JPH0955940A - Motion compensation module - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize an unnecessary motion compensation module of an additional device for external memory, which is used for a moving picture encoding device, without increasing data transfer quantity and to reduce the scale. SOLUTION: An MV detection means 53 detects motion information of a block being an encoding object from present picture data of a present picture memory 55 and reference picture data of a reference picture memory 56. A predicted picture generation means 51 generates predicted picture data from motion information and reference picture data. A predicted error picture generation means 54 obtains prediced error picture data as the difference of predicted picture data and present picture data and outputs it to an encoding module 6. A reproduced picture generation means 54 generates reproduced picture data from predicted picture data and prediced error picture data which is locally decoded from the encoding module 6 and outputs it to a reproduced picture memory 58. The motion compensation module 5 is constituted by such a way, and all processings which need access with an external picture memory 7 are realized in the motion compensation module 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for constructing a motion compensation module for economically realizing a moving picture coding apparatus.

[0002]

2. Description of the Related Art A moving image is composed of a plurality of temporally continuous still images. In this specification, each still image is referred to as a frame.

FIG. 8 is a block diagram showing the flow of a moving image coding process using motion compensation. In moving image coding, a current image is a frame to be coded and other frames constituting the moving image are reference images, and the current image is divided into rectangular small areas (image blocks) (hereinafter, a small area is referred to as an MB. The following processing is performed in combination.

(1) Motion Detection Processing This is processing for evaluating from which position of the reference image the current image MB can be moved, and detecting the movement information (motion information, hereinafter referred to as MV).

(2) Prediction error image generation process The MB in the reference image corresponding to the current image MB obtained from the MV obtained in the motion detection process (1) and the reference image is set as the prediction image MB, and the current image MB This is a process of generating a prediction error MB that is a difference from the predicted image MB.

(3) Encoding process This is a process for performing the same two-dimensional DCT and quantization as for a still image on the prediction error MB, and encoding the quantized data and MV.

(4) Local Decoding Process This is a process of reconstructing the prediction error MB 'by inversely transforming the data subjected to DCT and quantization.

(5) Reproduced image generation process The present image MB is reconstructed by reconstructing the prediction error MB 'obtained in the local decoding process (4) and the prediction image MB used in the prediction error image generation process (2). This is a process of generating the image MB.

The reproduced image obtained after encoding the entire current image is used as a reference image when the next frame to be encoded is the current image. Also, motion detection processing (1)
And the prediction error image generation process (2) are collectively referred to as motion compensation. The reference image, the predicted image, and the reproduced image are stored in the image memory 100.

FIG. 9 shows a first structural example of a conventional moving picture coding apparatus. The feature of this configuration example is that each process of moving image coding is performed by a motion detecting module 1 including a motion detecting circuit 10, a current image memory 12, and a reference image memory 13, a prediction error image generating device 20, an encoding device 21, That is, the encoding module 2 is divided into a local decoding device 22, a reproduced image generation device 23, a current image memory 24, a predicted image memory 25, and a reproduced image memory 26 [Tetsya].
Matsumura et. al. : A Chip
Set Architecture for Prog
Ramble Real-Time MPEG2
Video Encoder, IEEE 1995 CU
STOM INTEGRATED CIRCUITS
CONFERENCE, pp. 393-396 (199
5)].

In this first configuration example, since only the MV needs to be transmitted between the motion detection module 1 and the coding module 2, the flexibility of the internal configuration of the motion detection module 1 is high. Further, in the motion detection module 1, only the motion detection processing means having a large processing amount is processed by the encoding module 2.
Since a plurality of processing means having a relatively small processing amount are mounted, the difference in hardware scale between both modules is small. In particular, the motion detection module 1 is effective in reducing the scale of hardware. However, on the other hand, since the encoding module 2 generates the prediction error MB, the reference image (in this case, the prediction image MB) needs to be input again from the external memory 4, and the data transfer amount with the external memory 4 is growing.
Further, since the external memory 4 is accessed by the motion detection module 1 and the encoding module 2, use of an external device (additional device 3) for coping with access conflict,
It is necessary to deal with this by using multiple external memories that store the same data.

