried out. When performing, the macroblock MP (2,3), the macroblocks MI ′ (1,1) to MI ′ (6,8) of the I picture stored in the picture storage unit 21, and the macroblock MP ′ of the P picture. (1,1) to MP '(6,8), B
Picture macroblocks MB (1,1) to MB (2,2
2) The difference data with each is calculated to obtain macroblocks MI ′ (5,3) as macroblocks MI ′ (1,1) to M.
I '(6,8), macroblocks MP' (1,1) to MP '
(6, 8), macroblocks MB (1, 1) to MB (2,
In 2), when it is determined that the difference data with the macro block MB (2, 3) is the smallest and exceeds the predetermined similarity (the difference data is less than or equal to the set value), the macro block MI ′ ( 5, 3) are extracted as similar macroblocks.

Therefore, the motion estimation unit 12 encodes the macro block MB (2,3), but since the macro block MI '(5,3) is a similar macro block, the inter-frame forward prediction encoding is performed. A mode signal of a coding mode indicating that (previous picture, that is, a macro block in an I picture is encoded as a similar block) is output to the motion prediction unit 20, and the macro block MB
It outputs (2, 3) to the subtractor 13. At this time, the motion estimation unit 12 uses the motion vector indicating the relative position of the mask block MI ′ (5,3) corresponding to the position of the mask block MB (2,3) as the motion prediction unit 20 and the multiplexing unit 22. Output to.

Then, the motion predicting section 20 has the motion estimating section 1
Since the mode signal indicating the inter-frame forward predictive coding is input from 2, the macro block MB (2,3) corresponding to this mode signal is input to the subtractor 13,
The subtraction command is output to the subtractor 13, and the macro block MI ′ used for subtraction from the picture storage unit 21.
(5, 3) is read and output to the subtractor 13. At this time, the subtractor 13 receives the subtraction command and the data of the macroblock MI ′ (5,3) used for the subtraction from the motion prediction unit 20, so that the input macroblock MP (2,3) is input.
Macroblock MI '(5, 3) is subtracted from 3), and the subtraction result is used as residual prediction data MS (2, 3) to obtain D
Output to the CT unit 14.

Next, the DCT unit 14 outputs the residual prediction data MS (2,3) of this macroblock MB (2,3) to 4
It is divided into blocks of × 4 dots, DCT processing is performed on each block, and the processing result is set to macroblock MD (2,
3) to the quantizer 15. Then, the quantizer 15 quantizes the input macroblock MSD (2,3) and outputs the processing result as the macroblock MSQ (2,3). As a result, the variable length coding 16 performs variable length coding on the macro block MSQ (2,3) and outputs it as the macro block MSV (2,3). Next, the multiplexing unit 22 sets the mode signal of the coding mode indicating that the macroblock MSV (2,3) is the interframe forward predictive coding (that the picture to be coded is a B picture). (Including the identification signal indicating the type of picture), the mask block MB (2, 3) and the mask block MI ′.
It is multiplexed with the motion vector indicating the relative position of (5, 3) and output to the buffer 23.

Then, the buffer 23 outputs a bitstream in which the macroblock MSV (2,3), the mode signal and the motion vector are multiplexed. In addition, the inverse quantizer 1
7 is a macro block MSQ output from the quantizer 15.
Inverse quantization of (2,3) is performed and macroblock MSD '
(2, 3) is reproduced and output to the IDCT unit 18. Then, the IDCT section 18 receives the input macroblock MS
Inverse discrete cosine transform is performed on D ′ (2,3) to reproduce the residual prediction signal MS ′ (2,3), which is output to the adder 19.

At this time, since the prediction residual signal MS '(2, 3) corresponds to the inter-frame forward predictive-coded macroblock MB (2, 3), the adder 19
The motion prediction unit 20 outputs an addition command, and the motion prediction unit 2
0 is added from the macroblock MI ′ (5,3) read from the picture storage unit 21 and the prediction difference signal MS ′ (2,3) to decode the macroblock MB ′ (2,3). It Then, the adder 19 stores the decoded macro block MB ′ (2,3) in the area corresponding to the identification number of the B picture in the picture storage unit 21.

