JPH0678298A - Motion compensation prediction device for interlace animation - Google Patents

Abstract

PURPOSE:To improve coding efficiency and the picture quality by implementing field motion compensation using a motion vector per interlace block in the motion compensation prediction used for coding an interlace animation. CONSTITUTION:An input pattern 10 and a reference pattern 11 are respectively inputted to the same parity motion detector 12, a vicinity field motion detector 13 and an inter-field interpolation motion detector 14 respectively at every block, the motion is searched by one motion vector per block, predicted error signals E1, E2, E3 outputted from the motion detectors 12, 13, 14 are inputted to a comparator 15, where which motion detection mode is employed is decided and a selected mode ZM is outputted. A selector 16 selects any of motion vectors V1, V2, V3 based on the selected mode and outputs the selected vector as a ZV.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion compensation prediction device for coding an interlaced moving image. The present invention reduces the amount of information in motion compensation prediction in block units in an image transmission device such as a high-efficiency coding device for interlaced moving images and a storage device. Therefore, by using one motion vector per block, The amount of information is reduced, and the coding efficiency is improved by selecting either the same parity field motion compensation screen, the near field motion compensation screen, or the inter-field interpolation motion compensation screen.

[0002]

2. Description of the Related Art Conventionally, video communication such as video conference and C
In a high-efficiency encoding system for moving images for the purpose of storing moving images in a D-ROM or the like, in a frame or field screen, each screen is divided into blocks of, for example, 16 pixels × 16 lines, and in-plane coding is performed. High-efficiency coding is performed using inter-frame coding that codes the difference between the reference screen and the current screen by encoding or motion compensation.

FIG. 1 shows the configuration of a general encoder. Here, 71 is a subtracter, which obtains a difference between the input screen X1 and the prediction screen X2 to generate a prediction error screen X3. 7
2 is an encoder such as a discrete cosine transform (DCT) or a vector quantizer, 73 is a quantizer, 74 is an inverse quantizer, 7
Decoder 5 is an inverse discrete cosine transform (IDCT) or inverse vector quantizer. Reference numeral 76 denotes an adder, which is the prediction error screen X5 and the prediction screen X restored by the decoder 75.
2 is added to generate the local decryption screen X6. The local decryption screen X6 is used as a reference screen. As the reference screen, it is possible to use the uncoded original screen, that is, the screens before and after the input screen X1, instead of the local decoding screen X6.

The frame memory 77 has a local decoding screen X6.
And the input screen X1 is stored. The motion detector 78 detects motion in block units. The corresponding input block data 10 and the reference block data 11 of the area where the motion is searched are input from the frame memory 77 to the motion detection unit 78, and after the motion detection, the motion vector ZV and the selection flag ZM are output. The motion compensator 79 creates and outputs the prediction screen X2 from the reference block data 11 using the motion vector ZV obtained by the motion detection unit 78 and the selection flag ZM.

The output of the quantizer 73 is a variable length encoder 80.
And is multiplexed with the motion vector ZV obtained by the motion detector 78 and the selection flag ZM by the multiplexer 81, and output as the encoded output.

FIG. 2 shows an example of the configuration of the conventional motion detecting section 78. The frame motion detector 84 detects the motion of the frame block and outputs the prediction error signal ER and the motion vector VR. On the other hand, the field motion detector 85 detects the motion of the field block and outputs the prediction error signal EF and the motion vector VF. The prediction error signals ER and EF are compared by the comparator 87. Comparator 8
7 outputs to the selector 86 a selection flag ZM for selecting the smaller of the prediction error signals ER and EF. Selector 86
Operates in response to this, selects the motion detector having the smaller prediction error signal, and outputs the motion vector ZV from the selected motion detector. Thus, conventionally, the prediction error signal is obtained from the frame motion detector and the field motion detector, and motion compensation is performed using the motion vector of the motion detector having the smaller prediction error signal to improve the coding efficiency. ing.

[0007]

The motion detecting section 7 described above.
In the coding apparatus using 8, the motion detection is performed on the frame screen and the field screen, and the motion detector having the smaller prediction error is selected to perform the motion compensation. This conventional encoder has the following problems.