FIG. 10 is a second structural example of a conventional moving picture coding apparatus. The feature of this configuration example is that each process of moving image coding is performed by a motion compensation module 1 including a motion detection device 10, a prediction error image generation device 11, a current image memory 12, and a reference image memory 13, an encoding device 21, Local decoding device 22, reproduced image generation device 23, prediction image memory 2
5, it is divided into an encoding module 2 composed of a reproduced image memory 26 [Naoya Hayashi]
et. al. : A Compact Motion E
stimulator with a Simplelife
d VectorSearch Strategic M
ainting Encoded Picture
Quality, IEEE 1995 CUSTOM
INTEGRATED CIRCUITS CONNECT
RENCE, pp. 409-412 (1995)].

In the second configuration example, the hardware scale of the motion compensation module 1 is larger than that in the first configuration example, but the external memory 4 is not accessed again in order to generate the prediction error MB. Therefore, an increase in the data transfer amount with the external memory 4 can be suppressed. Also, the motion compensation module 1
The amount of data transferred from the coding module 2 to the coding module 2 increases, but since there is no data transferred directly from the coding module 2 to the motion compensation module 1, the motion compensation module 1
The degree of freedom of the internal configuration of the does not decrease so much. However, since the reproduced image MB needs to be written from the encoding module 2 to the external memory 4, the external memory 4 is accessed by both the device 1 and the device 2, and therefore, as in the first configuration example,
It is necessary to add an external device (additional device 3).

[0014]

[Problems to be Solved by the Invention] As described above,
In the configuration example of the conventional moving image coding apparatus, motion detection processing,
The prediction error image generation processing and the reproduction image generation processing are distributed to a plurality of modules to reduce the size of the module that executes motion detection with a large amount of processing. However, since the external memory is accessed from a plurality of modules, it is necessary to deal with an increase in the amount of data transfer with the external memory and a conflict of access to the external memory. Therefore, it is necessary to multiplex the external memory and to add an external device that arbitrates the competition, which causes a problem that the circuit configuration becomes complicated as the whole moving image coding system.

Further, when the motion detection process, the prediction error image generation process, and the reproduced image generation process are integrated in the same module, the hardware scale becomes very large, and there is a problem that the manufacturing yield of the module decreases.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to increase the data transfer amount with an external memory in a moving picture coding device without increasing the data transfer amount.
The purpose of the present invention is to provide a motion compensation module that does not require an additional device to access an external memory and to reduce the hardware scale thereof.

[0017]

According to the present invention, a motion compensation module includes (means 1) means for inputting an encoding target image from outside and inputting a reference image from an external memory; and the input encoding target image. For the coding target block cut out from
Between the encoding target image and the input reference image, motion information detecting means for detecting motion information of the encoding target block, and a prediction image for generating a prediction image block of a movement source from the detected motion information. Generating means, prediction error image generating means for generating a prediction error image block from the encoding target block and the generated prediction image block, and means for outputting the generated prediction error image block to an external encoding means. And a prediction error image block input from the outside, a reproduction image generation unit that generates a reproduction image block from the prediction image block, and a unit that outputs the generated reproduction image block to the external memory. To do.

(Means 2) In the above, the current image memory for storing the target block to be encoded until generation of the prediction error image block and all the reference image data referred to by the motion information detecting means are generated in the prediction error image block generation. A reference image memory that stores until the end, a prediction image memory that stores the prediction image data used for the prediction error image block generation until the end of the reproduction image block generation process, a prediction image block in the prediction image memory, and an external source. A reproduced image memory for storing a reproduced image block generated from the input locally decoded prediction error image block is provided.

(Means 3) In the above, the predicted image generation means creates a predicted image block from the data in the reference image memory used in the motion information detection means, and the prediction error image generation means means the generated predicted image block. From the prediction error image block, and at the same time, the prediction image generating means writes the prediction image block in the prediction image memory.

(Means 4) In the above, by storing pixel data corresponding to the same position in the predicted image block and the reproduced image block in the same memory space, the predicted image memory and the reproduced image memory are shared. .

(Means 5) In the above, hardware is shared by performing prediction image block generation by the prediction image generation means and reproduction image block generation by the reproduction image generation means in time division using the same arithmetic unit. The configuration.

By doing so, the above object is achieved.

The above-mentioned means of the present invention operate as follows.

