JP2914448B2 - Motion vector prediction encoding method and motion vector decoding method, prediction encoding device and decoding device, and recording medium recording motion vector prediction encoding program and decoding program - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector predictive encoding method and decoding method used for motion compensation inter-frame prediction in video coding, a predictive encoding apparatus and a decoding apparatus, and a motion vector predictive code. The present invention relates to a recording medium on which an encrypted program and a decryption program are recorded.

[0002]

2. Description of the Related Art Hitherto, in video coding, interframe prediction coding using a coded frame as a prediction signal has been known in order to reduce redundancy in the time direction. In order to increase the prediction efficiency of the temporal direction prediction, a motion-compensated inter-frame prediction method of performing prediction using a motion-compensated image signal as a prediction signal is used. The number and type of components of the motion vector used for motion compensation differ depending on the assumed motion model. For example, assuming a motion model of only translation, the motion vector is composed of a horizontal movement component and a vertical movement component. In addition, the motion vector of a motion model that takes into account enlargement and reduction in addition to translation is
It consists of three components, a horizontal / vertical movement component and a scaling component.

Normally, motion compensation is performed in units of divided areas such as small blocks obtained by dividing one screen into a plurality of screens. Therefore, each divided area has an individual motion vector. It is known that each of these motion vectors has a high correlation between neighboring regions in adjacent divided regions. Therefore, a method is used in which a motion vector of a coding target region is predicted from a motion vector of a region near the coding target region, and the prediction error is variable-length coded to reduce the redundancy of the motion vector.

[0004] The moving picture coding system ISO / IEC
In 11172-2 (MPEG-1), an encoding target image is divided into a plurality of small blocks, motion compensation is performed for each small block, and a small block to be encoded (hereinafter, referred to as an encoding target small block). The motion vector of the small block to be encoded is predicted from the motion vector of the encoded small block.

In the above-described MPEG-1, the only motion that can be compensated is translation. In a simple motion model having a small number of motion vector components such as MPEG-1, the motion cannot be compensated in some cases. Therefore, the prediction efficiency between frames can be improved by using a motion compensation method corresponding to a more complex motion model having a large number of motion vector components. However, if motion compensation of each small block is performed on a complex motion model using such a motion compensation method, the amount of codes generated when coding a motion vector increases.

Therefore, in such a case, in order to avoid an increase in the code amount, a motion compensation method that minimizes the prediction error of the small block to be encoded has been conventionally selected from a plurality of motion compensation methods. A method of encoding a motion vector is known. Here, as an example of such an encoding method, there are two motion compensation methods that respectively assume two different types of motion models, a translation model and a translation + enlargement / reduction model. An encoding method for selecting a compensation method and performing encoding will be described.

FIGS. 9A and 9B show examples of a parallel movement model and a parallel movement + enlargement / reduction movement model, respectively. The parallel movement model shown in FIG. 9A is a movement model that expresses the movement of a subject by a parallel movement component (x, y). Further, the parallel movement + enlargement / reduction movement model shown in FIG. 9B adds the parameter z indicating the amount of enlargement or reduction of the object as well as the parallel movement component (x, y). This is a motion model represented by (x, y, z). The parallel movement shown in FIG.
In the example of the enlargement / reduction motion model, the parameter z of the motion vector is a value representing reduction.

Therefore, the motion vector of the translation model is

(Equation 1) , And the motion vector of the parallel movement + enlargement / reduction model is

(Equation 2) It is represented by In each of the above equations, x is a horizontal component, y is a vertical component, and z is a scaling component. Also, the unit for performing motion compensation is a small block, and the motion compensation method is switched for each small block according to the prediction efficiency, and the prediction of the motion vector is performed from the motion vector of the encoded small block.

When the motion compensation method for the coded small block is the same as the motion compensation method for the small block to be coded, the prediction error of the motion vector is calculated by the following equation.