Then, the motion estimation unit 12 is further stored in the picture storage unit 21 and the macro block MB (5,3) when the coding is further advanced and the macro block MB (5,3) is coded. I picture macro blocks MI '(1,1) to MI' (6,8), P picture macro blocks MP '(1,1) to MP' (6,8), B picture macro block MB The difference data between (1, 1) to MB (5, 2) is calculated to obtain the macro block MB '.
(2, 6) is the macroblock MI '(1, 1) to MI'
(6, 8), macroblocks MP '(1, 1) to MP'
(6, 8), B picture macroblock MB (1,
Of 1) to MB (5,2), the macroblock MB (5,5)
When it is determined that the difference data with 3) is the smallest and exceeds the predetermined similarity (the difference data is less than or equal to the set value), the macro block MB '(2,6) is extracted as a similarity macro block. .

Therefore, the motion estimation unit 12 encodes the macro block MB (5,3), but since the macro block MB '(2,6) is a similar macro block, the same intra-frame prediction encoding ( A mode signal of a coding mode indicating that a picture currently being coded, that is, a macro block in a B picture is coded as a similar block is output to the motion prediction unit 20, and
The macro block MB (5,3) is output to the subtractor 13. At this time, the motion estimation unit 12 determines that the mask block MB
Mask block MB 'corresponding to the position of (5, 3)
The motion vector indicating the relative position of (2, 6) is output to the motion prediction unit 20 and the multiplexing unit 22.

Then, the motion predicting section 20 includes the motion estimating section 1
Since the mode signal indicating the same intra prediction coding is input from 2, the macro block MB corresponding to this mode signal is input.
When (5, 3) is input to the subtractor 13, the subtracter 1
3, a subtraction command is output, and the macro block MB ′ (2,6) used for subtraction from the picture storage unit 21.
Is read out and output to the subtracter 13. At this time, the subtractor 13 receives the subtraction command and the data of the macro block MB ′ (2,6) used for the subtraction from the motion prediction unit 20, so that the macro block MB (5,3) to the macro block is input. MB '(2,6) is subtracted, and the subtraction result is output to the DCT unit 14 as residual prediction data MS (5,3).

Next, the DCT unit 14 outputs the residual prediction data MS (5,3) of this macroblock MB (5,3) to 4
It is divided into blocks of × 4 dots, DCT processing is performed on each block, and the processing result is set to macroblock MD (5,
3) to the quantizer 15. Then, the quantizer 15 quantizes the input macroblock MSD (5,3) and outputs the processing result as the macroblock MSQ (5,3). As a result, the variable length coding 16 performs variable length coding on the macro block MSQ (5,3) and outputs it as the macro block MSV (5,3). Next, the multiplexing unit 22 encodes the macroblock MSV (5, 3) into a mode signal of an encoding mode indicating that the macroblock MSV (5, 3) is in the same frame predictive encoding (indicating that the image to be encoded is a B picture). , And a mask block MB (5, 3) and a mask block MB ′.
It is multiplexed with the motion vector indicating the relative position of (2, 6) and output to the buffer 23.

Then, the buffer 23 outputs a bitstream in which the macroblock MSV (5,3), the mode signal and the motion vector are multiplexed. In addition, the inverse quantizer 1
7 is a macro block MSQ output from the quantizer 15.
Inverse quantization of (5, 3) is performed and macroblock MSD '
(5, 3) is reproduced and output to the IDCT unit 18. Then, the IDCT section 18 receives the input macroblock MS
The inverse discrete cosine transform is performed on D ′ (5,3) to reproduce the residual prediction signal MS ′ (5,3), which is output to the adder 19.