(Means 1) By implementing all the processing that requires access to the external memory in the motion compensation module, the access to the external memory can be controlled by the motion compensation module alone, so that Enables control of external memory without providing additional devices.

(Means 2) Prediction error MB generation and reproduced image M
In order to perform B generation, it is necessary to hold the predicted image MB in the reference image until the reproduction image MB generation is completed. This reference image uses the fact that a part of the data in the reference image memory used for MV detection is sufficient, and is used in the reference image memory as a prediction image MB to be reused in the prediction error MB generation and the reproduction image MB generation. The increase in internal memory is minimized by securing only the necessary data in the prediction image memory.
After securing the predicted image MB in the predicted image memory,
Since the current image memory and the reference image memory can be used for MV detection, the memory increase is only required for the predicted image memory and the reproduced image memory when compared with the second conventional configuration example.

(Means 3) Prediction error is generated by creating prediction image data from the data of the reference image memory used for the MV detection process and generating prediction error image MB, and at the same time writing the prediction image data in the prediction image memory. The required prediction image memory capacity is reduced by shortening the period for securing the prediction image MB in the internal memory before generating all the predicted image MB data in the prediction image memory before generating the image block. And reduce the hardware scale.

(Means 4) If the reproduced image MB is generated,
By utilizing the fact that the predicted image MB data is not necessary and that each pixel data of the predicted image MB and each pixel data of the reproduced image MB have a one-to-one correspondence, each pixel data of the predicted image and its pixel are used. The pixel data of the corresponding reproduced image is set to the reproduced image M.
By overwriting B data in the same memory as the corresponding predicted image MB data, the reproduced image memory and the predicted image memory provided in (Means 2) are shared, and the hardware scale is reduced.

(Means 5) The subtraction of the prediction error image MB generation process and the addition of the reproduced image MB generation process are both realized by an adder / subtractor, and the amount of calculation in each process is larger than that in the detection of MV. Utilizing the few things, we will share the computing unit by time sharing and reduce the hardware scale. Due to this sharing, the increase in hardware compared to the conventional configuration example 2 is about 10 bits of registers, selectors, and address generators except for the memory, and the address generator also applies (Means 2). In this case, since the memory capacity (that is, the number of addresses) is reduced, downsizing is facilitated.

[0029]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing a basic configuration example of a moving picture coding apparatus using a motion compensation module according to the first embodiment of the present invention. The motion compensation module 5 according to the present embodiment is an example in which (Means 1) and (Means 2) of the present invention are applied, and the motion compensation module 5 is
Prediction image generation means 51, prediction error image generation means 52, M
V detection means 53, reproduced image generation means 54, current image memory 55, reference image memory 56, predicted image memory 57, reproduced image memory 58, external image memory control means 59, and is connected to the encoding module 6 and external memory 7. Has been done. The internal configuration of the encoding module 6 is arbitrary,
It has a coding device 61 and a decoding device 62, has a prediction error image input means 63 and a locally decoded data output means 64 as an interface of the motion compensation module 5, and does not have an interface with the external image memory 7.

The motion compensation module 5 stores the current image data input from the outside in the current image memory 55. The external image memory control means 5 is previously stored in the reference image memory 56.
The reference image data required for block matching is input from the external image memory 7 via the current image memory 55.
MV (motion information) detection processing is performed using the MV detection means 53 from the current image data in the reference image data and the reference image data in the reference image memory 56. Subsequently, the predicted image generation means 51
Reads the image data in the reference image memory 56 from the detected MV, generates a predicted image MB, and stores it in the predicted image memory 57. Next, the prediction error image generation means 52
However, the current image MB data is read from the current image memory 55, the predicted image MB data is read from the predicted image memory 57, the difference between the corresponding data is obtained, and the difference is output as prediction error image data to the encoding module 6. Next, the prediction error image data passes through the prediction error image input means 63, is encoded by the encoding device 61, is locally decoded by the local decoding device 62, and is locally decoded by the local decoding data output means 64 from the encoding module 6. The reproduced image MB is stored in the reproduced image memory 58 by using the reproduced image generating means 54 from the predicted error image data that has returned and the predicted image MB previously stored in the predicted image memory 57. Finally, the reproduced image MB stored in the reproduced image memory 58 is output to the external image memory 7 via the external image memory control means 59.