(1) When an image moves in an accelerating manner, the screen to be referred to differs between even lines and odd lines. Therefore, when the frame motion detector 84 is selected, the prediction error at the time of motion compensation becomes large, and the coding efficiency decreases.

(2) When the screen moves at a constant speed, the amount of motion in each field is almost the same. Therefore, when the field motion detector 85 is selected,
As compared with motion compensation on a frame screen, information about the amount of motion vectors is doubled, and as a result, coding efficiency is reduced.

Therefore, an object of the present invention is to improve the coding efficiency and image quality, which are the drawbacks of the above-mentioned conventional method.

[0011]

The feature of the present invention is that, for each block of the input screen and the reference screen, the same vector is used for each field between the field blocks of the input screen and the reference screen having the same parity. Detect motion,
The same parity field motion detecting means for obtaining a motion vector and for obtaining the sum of the respective field prediction errors from the motion detection, and for each block, in each field between the field blocks of the reference screen which is closest in time to the input screen. On the other hand, the same field is used to detect the motion, the motion vector is calculated, and the near field motion detection means for calculating the sum of the respective field prediction errors from the motion detection, and two field screens of the reference screen for each block. Motion detection using the same vector for each field between the combined screen and the input screen, the motion vector is obtained, and the sum of the respective field prediction errors is obtained from the motion detection. It is characterized by having means.

[0012]

According to the present invention, when motion detection is performed on the input screen in units of field blocks with respect to the reference screen, motion detection is performed between fields of the same parity, and a reference field closest in time to the input screen is obtained. Motion detection is performed between the two reference fields, the motion detection is performed between the two reference fields, and the three prediction error signals after the motion detection are compared, and the motion compensation form is determined based on the comparison result. Then, the motion vector is selected, and the selection flag and the motion vector are output.

According to the present invention, the motion vector information amount is reduced by using one motion vector per block, and the motion detection described above is adaptively used to detect the motion of only the conventional frame signal and the motion detection of only the field signal. It is possible to prevent a decrease in coding efficiency in, and to improve image quality and reduce the amount of transmitted information.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the block diagram of FIG. The motion detector of FIG. 3 is used as the motion detector 78 of the encoding device of FIG.

FIG. 3 shows the structure of an embodiment of the present invention.
0 is input block data and 11 is reference block data, both of which are interlaced screens. The block data 10 and 11 are subjected to motion detection by the same parity field motion detector 12, and the motion vector V1 and the block prediction error E1 are output. Further, the near field motion detector 13 also detects the motion, and the motion vector V2
And the prediction error E2 is output. Further, the inter-field interpolation motion detector 14 also performs motion detection, and outputs the motion vector V3 and the prediction error E3.

Identical parity field motion detector 12,
The prediction errors E1, E2 and E3 output from the near field motion detector 13 and the inter-field interpolation motion detector 14 are compared by a comparator 15. The comparator 15 selects the smallest one of the prediction errors E1, E2 and E3 and outputs a selection flag ZM which is a signal indicating which prediction error is selected. Selector 16 has selection flag ZM
The motion vector is selected in accordance with the above, and the motion vector from the motion detector that has output the smallest prediction error is output as the motion vector ZV.

After the motion vector ZV is selected by the selector 16 based on the selection flag ZM, the motion compensator 79 of FIG. 1 performs motion compensation using the reference block corresponding to the motion vector as a prediction signal. At this time, the value of the motion vector is used as it is for the luminance signal. On the other hand, for the color signal, as will be apparent from the description below, since the block size is half the luminance in the horizontal direction, the horizontal motion vector is used in half, and the luminance signal and the chromaticity signal are used. A prediction signal is created from

Incidentally, supplementing the outline of the processing on the decoder side (not shown) for receiving and decoding the encoded output output from the multiplexer 81, the decoder is sent from the encoder. Based on the type of motion detection and the amount of motion vector, the reference block is searched for, motion compensation is performed, and a prediction signal is created.