At this time, since the prediction residual signal MS ′ (5,3) corresponds to the macroblock MB (5,3) that has been predictively coded in the same frame, the adder 19 performs the motion prediction. The addition command is output from the unit 20, and the motion prediction unit 20
The macroblock M read from the picture storage unit 21
B '(2,6) and the prediction difference signal MS' (5,3) are added to decode the macroblock MB '(5,3). Then, the adder 19 stores the decoded macroblock MB ′ (5,3) in the area corresponding to the identification number of the B picture in the picture storage unit 21.

Then, the motion estimation unit 12 is further encoded, and when the macroblock MB (6, 6) shown in FIG. 5 is encoded, the macroblock MB (6, 6),
Macroblocks MI ′ (1,1) to MI ′ (6,8) of I pictures and macroblocks MP ′ (1,1) to MP ′ (6,8) of P pictures stored in the picture storage unit 21. ，
B picture macroblocks MB (1, 1) to MB (6,
5) The difference data with each is calculated to obtain macroblocks MP '(5,3) as macroblocks MI' (1,1) to M '.
I '(6,8), macroblocks MP' (1,1) to MP '
(6, 8), macro blocks MB (1, 1) to MB (6,
In 5), when it is determined that the difference data with the macro block MB (6, 6) is the smallest and exceeds the predetermined similarity (the difference data is less than or equal to the set value), the macro block MP ′ ( 5, 3) are extracted as similar macroblocks.

Therefore, the motion estimation unit 12 encodes the macro block MB (6, 6), but since the macro block MP '(5, 3) is a similar macro block, the inter-frame backward prediction encoding is performed. A mode signal of a coding mode indicating that (a future picture, that is, a macro block in a P picture is encoded as a similar block) is output to the motion prediction unit 20 and the macro block MB
(6, 6) is output to the subtractor 13. At this time, the motion estimation unit 12 uses the motion vector indicating the relative position of the macro block MP ′ (5,3) corresponding to the position of the mask block MB (6,6) as the motion prediction unit 20 and the multiplexing unit 22. Output to.

Then, the motion predicting section 20 includes the motion estimating section 1
Since the mode signal indicating the inter-frame backward prediction coding is input from 2, the macro block MB (6, 6) corresponding to this mode signal is input to the subtractor 13,
The subtraction command is output to the subtractor 13, and the macroblock MP ′ used for subtraction from the picture storage unit 21.
(5, 3) is read and output to the subtractor 13. At this time, the subtractor 13 receives the subtraction command and the data of the macro block MP '(5, 3) used for the subtraction from the motion prediction unit 20, and thus the input macro block MB (6,
The macro block MP '(5, 3) is subtracted from 6), and the subtraction result is used as residual prediction data MS (6, 6) to obtain D
Output to the CT unit 14.

Next, the DCT unit 14 outputs the residual prediction data MS (6,6) of this macroblock MB (6,6) to 4
It is divided into blocks of × 4 dots, DCT processing of each block is performed, and the processing result is set to the macro block MSD (6,
6) to the quantizer 15. Then, the quantization unit 15 quantizes the input macroblock MSD (6, 6) and outputs the processing result as the macroblock MSQ (6, 6). As a result, the variable length coding 16 performs variable length coding on the macroblock MSQ (6, 6) and outputs it as the macroblock MSV (6, 6). Next, the multiplexing unit 22 encodes the macroblock MSV (6, 6) into a mode signal in a coding mode indicating that the macroblock MSV (6, 6) is backward predictive coding between frames (that the picture to be coded is a B picture. (Including an identification signal indicating the type of picture), a mask block MB (6, 6) and a mask block MP '.
It is multiplexed with the motion vector indicating the relative position of (5, 3) and output to the buffer 23.

Then, the buffer 23 outputs a bit stream in which the macro block MSV (6, 6), the mode signal and the motion vector are multiplexed. In addition, the inverse quantizer 1
7 is a macro block MSQ output from the quantizer 15.
Inverse quantization of (6, 6) is performed and macroblock MSD '
(6, 6) is reproduced and output to the IDCT unit 18. Then, the IDCT section 18 receives the input macroblock MS
The inverse discrete cosine transform is performed on D ′ (6,6) to reproduce the residual prediction signal MS ′ (6,6), which is output to the adder 19.