In the case of the parallel movement model, d1x, y = v1x, y (i) −v1x, y (i-1) (1) In the case of the parallel movement + enlargement / reduction model, d2x, y, z = V2x, y, z (i) -v2x, y, z (i-1) (2)

Here, v1x, y (i) and v2x, y, z (i)
Is the motion vector component of the encoding target small block, v1
x, y (i-1) and v2x, y, z (i-1) are components of the motion vector of the small block one frame before.

As described above, the prediction error d of the motion vector
x, y and dx, y, z are calculated and encoded, and transmitted to the decoding side.
Note that, even if the motion compensation method is different in the size of each small block or the like, the same motion vector predictive encoding is performed as long as the motion model is the same.

When the motion compensation method of the coded small block is different from the motion compensation method of the small block to be coded, and in the case of intra-frame coding, the prediction value of each component is set to 0, Each component of the motion vector of the small block to be converted is transmitted to the decoding side as it is.

According to such an encoding method, the redundancy of the motion vector in the motion compensation inter-frame predictive encoding can be reduced, and therefore, the generated code amount of the motion vector can be suppressed.

On the other hand, to decode a motion vector encoded by the above-described encoding method, a prediction error is extracted from the encoded data sequence and the motion vector of the decoded small block and the prediction error are added as shown in the following equation. To decode the motion vector of the small block to be decoded.

In the case of the translation model, v1x, y (i) = v1x, y (i-1) + d1x, y (3) In the case of the translation + enlargement / reduction model, v2x, y, z ( i) = v2x, y, z (i-1) + d2x, y, z (4)

Here, v1x, y (i) and v2x, y, z (i)
Is the motion vector component of the small block to be decoded, v1x, y
(I-1) and v2x, y, z (i-1) are motion vector components of the decoded small block.

[0018] Also, ISO / IEC 14496-2 is working on international standardization in January 1999.
The same motion compensation method is adopted in the (MPEG-4) verification model. MPEG-4 employs global motion compensation which predicts the motion of the entire screen due to panning, tilt and zoom of the camera (MPEG Vide
o Group, “MPEG-4 Video Verification Model Versio
n 7.0, “IS0 / IEC JTC1 / SC29 / WG11N1682, April, 199
7． ). Hereinafter, the configuration and processing flow of an encoder using global motion compensation will be briefly described with reference to FIG.

First, the encoding target image 31 is input to a global motion detector 34, where a global motion parameter 35 for the entire screen is obtained. MPEG-
In No. 4, projective transformation and affine transformation can be used for the motion model. Assuming that the coordinates of the current point are (x, y) and the coordinates of the corresponding point are (x ′, y ′), the projective transformation is represented by the following equations (5) and (6).

X ′ = (ax + by + tx) / (px + qy + s) (5) y ′ = (cx + dy + ty) / (px + qy + s) (6)

In general, the case where s = 1 is often called a projective transformation. The projective transformation is a general expression of two-dimensional transformation, and the following equations (7) and (8), where p = q = 0 and s = 1, are affine transformations.

X '= ax + by + tx (7) y' = cx + dy + ty (8)

In the above equations, tx and ty represent the horizontal and vertical translation amounts, respectively. The parameter a represents horizontal enlargement / reduction or inversion, and the parameter d
Represents vertical enlargement / reduction or inversion. Also,
Parameter b represents horizontal shear, and parameter c represents vertical shear. Further, a = cos θ, b = si
In the case of nθ, c = −sin θ, and d = cos θ, it represents rotation of the angle θ. The case where a = d = 1 and b = c = 0 is equivalent to a conventional parallel movement model.

As described above, by using the affine transformation for the motion model, various motions such as parallel movement, enlargement / reduction, inversion, shearing, rotation, and any combination thereof can be expressed. The projective transformation with eight or nine parameters can express more complicated movements and deformations.

The global motion parameter 35 obtained by the global motion detector 34 is input to the global motion compensator 36 together with the reference image 33 stored in the frame memory 32. The global motion compensator 36
A motion vector for each pixel obtained from the global motion parameter 35 is applied to the reference image 33 to generate a global motion compensation prediction image 37.