A specific example of the structure of the main part of this embodiment will be described in detail below. First, the configuration of the block data of the input screen will be described in detail with reference to FIG. Regarding the size of the input block, the luminance signal is 16 pixels × 16 lines, and the two color difference signals are each 8 pixels × 16 lines.
Lines are collectively called a macro block.
A series of encoding processes is performed for each macroblock.

As shown in FIG. 4, the block is composed of odd field data (O) existing in odd lines and even field data (□) existing in even lines. The frame block is
The field block includes data in which odd lines and even lines alternate, and the field block includes data in which only odd line data is collected or data in which even line data is collected.

For the reference block used for motion detection, only the luminance signal is used. The block size changes depending on the search range. For example, if the search range is ± 7 pixels in the main and sub-scanning directions, it will be 30 pixels × 30 lines. The size of the prediction error data obtained after motion compensation is the same as that of the input block in FIG. 5 for both the luminance signal and the color difference signal.

FIG. 5A shows the block data of the input screen and the block data of the reference screen in the vertical direction and the time axis direction. The data (◯) in the odd field f1 exists in the odd line, and the even field f2
Data (□) exist on even lines.

Next, a specific example of the same parity field motion detector 12 will be described with reference to FIG. The same parity field means that it is an odd field or an even field. The input block data 10 and reference block data 11 are input block memory 20 and reference block memory 21, respectively.
Is once stored in. According to the address from the address generating circuit 26, each field data of the area where motion detection is performed is extracted from the input block memory 20 and each field data of the search area is extracted from the reference block memory 21, and the first and second field prediction error calculation circuits 23 are extracted. , 24 calculates the error.

In this case, for an input block signal of 16 pixels × 16 lines, the same vectors MV1 and MV2 are used for the odd field f1 and the even field f2 of the input screen as shown in FIG. A prediction error signal is calculated between the input data of 1 and the reference data of the odd field, and the input data of the even field and the reference data of the even field. This prediction error signal can be obtained by a cumulative sum of absolute differences or a cumulative sum of squared differences. In the first and second field prediction error calculation circuits 23 and 24, the prediction error signal is obtained between the same parity fields and using the same vector as described above.

The optimum vector determination circuit 25 is composed of the first and second
The sum of the two prediction error signals obtained by the field prediction error calculation circuits 23 and 24 is calculated and stored as the total prediction error. The optimum vector determination circuit 25 searches all the search points by the address generation circuit 26, then compares the total prediction error at each search point to find the position where the error value is the minimum, and calculates the motion vector V1 and the prediction error. Outputs E1.

In this embodiment, furthermore, it is possible to improve the accuracy of motion detection by performing motion detection on the interpolated pixel block with decimal point accuracy. As shown in FIG. 5B, with respect to the original pixel signals A, B or C, D, for example, an interpolation signal such as p or q is created between upper and lower lines in the same field with 1/2 pixel precision. It can be an interpolated pixel block. In this case, as the interpolation signal between the upper and lower lines, p = (A + B) / 2, q is created by a method of creating odd lines between odd lines and even lines between even lines.
= (C + D) / 2, ... These calculations can be rounded to improve accuracy. In addition, it is possible to further improve the motion detection accuracy by making the interpolation between the upper and lower lines further 1/4 pixel accuracy. In this case, s = (3 × A + B) / for the points s, t, u and v as shown in FIG.
Interpolation pixels are added as in 4. Note that the interpolation signal between pixels in the horizontal direction is up to 1/2 pixel precision, and the interpolation pixel is created by averaging the left and right pixels.

In the case of FIG. 6, the intra-field interpolation circuit 22 creates interpolated pixels between adjacent lines and adjacent pixels in each field to obtain an interpolated pixel block with decimal point precision. FIG. 8B shows a state in which interpolation pixels p and q are created between adjacent lines, a prediction error signal is obtained using the same vectors MV1 and MV2, and motion prediction is performed.