At this time, since the prediction residual signal MS '(6, 6) corresponds to the macroblock MB (6, 6) which has been subjected to the inter-frame backward prediction coding, the adder 19
The motion prediction unit 20 outputs an addition command, and the motion prediction unit 2
0 is added from the macro block MP ′ (5,3) read from the picture storage unit 21 and the prediction difference signal MS ′ (6,6) to decode the macro block MB ′ (6,6). It Then, the adder 19 stores the decoded macroblock MB '(6, 6) in the area corresponding to the identification number of the B picture in the picture storage unit 21.

Then, the motion estimation unit 12 is further stored in the macro block MB (6,8) and the picture storage unit 21 when the coding is further advanced and the macro block MB (6,8) is coded. I picture macro blocks MI '(1,1) to MI' (6,8), P picture macro blocks MP '(1,1) to MP' (6,8), B picture macro block MB The macro block MI 'is calculated by calculating the difference data with each of' (1, 1) to MB '(6, 7).
(5, 6) is the macroblock MI '(1, 1) to MI'
(6, 8), macroblocks MP '(1, 1) to MP'
(6, 8), macroblock MB 'of B picture (1,
Macro block MB among 1) to MB '(6, 7)
It is determined that the difference data with (6, 8) is the smallest.

Next, when the motion estimation unit 12 determines that the difference data does not exceed the predetermined similarity (the difference data exceeds the set value), the code indicating that intra coding is to be performed. The mode signal of the conversion mode is output to the motion prediction unit 20. The following encoding processing is the same as the processing of the macro block MI (1,1) and the macro block MP (6,3) already described, and thus the description thereof is omitted. Finally, the buffer 23 stores the macroblock MSV
(6, 8), A bitstream in which mode signals are multiplexed is output. Then, the adder 19 outputs the decoded macroblock MB ′ (6,8) to the picture storage unit 2
It is stored in the area corresponding to the identification number of B picture 1.

Next, the motion estimation section 12 determines the B picture (n
When the encoding of the macro block MB (6, 8) in +1) is completed, it is determined that the encoding of all macro blocks is completed, and the process is returned to step S1. At this point, the picture storage unit 21 is once coded (predictive coding, DC
Image data of I picture (n), P picture (n + 3), and B picture (n + 1) in a decoded state after being subjected to T processing, quantization, etc. is stored.

Then, the motion estimator 12 returns to step 1 in order to perform image coding of the next B picture (n + 2), and performs the above B picture (n + 1) as shown in FIG.
The process of each step of the flowchart of (1) is performed on the B picture (n + 2). That is, FIG. 7 described above
Image encoding processing based on the flowchart of {n, n
+3, n + 1, n + 2, n + 6, n + 4, n + 5} are repeated for each frame image input.

<Decoder> Next, the configuration of the decoder shown in FIG. 8 will be described. For example, in the decoder 31, the bit stream is input in the order of encoding, that is, I picture {n}, P picture {n + 3}, B picture {n + 1, n + 2}, ... ,
It will be decrypted. The buffer 41 receives the multiplexed bitstream generated by the encoder 1 and output to the transmission path, and outputs the received bitstream to the separation unit 42. The separation unit 42 separates the input bit stream into a macro block MSV, a motion vector, and a mode signal, and outputs the encoded macro block data, the macro block MSV, to the variable length decoding unit 43. . Further, the separation unit 42 outputs the motion vector and the mode signal (identification number indicating the type of each coded picture) to the motion prediction unit 49.

At this time, the separation unit 42 extracts the quantization step size included in the bit stream and uses it in the inverse quantization performed by the inverse quantization unit 44. Output the quantization step size. Then, the variable-length decoding unit 43 performs variable-length decoding which is the reverse process of the variable-length coding used in the above-described variable-length coding unit 15, and inputs the macroblock M.
The SV is decoded and is output as a macroblock MSQ as a decoding result. The inverse quantization unit 44 performs an inverse process on the macroblock MSQ output from the variable length decoding unit 43 with respect to the quantization process used by the quantization unit 15 in the encoder 1, that is, the above process. Inverse quantization processing is performed using the quantization step size, and as a result of inverse quantization of the input macroblock MSQ, it is output as a macroblock MSD.