In the configuration example of the conventional moving picture coding apparatus shown in FIGS. 9 and 10, it is necessary to access the external memory 4 from both the motion compensation module 1 and the coding module 2, so that the additional apparatus 3 It was necessary to attach. However, by using the motion compensation module 5 of the present embodiment, the coding module 6 does not need to access the external image memory 7, and only the external image memory control means 59 included in the motion compensation module 5 can be used. It is possible to control access to the external image memory 7. Therefore, it is not necessary to attach an additional device around the external image memory 7. Further, the transfer data to and from the external image memory 7 are only the reference image data for MV detection and the finally created reproduced image data. Compared to the conventional first configuration example, these data do not need to access the predicted image again, so the data transfer amount is reduced, and even when compared to the conventional second configuration example, the external memory There is no increase in the amount of data transferred.

The capacity required for the predictive image memory 57 is determined by starting the transfer of the predictive error image data to the encoding module 6 and then passing through the predictive error image input means 63 to pass through the encoding device 6
The prediction error image data encoded by 1 and locally decoded by the local decoding device 62 is locally decoded data output means 64.
Depending on the degree of delay before returning from the encoding module 6. Normally, in the hardware used in the encoding module, this delay is about a processing period of 1 MB, and considering that pipeline processing is performed to handle a plurality of MBs at the same time, the prediction image memory 57 has 2 MB. It is enough if there is enough capacity to store minute data.
It is about 768 [W] × 8 [bit].

Further, the required capacity of the reproduced image memory 58 is sufficient if there is a capacity corresponding to an image of 1 MB (384 [W] × 8 [bit]).

FIG. 2 is a block diagram of a moving picture coding apparatus using the second embodiment of the motion compensation module according to the present invention, in which (Means 3) of the present invention is applied to the basic configuration of FIG. is there. The difference from the basic configuration of FIG. 1 is that the prediction error image generating means 52 generates prediction error image data before storing the output of the prediction image generating means 51 in the prediction image memory 57 and outputs it to the encoding module 6. That is the configuration.

In the case of the basic configuration of FIG. 1, 384 bytes are written to the predicted image memory 57 for writing predicted image data when generating predicted image data, and 384 bytes for reading predicted image data when generating predicted error image data. 384 by for reading predicted image data when generating image data
te, the total data transfer amount of 1152 bytes is 1 MB.
Required for each process.

On the other hand, in the second embodiment,
Since the output of the prediction image generation unit 51 is directly used by the prediction error image generation unit 52, it is not necessary to read out the prediction image data when generating the prediction error image data, and the prediction when generating the prediction error image data is performed per 1 MB process. A total of 768 bytes of 384 bytes for writing the image data and 384 bytes for reading the predicted image data at the time of the reproduced image generation data.
Processing can be performed by data transfer of e. As a result, there is a margin in the schedule for accessing the predicted image memory 57, and when using a memory having the same access speed as the memory used in the basic configuration, it is possible to easily share the memory. For example, when the predicted image memory 57 and the reproduced image memory 58 are shared based on the second embodiment, the amount of data transferred to the shared memory is 1 MB processing, and the reproduced image when the reproduced image data is generated. A total of 768 bytes of 384 bytes for writing data and 384 bytes for reading when outputting reproduced image data to an external memory.
te, and the total data transfer amount is 1536 bytes /
Become MB.

Generally, the operation frequency of the motion compensation module 5 has a processing cycle of about 2000 cycles / Mb, and data transfer of 1536 bytes / MB can be realized by a one-port memory operating at the operation frequency of the motion compensation module. Is. On the other hand, when sharing the same memory based on the basic configuration, the increase in the data transfer amount is
This is the same as when sharing is performed based on the second embodiment, and the total data transfer amount to the memory after sharing is 19
20 bytes / MB, 2000 cycles /
In order to perform processing within a processing cycle of about Mb, it is necessary to simultaneously improve the access capability of the memory itself, such as multiplexing the ports of the memory and raising the operating frequency of the memory part.

FIG. 3 is a moving picture coding using the third embodiment of the motion compensation module according to the present invention in which (Means 4) of the present invention is applied to the second embodiment of FIG. It is a block diagram of an apparatus. The third embodiment is the second embodiment.
The prediction image memory and the reproduction image memory described in the above embodiment are shared and the same memory is used. The memory that shares the predicted image memory and the reproduced image memory is the compensation image memory 81.