Next, the internal structure of the near field motion detector 13 will be described with reference to the block diagram of FIG. Input block signal 10 and reference block signal 11
Are temporarily stored in the input block memory 20 and the reference block memory 21, respectively. According to the address from the address generation circuit 26, the respective field data of the area in which the motion detection is performed from the input block memory 20 and the data of the search target area in the field at the position temporally close to the input frame from the reference block memory 21 are stored. take out,
The error is calculated in the first and second field prediction error calculation circuits 23 and 24.

In this case, the input block signal of 16 pixels × 16 lines is parallel to both the input data of the odd field f1 and the even field f2 of the input screen as shown in FIG. 9A. Vectors MV1 and MV
2 using the reference data of the near field (FIG. 9A).
Then, the prediction error signal is obtained between the field f2).
Then, the optimum vector determination circuit 25 obtains the sum of the two prediction errors obtained by the first and second field prediction error calculation circuits and stores it as the total prediction error. The total prediction error at each search point is compared to find the position where the error value is the minimum, and the motion vector V2 and the prediction error E2 are output.
As the prediction error signal, the cumulative sum of absolute differences or the cumulative sum of squared differences can be used.

As the motion vector for each field, one of the vectors is used as a basis, and the other vector uses a value converted by a temporal distance ratio. For example,
In FIG. 9A, MV1 is used as a basic vector, and as MV2, a vector obtained by converting the basic vector MV1 by a temporal distance ratio k is used. That is, MV2 = k × MV1
And

In FIG. 9A, since the distance ratio is 1: 2, k = 2. Since the distance ratio is uniquely determined according to the reference screen, only one of the motion vectors is encoded.

In order to improve the accuracy of motion detection,
Motion detection can be performed on the interpolated pixel block with decimal point precision. In this case, the inter-field interpolation circuit 2
In step 2, interpolation pixels are created between adjacent lines and between adjacent pixels in the neighboring field to form an interpolation pixel block. FIG. 9B shows a state in which interpolation pixels v and q are created between adjacent lines and motion prediction is performed.

Next, the internal structure of the inter-field interpolation motion detector 14 will be described with reference to FIG. The input block signal 10 and reference block signal 11 are input block memory 20 and reference block memory 21, respectively.
Is once stored in. According to the address from the address generation circuit 26, each field data of the area where the motion is detected from the input block memory 20 and the reference block memory 2
Each field data in the search area is extracted from 1, and the reference field data is synthesized by the field synthesizing circuit 27 between the fields. Then, the first and second field prediction error calculating circuits are used as prediction data. Two
The prediction error is calculated at 3 and 24.

In this case, the block signal is 16 pixels × 16
As shown in FIG. 11A, for the input block signal of the line, the reference pixel w, x obtained by combining the data of both the odd field and the even field in the reference screen and the odd field and the even field of the input screen are combined. In between, a prediction error signal is obtained using parallel vectors. Then, the optimum vector determination circuit 25 obtains the sum of the two prediction errors obtained by the first and second field prediction error calculation circuits and stores it as the total prediction error. The total prediction error at each search point is compared to find the position where the error value is the minimum, and the motion vector V3 and the prediction error E3 are output. For synthesizing the even field and odd field data of the reference screen, a simple average or an average of data weighted according to a temporal distance can be used. In this case, the motion vector can use the same parity, and the prediction error signal can use the cumulative sum of absolute differences or the cumulative sum of squared differences.

As a motion vector for each field, one of the vectors is used as a basis, and the other vector is a value converted by a temporal distance ratio. For example,
In FIG. 11A, the basic vector is set between f1 on the input screen and f1 on the reference screen, and this is designated as MV1. The reference data required for f1 of the input screen is f of the reference screen by MV1.
The combination of the data of 1 and the data of f2 at the position where MV1 is projected on f2 of the reference screen is used.

If the vector between f2 on the input screen and f1 on the reference screen is MV2, MV2 is MV2.
A value obtained by converting 1 into a temporal distance ratio k is used. That is, MV2 = k × MV1. In FIG. 11A, since the distance ratio is 2: 3, k = 3/2. The reference data required for f2 of the input screen is a combination of the data of f1 of the reference screen by MV2 and the data of f2 at the position where MV2 is mapped to f2 of the reference screen.