The IDCT unit 45 applies the DC of the encoder 1 to the macroblock MSD output from the inverse quantization unit 44.
The process opposite to that used in the T unit 14 is performed, the inverse cosine transform process of the macroblock MSD is performed, and the macroblock MSD is input (input in the encoded order).
For each of them, an adder 46
Output to. The adder 46, based on the mode signal (coding mode and the identification number indicating the top of the picture) and the macroblock address input from the separation unit 41, performs IDC
The residual prediction data Ms input from the T unit 45 and the already decoded I picture, P picture, or B picture stored in the picture accumulating unit 48 are added, and the input residual The image corresponding to the difference prediction data MS is decoded.

Further, the adder 46 stores the decoded image (I picture, P picture and B picture) in the picture accumulating section 48 and outputs it to the motion predicting section 49 and the picture rearranging section 47. . The picture rearrangement unit 47 collects the decoded macroblocks of each picture, which are input from the adder 46, on a picture-by-picture basis, reproduces the images of the pictures, and sequentially stores the images of the pictures. Accumulate to 48.

Then, the picture rearrangement unit 47
In P units, I picture {n}, P picture {n +
3}, B pictures {n + 1, n + 2}, ... Input pictures in the order I picture {n}, B picture {n
+1, n + 2} and P pictures {n + 3} are rearranged in this order, that is, the picture rearrangement unit 1 in the encoder 1
1 is rearranged in the order of the image signals input and output as an output image to an external device.

Next, the decoding in the motion prediction unit 49 will be described. The motion prediction unit 49 includes a picture type (I picture, B picture, P picture), motion vector, and intra coding that is decoded by a mode signal, intra-frame predictive coding, inter-frame forward predictive coding, and frame. The decoding process corresponding to the image to be decoded is performed based on either the inter-reverse prediction coding or the identification number indicating the type of the picture to be decoded. For example, the motion prediction unit 49 receives the residual prediction data MS from the IDCT unit 45.
(4, 7) is input, the mode signal indicates the same intraframe predictive coding, and the macroblock MI shown in FIG.
When (4, 7) (macroblock address) is decoded, the macroblock MI ′ already composited and stored internally from the identification signal and the motion vector indicating the type of the composite picture included in the mode signal. The image of (2, 4) is read and output to the adder 46 together with the addition signal.

As a result, the adder 46 adds the macro block MI '(2, 4) to the input residual prediction data MS (4, 7) and outputs the addition result as the macro block MI'. (4, 7) is output to the motion prediction unit 49. Then, when the coding of MI ′ (6,8) is completed, the motion prediction unit 49 assembles each decoded macroblock MI, reproduces the I picture {n}, and rearranges the I pictures. The image is output to the unit 47 and stored in the picture storage unit 48.

Next, the motion prediction section 49 has the IDCT section 45.
Residual prediction data MS (1, 6) is input from the mode signal, and the mode signal indicates the inter-frame forward prediction coding.
In the case of decoding MP (1, 6) shown in (1), since the identification signal indicating the type of picture included in the mode signal is a P picture, the I already stored in the picture storage unit 48 is already combined. The picture image is read out, and based on the read out I picture and motion vector, I
Picture (reference image) macroblock MI '(5, 6)
Is output as a reference macroblock to the adder 46 together with the addition signal.

As a result, the adder 46 adds the macro block MI '(5, 6) to the input residual prediction data MS (1, 6) and outputs the addition result as the macro block MP'. (1,6) is output to the motion prediction unit 49. Then, when the residual prediction data Ms (6, 3) is input from the IDCT unit 45 and the mode signal indicates intra coding, the motion prediction unit 49 receives the residual prediction data M.
Recognizing that S (6,3) has been encoded without a reference image, the macroblock MP '(6,3) is decoded only from this residual prediction data MS (6,3).