On the other hand, the reference image 33 stored in the frame memory 32 is input to the local motion detector 38 together with the input image 31. In the local motion detector 38, 16
For each macroblock of pixels × 16 lines, the input image 31
A motion vector 39 between the image and the reference image 33 is detected.
The local motion compensator 40 generates a local motion compensated prediction image 41 from the motion vector 39 for each macroblock and the reference image 33. This is the motion compensation method itself used in conventional MPEG and the like.

Next, the coding mode selector 42 selects, for each macroblock, one of the global motion compensation predicted image 37 and the local motion compensated predicted image 41 whose error from the input image 31 is small. In a macroblock for which global motion compensation has been selected, no local motion compensation is performed, and thus the motion vector 39 is not encoded. The predicted image 43 selected by the encoding mode selector 42 is subtracted by the subtractor 4
The difference image 45 between the input image 31 and the prediction image 43 is converted into a DCT coefficient 47 by the DCT unit 46.

Next, the DCT coefficient 47 is converted into a quantization index 49 by a quantizer 48. The quantization index 49 is a quantization index encoder 57, the encoding mode selection information 56 is an encoding mode encoder 58, the motion vector 39 is a motion vector encoder 59, and the global motion parameter 35 is a global motion parameter encoder. After being individually encoded at 60, they are multiplexed to provide encoder output.

Further, in order to obtain the same decoded image as that of the decoder in the encoder, the quantization index 49 is returned to the quantization representative value 51 by the inverse quantizer 50, and further the difference image is converted by the inverse DCT unit 52. It is converted back to 53. The difference image 53 and the prediction image 43 are added by an adder 54, and a local decoded image 55
Becomes The local decoded image 55 is stored in the frame memory 32 and used as a reference image when encoding the next frame.

Next, the operation of the corresponding MPEG-4 decoder will be described with reference to FIG. The multiplexed coded bit streams are separated and individually decoded. The quantization index decoder 61 stores the quantization index 49, the coding mode decoder 62 stores the coding mode selection information 56,
The motion vector decoder 63 decodes the motion vector 39, and the global motion parameter decoder 64 decodes the global motion parameter 35.

The reference image 33 stored in the frame memory 68 is input to the global motion compensator 69 together with the global motion parameter 35, and a global motion compensated image 37 is generated. In addition, the reference image 33 is input to the local motion compensator 70 together with the motion vector 39, and a local motion compensation prediction image 41 is generated. The encoding mode selection information 56 causes the switch 71 to operate, and outputs either the global motion compensation image 37 or the local motion compensation image 41 as the prediction image 43 for each macroblock.

On the other hand, the quantization index 49 is returned to the quantization representative value 51 by the inverse quantizer 65, and further the inverse DC
In the T unit 66, the image is inversely converted into the difference image 53. Difference image 53
And the predicted image 43 are added by the adder 67 to become the local decoded image 55. The local decoded image 55 is stored in the frame memory 6
8 is used as a reference image when encoding the next frame.

The above-described global motion compensation prediction method in MPEG-4 selects the smaller prediction error for each macroblock from the prediction image based on global motion compensation and the prediction image based on local motion compensation, and predicts the prediction efficiency of the entire frame. Is to increase. Also, in a macroblock for which global motion compensation has been selected, since a motion vector is not coded, the amount of code conventionally required for coding a motion vector can be reduced.

[0034]

Conventionally, in an encoding method for switching between a plurality of motion compensation methods having different motion models, prediction between motion vectors of different motion models is not performed. For example, when considering a coding method for switching between a motion compensation method assuming a parallel motion model and a motion compensation method assuming a parallel motion + enlargement / reduction motion model, the number of parameters in both motion vectors is considered. Because of the difference, it is not possible to predict the motion vector of the parallel motion model from the motion vector of the parallel motion + enlargement / reduction motion model by a simple difference.