Since the distance ratio is uniquely determined according to the reference screen, only one of the motion vectors is coded. In addition, in order to improve the motion detection accuracy, it is possible to perform the motion detection for the interpolation pixel block with the decimal point accuracy. In this case, the inter-field interpolation circuit 22 creates an interpolated pixel between adjacent lines and adjacent pixels in each field to form an interpolated pixel block. FIG. 11B shows a state in which interpolation pixels y and z are created between adjacent lines and between adjacent pixels to perform motion prediction.

Various modifications are possible in carrying out the present invention. For example, the block size is 16 pixels ×
The size is not limited to 16 lines, and various sizes such as 32 pixels × 32 lines can be applied. For the color signal block, for example, in the case of 8 pixels × 8 lines, both the horizontal direction motion vector and the vertical direction vector obtained by the motion detection may be halved to create the prediction signal.

Next, a second embodiment of the present invention will be described. In this embodiment, as shown in FIG. 12, the same parity field motion detector 12 and the near field motion detector 13 are used.
The inter-field interpolation motion detector 14 is combined with the frame motion detector 17 and the field motion detector 18. In this embodiment, each of the detectors 1
The prediction error values E1, E2, E3, ER and EF detected at 2, 13, 14, 17 and 18 are input to the comparator 15. The comparator 15 selects the smallest prediction error value from these prediction error values and outputs the selected plug ZM. The selector 16 determines the motion vectors V1, V2, V of the motion detector having the smallest prediction error value based on the selection flag ZM.
3, VR or VF is selected and output as ZV.

Here, the internal structure of the frame motion detector 17 will be described with reference to FIG. Here, the input block data 10 and the reference block data 11 are the input block memory 20 and the reference block memory 21, respectively.
Is once stored in. In accordance with the address from the address generation circuit 26, the frame data of the area in which motion detection is performed is extracted from the input block memory 20 and the frame data of the area for search is extracted from the reference block memory 21, and the error is calculated by the frame prediction error calculation circuit 29. .

In this case, as shown in FIG. 14A, in the frame screen F composed of two fields (O and ● are odd field data and □ and ■ are even field data) alternately arranged, the input screen is displayed. Prediction error signal between the frame and the frame of the reference screen is obtained. The optimum vector determination circuit 25 compares the prediction error signal at each search point obtained by searching the reference screen with the address from the address generation circuit 26, obtains the position where this error is the minimum, and calculates the motion vector VR. And the prediction error ER is output.

In order to improve the accuracy of motion detection,
Motion detection can be performed on the interpolated pixel block with decimal point precision. In this case, the intra-frame interpolation circuit 28
Then, interpolation pixels are created between adjacent lines and between adjacent pixels in the same frame to form an interpolation pixel block with decimal point precision. FIG. 14B shows a state in which interpolation pixels are created between adjacent lines and motion prediction is performed.

Next, the internal structure of the field motion detector 17 will be described with reference to FIG. Here, the input block data 10 and the reference block data 11 are the input block memory 20 and the reference block memory 21, respectively.
Is once stored in. According to the address from the address generation circuit 26, the field data of the area where the motion detection is performed and the field data of the search area are retrieved from the reference block memory 21 from the input block memory 20,
The field prediction error calculation circuit 23 and the second field prediction error calculation circuit 24 calculate the prediction error of each field.

In this case, as shown in FIG. 16A, different vectors (MV1, MV2) are used for the odd field f1 and the even field f2 of the input screen, and the input data of the odd field is the odd field or the even field. A prediction error signal is obtained between the reference data and the reference data of the even field or the odd field of the input data of the even field. The optimum vector determination circuit 25 finds the sum of the two prediction errors obtained by the first and second field prediction error calculation circuits and stores it as a total predicted value. The total prediction error at each search point is compared to find the position where the error value is the minimum, and the motion vector VF and the prediction error EF for the input field are output.