At this time, the motion prediction unit 49 uses the adder 46
The macro block of the reference image and the addition signal are not output. Then, when the encoding of MP '(6,8) is completed, the motion prediction unit 49 assembles each decoded macroblock MP, reproduces the P picture {n + 3}, and rearranges this P picture in picture order. While outputting to the section 47,
The picture is stored in the picture storage unit 48.

Next, the motion prediction section 49 has the IDCT section 45.
Residual prediction data MS (6, 6) is input from the mode signal, and the mode signal indicates the inter-frame backward prediction coding.
When decoding MB (6, 6) shown in (1), P is already combined and stored in the picture accumulating unit 48 based on the fact that the identification signal indicating the type of picture included in the mode signal is a B picture. An image of a picture (reference image) is read, and based on the read P picture and motion vector, macroblock MP '(5, 3) of the P picture
Is output as a reference macro vector to the adder 46 together with the addition signal.

As a result, the adder 46 adds the macro block MP '(5, 3) to the residual prediction data MS (6, 6) that is input, and outputs the addition result to the macro block MB'. (6, 6) is output to the motion prediction unit 49. Then, when the encoding of MP '(6,8) is completed, the motion prediction unit 49 assembles each decoded macroblock MP, reproduces the P picture {n + 3}, and rearranges the P pictures in the picture rearrangement. While outputting to the section 47,
The picture is stored in the picture storage unit 48.

As described above, the motion prediction unit 49
The image of each picture is decoded from the residual cosine transformed residual prediction signal Ms based on the motion vector and the prediction coding method and reference image indicated by the mode signal. The coding and decoding processes other than the intra-frame predictive coding in the same coding mode are the same as those in MPEG2.

As described above, in the moving picture coding method according to the embodiment of the present invention, an I picture, P picture or B picture to be coded is set as a target picture, and the target picture is set to a plurality of macroblocks. Divide into MB. Then, the macroblocks MB are subjected to coding processing in a predetermined order to generate coding information of the target picture T.

Therefore, when the moving picture coding apparatus of the present invention codes the image of the target picture, if the target picture is an I picture, it is already in the same frame (I picture being coded). The encoded macroblock is used as a reference block for encoding, and if the target picture is a P picture, the past I picture (reference image) and the same frame (encoding P picture) If a macroblock that has already been encoded is used as a reference block for encoding, and if the target picture is a B picture, not only the past I picture or future P picture is used as a reference image, but also the same frame (encoding B-pictures that are being processed) refer to already-encoded macroblocks Since it is used for encoding, the amount of information of the image referred to in the encoding process of each picture increases, the probability that macro blocks with similarities can be searched for increases, and the encoding efficiency increases. In addition, since the reference macroblock (or the reference image) having high similarity can be extracted, the image after decoding can be brought into a state closer to the image before encoding. In the above-described embodiment of the present invention, the procedures executed by each of the moving picture encoder and the decoder are recorded in a computer-readable recording medium, and the program recorded in this recording medium is read into the computer system. Then, the user attribute registration information automatic updating apparatus of the present invention is realized by executing the above. The computer system mentioned here includes an OS and hardware such as peripheral devices.

Furthermore, "computer system" is a WW
If the W system is used, the homepage providing environment (or display environment) is also included. The "computer-readable recording medium" means a flexible disk, a magneto-optical disk, a ROM, a CD-R.
A portable medium such as an OM or a storage device such as a hard disk built in a computer system. Further, the "computer-readable recording medium" means a volatile memory (RA) inside a computer system which is a system or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line.
As in M), the one holding the program for a certain period of time is also included.

The above program may be transmitted from a computer system that stores the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. Further, the program may be a program for realizing some of the functions described above. further,
It may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

Although one embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and design changes within the scope not departing from the gist of the present invention. Even so, it is included in the present invention.