However, since the motion models are different, it cannot be said that the motion vectors have no redundancy. FIG. 10 shows the correlation between the motion vector of the parallel movement model and the motion vector of the parallel movement + enlargement / reduction movement model.
Consider the motion vector shown in FIG. In FIG. 10, in performing motion compensation of the encoding target small blocks Boa and Bob, the encoding target small block Boa refers to the small block Bra included in the reference frame, and a motion compensation method assuming a parallel motion model. , And the encoding target small block Bob performs motion compensation by a motion compensation method assuming a parallel movement + enlargement / reduction motion model with reference to the small block Brb included in the reference frame.

In this case, the motion vector shown in FIG.

(Equation 3) Is a parallel movement model,

(Equation 4) Is a parallel movement + enlargement / reduction movement model. At this time, in the motion compensation of the encoding target small block Bob, the small block Brb included in the reference frame is enlarged and referred to. Therefore, the motion vector shown in FIG.

[Outside 1] In, the translation components have substantially the same value, indicating that there is redundancy.

However, in the conventional method, since a motion vector of a motion model different from the motion vector of a certain motion model is not predicted, the redundancy between the motion vectors of the motion models cannot be reduced. .

In the above-described MPEG-4, predictive coding is used to efficiently code a motion vector. For example, the motion vector encoder 59 shown in FIG.
The operation of is as follows. That is, as shown in FIG. 13, the vector MV of the current block is the motion vector MV1 of the left block, and the motion vector M
V2 and MV3 of the upper right diagonal block are referred to, and the median (median) of them is used as the predicted value. If the predicted value of the vector MV of the current left block is PMV, PM
V is defined by the following equation (7). PMV = median (MV1, MV2, MV3) (9)

However, when the reference block is in the intra-frame encoding mode, there is no motion vector, so the vector value at the corresponding position is set to 0 and the median value is obtained. Further, even when the reference block is predicted by global motion compensation, there is no motion vector, so that the median value is calculated by setting the vector value at the corresponding position to 0. For example, if the left block is local motion compensation, the block directly above is global motion compensation, and the upper right block is within a frame, MV2 =
MV3 = 0. When all three reference blocks are global motion compensated, MV1 = MV2 = MV3 = 0
And the median is also 0, so the predicted value is 0. In this case, the motion vector of the current block is equivalent to not performing predictive coding, and the coding efficiency is reduced.

In MPEG-4, seven ranges in Table 1 are defined as the range (range) of the magnitude of the local motion vector, and which range is used for each codeword in the bit stream called fcode. To the decoder.

[Table 1]

However, since the global motion parameters in MPEG-4 can take a wide range from -2048 to +2047.5, the motion vectors obtained from the global motion vectors can also take values from -2048 to +2047.5. However, the range of the local motion vector is smaller than that, and a case occurs in which the prediction deviates greatly. For example, if fcode = 3, the motion vector of the current block is (Vx, Vy) = (+ 48, +36.5), and the prediction vector obtained from the global motion vector is (Px
When (Vx, PVy) = (+ 102, +75), the prediction error is (MVDx, MVDy) = (− 54, −38.5), and its absolute value is larger than the original (Vx, Vy). The word length of the code word assigned to the prediction error (MVDx, MVDy) is shorter as the absolute value is smaller. Therefore, there is a disadvantage that the code amount is increased by the motion vector prediction.

Accordingly, an object of the present invention is to provide a motion vector prediction encoding method and a motion vector decoding method, a prediction encoding apparatus and a decoding apparatus, and a motion vector prediction method for suppressing the amount of generated motion vectors and improving the efficiency of motion vector prediction. It is an object of the present invention to provide a computer-readable recording medium on which a motion vector predictive encoding program and a motion vector decoding program are recorded.