In order to improve the accuracy of motion detection,
Motion detection can be performed on the interpolated pixel block with decimal point precision. In this case, the inter-field interpolation circuit 2
In step 2, interpolation pixels are created between adjacent lines and adjacent pixels in each field to form decimal point precision interpolation pixel blocks. FIG. 16B shows a state in which interpolation pixels are created between adjacent lines and motion prediction is performed.

According to this embodiment, the frame motion detector 1
Since 7 and the field motion detector 18 are added, it is possible to further improve the prediction accuracy as compared with the first embodiment. However, in this case, an increase in the selection flag ZM of the motion detection method and an increase in calculation time are expected.

Next, a third embodiment of the present invention will be described. In this embodiment, the same parity field motion detector 12, the near field motion detector 13 and the inter-field interpolation motion detector 14 are provided with the inverse parity field motion detector and the far field detector as shown in FIGS. It is a combination of field motion detection.

In this case, contrary to the same parity field motion detection shown in FIG. 7, the reverse parity field motion detection shown in FIG. 17 uses the same vector for the odd field and the even field of the input screen to input the input data of each odd field. Is for detecting the motion between the reference data of the even field and the input data of the even field for the reference data of the odd field.

Further, in the far field motion detection in FIG. 18, contrary to the near field motion detection in FIG. 9, the same vector is used in the odd field and the even field of the input screen to input data of the odd field and the even field. On the other hand, the motion detection is performed with respect to the reference data (field f1 in FIG. 18) of the field far away in time.

Further, in the present invention, the temporal distance or position between the reference screen and the input screen is arbitrary. Eg 6
It is possible to refer to a reference screen that is apart from the frame or to refer to a reference screen that is located later in time. In the latter case, motion compensation in the opposite direction in time is performed.

The number of reference screens is also free. For example, an optimum prediction screen is obtained by using the present invention for each of a plurality of reference screens that are temporally positioned before and after the input screen. Then, a new screen that is synthesized by applying a temporal filter to each optimum prediction screen can be used as the prediction screen.

[0052]

As described above, the present invention adaptively uses the same parity field motion detection, the near field motion detection, and the inter-field interpolation motion detection, which causes a problem in the motion compensation of the conventional frame signal only. It is possible to detect motion with high accuracy even for an image of accelerated motion or fast motion. Further, in the present invention, since only one vector is required per block in any case, it is possible to reduce the vector amount load, which has been a problem in the conventional motion compensation of only the field signal, to half. It is possible to improve and reduce the amount of transmitted information. As an example of the effect, ISO
Test moving image (Flower Garden, Bicy
cle) in the CCIR601 image format,
The image quality (S / N ratio) at a bit rate of 4 Mbit / s is improved by 0.5 to 1.0 dB as compared with the method that adaptively uses frame motion compensation and field motion compensation, and the amount of motion vector information transmission is 30 to It was possible to reduce by 60%.

[Brief description of drawings]

FIG. 1 is a block diagram of a conventional motion compensation prediction apparatus for interlaced moving images.

FIG. 2 is a block diagram of a conventional motion detection unit.

FIG. 3 is a block diagram of a motion detecting unit according to an embodiment of the present invention.

FIG. 4 is an explanatory diagram showing an example of a configuration of block data on an input screen.

FIG. 5 is a configuration diagram of screen data showing a part of the block data of the input screen and the block data of the reference screen in a vertical direction and a time axis direction.

FIG. 6 is a block diagram showing a specific example of the same parity motion detector.

FIG. 7 is an explanatory diagram of an operation principle of the same parity motion detector.

FIG. 8 is a block diagram showing a specific example of a near field motion detector.

FIG. 9 is an explanatory diagram of an operation principle of a near field motion detector.

FIG. 10 is a block diagram showing a specific example of an inter-field interpolation motion detector.

FIG. 11 is an explanatory diagram of an operation principle of an inter-field interpolation motion detector.

FIG. 12 is a block diagram of a motion detecting unit according to a second embodiment of the present invention.

FIG. 13 is a block diagram showing a specific example of a frame motion detector.

FIG. 14 is an explanatory diagram of the operation principle of the frame motion detector.

FIG. 15 is a block diagram showing a specific example of a field motion detector.