[0106]

According to the moving picture coding apparatus of the present invention, when a picture of a target picture is coded, not only a past I picture or a future P picture is used as a reference picture but also the target picture itself is used. Since the already-encoded macroblock is used as the reference image, the amount of information to be referred to in the encoding process increases, and the encoding efficiency of the moving image can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of an encoder 1 according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram showing a moving image coding method according to an embodiment of the present invention.

FIG. 3 is a conceptual diagram showing an I-picture encoding method in the moving image encoding method according to the embodiment of the present invention.

FIG. 4 is a conceptual diagram showing a P picture encoding method in the moving image encoding method according to the embodiment of the present invention.

FIG. 5 is a conceptual diagram showing a B picture encoding method in the moving image encoding method according to the embodiment of the present invention.

FIG. 6 is a conceptual diagram showing an example of a method of dividing a target picture into macroblocks in the moving picture coding method according to the embodiment of the present invention. .

FIG. 7 is a flowchart showing each step of the moving picture coding method according to the embodiment of the present invention.

FIG. 8 is a functional block diagram showing a configuration example of a decoder used in the moving picture coding method according to the embodiment of the present invention.

FIG. 9 is a conceptual diagram showing a conventional moving image encoding method.

[Explanation of symbols]

1 encoder 2 decoder 11,47 Picture rearranging unit 12 Motion estimation section 13 Subtraction unit 14 DCT section 15 Quantizer 16 Variable length coding unit 17,44 Dequantizer 18,45 IDCT section 19,46 adder 20,49 Motion prediction unit 21,48 picture storage 22 Multiplexer 23,41 buffer 42 Separation part 43 Variable Length Decoding Unit

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Ryuji Tokunaga             4-201-508 Azuma, Tsukuba, Ibaraki (72) Inventor Miyuki Kawamura             1-5-1-5 Kamifenjaku, Mitaka City, Tokyo               Sedge Ltd. F-term (reference) 5C059 MA00 MA04 MA05 MA14 MA23                       MC11 MC38 ME01 NN21 PP05                       PP06 PP07 PP16 UA02 UA05                       UA33 UA38                 5J064 AA02 AA03 AA05 BA09 BA16                       BB01 BB03 BC01 BC08 BC14

Claims (4)

[Claims] 1. A moving picture coding using an I picture for intra coding, a B picture for interframe forward predictive coding and an interframe backward predictive coding, and a P picture for interframe forward predictive coding. When encoding an I picture, a B picture, and a P picture, which is a device, in addition to the interframe forward predictive coding, the interframe backward predictive coding, and the intra coding, the coding in the same frame is performed. An encoding device characterized by using the completed macroblock as a reference image for predictive encoding. 2. A moving picture coding using an I picture for intra coding, a B picture for interframe forward predictive coding and an interframe backward predictive coding, and a P picture for interframe forward predictive coding. In the process of encoding I-picture, B-picture and P-picture, in addition to the process of inter-frame forward predictive coding, inter-frame backward predictive coding and intra-coding,
The encoded macroblock in the same frame is
An encoding method comprising the step of using as a reference image for predictive encoding. 3. An I picture for performing intra coding, a B picture for performing interframe forward predictive coding and interframe backward predictive coding, and a P picture for performing interframe forward predictive coding are used. 1 is a moving picture coding program for operating the coding apparatus according to 1. In addition to the encoding process,
The encoded macroblock in the same frame is
An encoding program for causing a computer to execute an encoding process including a process used as a reference image for predictive encoding. 4. An I picture for performing intra coding, a B picture for performing interframe forward predictive coding and interframe backward predictive coding, and a P picture for performing interframe forward predictive coding are used. 1 is a recording medium in which a moving picture coding program for operating the coding apparatus described in 1 is recorded. In a process of coding an I picture, a B picture, and a P picture, interframe forward predictive coding and interframe reverse coding are performed. In addition to directional predictive coding and intra coding,
The encoded macroblock in the same frame is
A computer-readable recording medium in which a moving picture coding program for causing a computer to execute a coding process including a process used as a reference image for predictive coding is recorded.