[0043]

According to the present invention, there is provided a motion vector predictive coding method, a predictive coding apparatus, and a motion vector predictive coding program recorded on a computer-readable recording medium. Is divided into small blocks, and a motion vector of the small block to be encoded is encoded into small blocks in a motion vector prediction encoding method in which one motion compensation method can be selected from a plurality of motion compensation methods for each small block to be encoded. When predicting from the motion vector of a block, the motion compensation method of the encoding target small block is locally
Encoded small blocks used for prediction in the motion compensation method
If the motion compensation method is global motion compensation,
Local motion of coded small block from motion vector
Calculate the vector and calculate the motion vector of the encoding target small block.
The prediction error of the motion vector is obtained, and the prediction error of the motion vector is encoded.

A decoding method for decoding a motion vector encoded by the above-described motion vector predictive encoding method,
Decoder, and the decoding program of motion vectors recorded in a computer-readable recording medium is decoded pairs
The motion compensation method of the elephant small block is a local motion compensation method,
Motion compensation method for decoded small blocks used for prediction is global
Decoding from global motion vectors
Calculates and decodes local motion vector of already-completed small block
Find the motion vector prediction vector of the target small block ,
The motion vector is decoded by adding the prediction error to the prediction vector.

[0045]

[0046]

[0047]

The above-described motion vector predictive encoding method
Method, predictive coding device and computer readable
Prediction coding of motion vectors recorded on various recording media
In the gram, when local motion compensation predicts the motion vector of the selected small block to be encoded from the global motion parameters, if the size of the predicted vector exceeds a predetermined range, the predicted vector is Clip within specified range. Or
In predicting the motion vector of the block selected by the local motion compensation from the global motion parameter, if the magnitude of the predicted vector exceeds a predetermined range, the predicted vector is set to 0.

On the other hand, the above-described series of motion vector predictive coding methods, predictive coding apparatuses, and motion vector predictive coding programs for motion vectors recorded on a computer-readable recording medium are decoded. to process the decoding, decoding device, and, in the decoding program recorded in a computer-readable recording medium, decoding
The motion compensation method for the target small block is local motion compensation.
Therefore, the motion compensation method of the decoded small block used for prediction is
In the case of global motion compensation, from the global motion vector
Calculate the local motion vector of the decoded small block,
A prediction vector of a motion vector of a small block to be decoded is obtained, and a prediction error is added to the prediction vector to decode the motion vector.

[0050]

Then, the above-described motion vector coding or decoding method, predictive coding apparatus or decoding apparatus, or
When a motion vector predictive encoding program or decoding program converts a motion vector of an arbitrary motion model of a small block into a motion vector of a parallel motion model, a motion vector of a parallel motion model obtained for each pixel in the small block. , A representative motion vector of the small block is calculated. When calculating the above-described representative motion vector, a value (an average value, an intermediate value, a median value, a most frequent value) that is statistically calculated for each component of the motion vector of the parallel motion model for each pixel in the small block (Value, maximum value, minimum value) are the values of the components of the representative motion vector.

Further, at the time of decoding, if the predicted value of the local motion vector obtained from the global motion parameter exceeds a predetermined range, the predicted vector is clipped within the predetermined range. Alternatively, when the predicted value of the local motion vector obtained from the global motion parameter exceeds a predetermined range, the predicted vector is set to 0.

The motion vector prediction encoding method and decoding method of the present invention, the prediction encoding apparatus and the decoding apparatus, and
According to the motion vector predictive encoding and decoding programs, it is possible to predict a motion vector between motion vectors having different motion models, and to predict a motion vector between global motion compensation and local motion compensation. Since it is possible to perform this operation, it is possible to reduce the generated code amount of the motion vector.

When the prediction vector obtained from the global motion parameter exceeds the range of the local motion vector and clips to the minimum or maximum value, for example, fcode = 3 (the range of the motion vector is -64 to in Tasu63.5 pixels), motion vector of the current block is _{(V x, V y) =} (+ 48, + 36.5), the predicted vector obtained from the global motion parameters (PVx, PVy) = (+ 102, + 75) , The predicted vector is (PVx, PVy) = (+ 63.5, +63.
Clipped to 5). Prediction error (MVD _x , MVD _y )
= (-15.5, -27), which is (-5) of the conventional method.
4, -38.5). Since the smaller the absolute value of the code word assigned to the motion vector difference is, the shorter the word length is, the total code amount can be reduced.