FIG. 16 is an explanatory diagram of an operation principle of the field motion detector.

FIG. 17 is an explanatory diagram of the operating principle of the inverse parity field motion detector used in the third embodiment of the present invention.

FIG. 18 is an explanatory diagram of an operation principle of the far field motion detector.

[Explanation of symbols]

 10 Input Screen 11 Reference Screen 12 Same Parity Field Motion Detector 13 Near Field Motion Detector 14 Interfield Interpolation Motion Detector 15 Comparator 16 Selector 71 Subtractor 72 Encoder 73 Quantizer 74 Inverse Quantizer 75 Decoding 76 Adder 77 Frame Memory 78 Motion Detection Unit 79 Motion Compensator 80 Variable Length Encoder 81 Multiplexer 84 Frame Motion Detector 85 Field Motion Detector 86 Selector 87 Comparator

Claims (7)

[Claims] 1. A motion compensation prediction apparatus for interlaced moving images, which performs motion compensation in block units using an input screen and a reference screen, and means for storing the input screen and the reference screen decomposed in field block units, and for each block. In addition, between the field blocks of the input screen and the reference screen having the same parity, motion is detected using the same vector for each field, the motion vector is obtained, and the total of the field prediction errors from the motion detection is calculated. The same parity field motion detection means to be obtained, and for each block, the motion is detected by using the same vector for each field between the field blocks of the reference screen that is closest in time to the input screen, and the motion vector is obtained. , Near-field motion detection that obtains the sum of the respective field prediction errors from the motion detection For each block, a motion is detected by using the same vector for each field between the output unit and the screen in which the two field screens of the reference screen are combined, and the input screen, and the motion vector is obtained. Inter-field interpolation motion detection means for obtaining the total of each field prediction error from the detection, comparing the prediction error output from each of the motion detection means,
A selection flag indicating the smallest prediction error and a means for selecting a motion vector from the motion detection means that has output the smallest prediction error, and which motion detection means is included in the encoded prediction error. A motion compensation prediction apparatus for an interlaced moving image, characterized in that a selection flag indicating whether or not an output from the above is selected and a motion vector corresponding to the selection flag are added and transmitted. 2. The motion compensation prediction apparatus for an interlaced moving image according to claim 1, further comprising a frame for obtaining a prediction error signal between the frame screen of the reference screen and the frame screen of the input screen for each block. A motion-compensated prediction apparatus for interlaced video, characterized by adding motion detection means. 3. The motion compensation prediction apparatus for interlaced video according to claim 1, further comprising a field for obtaining a prediction error signal between the field screen of the reference screen and the field screen of the input screen for each block. A motion-compensated prediction apparatus for interlaced video, characterized by adding motion detection means. 4. The motion-compensated prediction apparatus for interlaced video according to claim 1, further, using the same vector for the odd field and the even field of the input screen for each block, input data of the odd field respectively. Is added to the reference data of the even field, and the input data of the even field is added with the inverse parity field motion detection means for detecting the motion of the reference data of the odd field. Motion compensation prediction device. 5. The motion-compensated prediction apparatus for interlaced moving pictures according to claim 1, further comprising using the same vector for the odd field and the even field of the input screen for each of the blocks so that either the odd field or the even field is used. A motion-compensated prediction apparatus for interlaced moving pictures, characterized in that a far-field motion detection means for detecting motion between time-distance field reference data is also added to the input data. 6. The motion-compensated prediction device for interlaced moving images according to claim 1, further comprising: a unit for creating an interpolated pixel block with decimal point precision for a block of a reference screen, A motion-compensated prediction device for interlaced video, characterized in that detection accuracy is improved. 7. The motion-compensated prediction apparatus for interlaced video according to claim 6, wherein, when performing motion detection, as a first step, motion detection with integer pixel precision is performed on a block of a reference screen. Then, the motion vector is obtained, and in the second step, the motion detection is performed using the interpolation pixel of the decimal point precision in the vicinity of the start position of the motion vector obtained in the first step to shorten the motion detection processing time. A motion-compensated prediction device for interlaced moving images, characterized in that