In the third embodiment, due to the function and effect of (Means 4) of the present invention, compared with the memory capacity before sharing,
The feature is that the memory capacity after sharing can be reduced. FIG.
Shows the processing schedule of the motion compensation module 5 and the usage state of the compensation image memory 81. The knitted portion in the figure shows the period in which the data needs to be stored in the memory. The data in the predicted image memory becomes unnecessary for the subsequent processing when it is read out when the reproduced image data is generated. Therefore, if the reproduced image data obtained by using the predicted image data is stored in the memory area from which the predicted image data is read, the reproduced image memory can be stored without increasing the memory area for storing the reproduced image data. Can achieve the same function as when there is. Therefore, it is possible to reduce the amount of hardware corresponding to the memory of 384 bytes used as the reproduced image memory.

The third embodiment shown in FIG. 3 is the second embodiment.
This is an example in which the prediction image memory having a sufficient access capacity and the reproduction image memory are shared according to the embodiment of the embodiment. However, if the memory to be used has a sufficient access capacity, the same method is used. The predicted image memory and the reproduced image memory of the basic configuration can be shared, and the memory capacity used can be reduced.

FIG. 5 is a moving picture coding using the fourth embodiment of the motion compensation module according to the present invention in which (Means 5) of the present invention is applied to the third embodiment of FIG. It is a block diagram of an apparatus. The difference from the third embodiment is that
The prediction error image generation means and the reproduction image generation means are shared to be the compensation image generation means 82.

Further, FIG. 6 shows a prediction error image generating means 52 and a reproduced image generating means 5 in the third embodiment.
4 shows the structure of the compensation image generating means 82 in the fourth embodiment.

As shown in FIG. 6, the prediction error image generating means 52 and the reproduced image generating means 53 have the same configuration except that the arithmetic unit is the adder 52a or the subtractor 54a.
The input registers 52b, 52c, 54b, 54c and the output registers 52d, 54d are basically the same as each other. Therefore, in the fourth embodiment, as shown in FIG. 7, the arithmetic unit is the adder / subtractor 82a, and the prediction image generation means 51 and the compensation image memory 51 for each input register 82b are controlled by the time division control means (not shown). Inputs from the current image memory 55 and the encoding module 6 to the input register 82c are input to the selectors 82e and 8 respectively.
By using the compensation image generation means 82 configured to switch the time-division using 2f and switch the adder / subtractor 82a in synchronization with the switching, the calculation unit used in the prediction error image generation and the reproduction image generation is changed. It can be shared and the amount of hardware can be reduced.

Further, as shown in FIG. 4, when the processing cycle of the prediction error image generation and the processing cycle of the reproduced image generation do not overlap with each other, the current image MB data in the current image memory, which is read at the time of generation of the prediction error image, , If the addresses are assigned to the predicted image MB data in the compensation image memory that is read out when the reproduced image is generated, in the order referred to in the predicted image generation process and the reproduced image generation process, the number of accessed data is exactly the same. Generation of the read address to the current image memory when generating the predicted image and generation of the read address to the compensation image memory when generating the reproduced image can be easily shared by using the same address generator in a time-sharing manner. . According to (Means 4) of the present invention, if the read address and the write address for the compensation image memory are the same, the read address generator for the current image memory and the write address generator for the compensation image memory may be shared. .

[0046]

As described above, according to the present invention, since it is not necessary for the encoding module to access the external image memory used in the moving image encoding system, it is possible to arbitrate access conflicts. No additional equipment is required.

Further, when only the necessary data in the reference image memory is secured in the predicted image memory as the predicted image MB to be reused in the prediction error MB generation and the reproduction image MB generation, in particular, the internal memory Can be kept to a minimum.

Predicted image data is created from the reference image memory data used in the MV detection process, and the predicted error image M
When the predicted image data is written in the predicted image memory at the same time as B is generated, the period for securing the predicted image MB in the internal memory can be shortened, and the required predicted image memory capacity can be reduced. It will be possible.

When the reproduced image MB is generated, the predicted image MB data becomes unnecessary, and the fact that each pixel data of the predicted image MB and each pixel data of the reproduced image MB have a one-to-one correspondence is used. Then, when the reproduced image MB data is overwritten in the same memory as the corresponding predicted image MB data, the reproduced image memory and the predicted image memory can be shared.

Furthermore, the subtraction of the prediction error image MB generation processing and the addition of the reproduction image MB generation processing are both realized by an adder / subtractor, and the amount of calculation in each processing is smaller than that in the detection of MV. In the case where the operation unit is shared by using time division, especially, the hardware scale can be further reduced, and the miniaturization is facilitated.

As described above, the moving picture coding system using the motion compensation module of the present invention does not require any components other than the motion compensation module, the coding module, and the external memory, and has a minimum number of modules. This is possible, and since the increase in the size of the motion compensation module can be suppressed to a low level, it has a great effect on the cost reduction of the moving picture coding apparatus.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a moving picture coding apparatus using a motion compensation module according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing the configuration of a moving picture coding apparatus using a motion compensation module according to a second embodiment of the present invention.

FIG. 3 is a block diagram showing a configuration of a moving picture coding apparatus using a motion compensation module according to a third embodiment of the present invention.

FIG. 4 is a diagram showing an example of a processing schedule of the image compensating means in the third embodiment.

FIG. 5 is a block diagram showing a configuration of a moving picture coding apparatus using a motion compensation module according to a fourth embodiment of the present invention.

FIG. 6 is a diagram showing a configuration of a prediction error image generating means and a reproduced image generating means in the third embodiment.

FIG. 7 is a diagram showing the configuration of a compensation image reproducing means in the fourth embodiment.

FIG. 8 is a block diagram showing the flow of a moving image coding process using motion compensation.

FIG. 9 is a block diagram of a moving picture coding apparatus showing a first configuration example of a conventional motion compensation module.

FIG. 10 is a block diagram of a moving picture coding apparatus showing a second configuration example of a conventional motion compensation module.

[Explanation of symbols]

 5 ... Motion compensation module 6 ... Encoding module 7 ... External image memory 51 ... Predicted image generation means 52 ... Prediction error image generation means 53 ... MV detection means 54 ... Reproduced image generation means 55 ... Current image memory 56 ... Reference image memory 57 Prediction image memory 58 ... Reproduction image memory 59 ... External image memory control means 61 ... Encoding device 62 ... Local decoding device 63 ... Prediction error image input means 64 ... Local decoded data output means 81 ... Compensation image memory 82 ... Compensation Image generation means

Claims (5)

[Claims] 1. A means for inputting an encoding target image from the outside and a reference image from an external memory; and an encoding target image for the encoding target block cut out from the input encoding target image. Motion information detecting means for detecting motion information of a block to be encoded between the input reference image, predictive image generating means for generating a predicted image block of a movement source from the detected motion information, and the code A prediction error image generating unit that generates a prediction error image block from the block to be encoded and the generated prediction image block; a unit that outputs the generated prediction error image block to an external encoding unit; A prediction error image block, a reproduction image generating unit that generates a reproduction image block from the prediction image block, and the generated reproduction image block. A motion compensation module, which further comprises: a means for outputting to the external memory. 2. A current image memory for storing an encoding target block until the end of generation of a prediction error image block, and all reference image data referred to by the motion information detecting means,
A reference image memory that stores until the end of the prediction error image block generation, a prediction image memory that stores the prediction image data used for the prediction error image block generation until the end of the reproduction image block generation process, and The motion compensation module according to claim 1, further comprising: a reproduced image memory that stores a reproduced image block generated from the predicted image block and a locally-decoded prediction error image block that is input from the outside. 3. The prediction image generation means creates a prediction image block from the data of the reference image memory used in the motion information detection means, and the prediction error image generation means calculates the prediction error image block from the generated prediction image block. The motion compensation module according to claim 2, wherein the prediction image generation means writes the prediction image block in the prediction image memory at the same time when the prediction image block is generated. 4. The motion compensation module according to claim 2, wherein pixel data corresponding to the same position in the predicted image block and the reproduced image block are stored in the same memory space. 5. The prediction image block generation by the prediction image generation means and the reproduction image block generation by the reproduction image generation means are time-divisionally performed using the same arithmetic unit. The motion compensation module according to any one of 1.