In the case where the prediction vector is cleared to 0, for example, if fcode = 3 (the range of the motion vector is −64 to +63.5 pixels) and the motion vector of the current block is (V _x , V _y) ) = (+ 48, +36.
5) If the predicted vector obtained from the global motion parameters is (PVx, PVy) = (+ 102, +75), the predicted vector is cleared to (PVx, PVy) = (0, 0). The prediction error is (MVD _x , MVD _y ) = (+ 4
8, +36.5), which is (-15.
Although the absolute value is larger than (5, -27), the absolute value is smaller than (-54, -38.5) of the conventional method.

As still another example, if fcode = 1 (the range of the motion vector is −16 to +15.5 pixels), and the motion vector of the current block is (V _x , V _y ) = (+ 3, +
1.5), the predicted vector obtained from the global motion parameter is (PVx, PVy) = (+ 102, +75)
Consider the case The clip method, prediction vector (PVx, PVy) = (+ 15.5, + 15.5) , and the prediction error _{(MVD x, MVD y) =} (- 12.5, -
14). On the other hand, in the 0 clear method, the prediction vector becomes (PVx, PVy) = (0, 0), and the prediction error (M
_{VD x, MVD y) = (} + 3, a + 1.5). In this example, the absolute value can be smaller in the 0 clear method than in the clip method.

[0057]

Next, embodiments of the present invention will be described with reference to the drawings. In this embodiment, the global motion compensation method (GMC) and the local motion compensation method (LMC) can be switched for each small block. The motion model of the global motion compensation method is a motion model in consideration of parallel movement and enlargement / reduction, and the motion model of the local motion compensation method uses a translation motion model.

Further, since the global motion compensation method does not have a motion vector for each small block, the motion vector is predicted based on the motion vector used for local motion compensation. When predicting a motion vector, the format of the motion vector is converted into a motion vector of a parallel motion model. When converting a motion vector for global motion compensation into a motion vector for a parallel motion model, a representative vector of a small block is calculated. The representative vector of the small block is calculated as the average value of the motion vectors calculated for each pixel of the small block. The size of the small block is M × N pixels.

[First Embodiment]

(1) Motion vector prediction encoding method and decoding method: Next, a motion vector prediction encoding method and a decoding method according to the first embodiment of the present invention will be described with reference to FIG.
This will be described with reference to the flowchart shown in FIG.

(1-1) Motion vector predictive encoding method:
First, in step S1 of FIG. 1, the motion compensation mode of the small block to be encoded is determined. Here, when the motion compensation mode of the small block to be coded is GMC, the prediction coding of the motion vector is not performed, and the motion vector coding process ends.

On the other hand, if the motion compensation mode of the small block to be coded is LMC in step S1,
Proceeding to step S2, the encoding mode of the encoded small block is determined. If the coding mode of the coded small block is determined to be the intra-frame coding mode, the process proceeds to step S3, and the prediction vector vp is set to 0, that is, vp
= (0, 0), and the process proceeds to step S7 (described later).

In step S2, the encoded
The coding mode of the small block is the inter-frame coding mode.
If it is determined, the process proceeds to step S4, and the encoded small
Determine the motion compensation mode of the block. And encoding
The motion compensation mode of the completed small block is determined to be LMC
The motion vector v of the small block_i-1= (X_i-1, Y
_i-1) As the prediction vector vp and proceed to step S7.
No.

On the other hand, if it is determined in step S4 that the motion compensation mode of the coded small block is GMC, the process proceeds to step S5, where the global motion vector GMV = (X, Y, Z) The motion vector v of the parallel motion model for each pixel of the small block
_i-1 (m, n) = (x _i-1 (m, n), y _i-1 (m, n
n)). Here, m and n represent the positions of the pixels in the small block.

Then, the process proceeds to a step S6, wherein a step 1
5, a representative motion vector is calculated from the motion vector vi _-1 (m, n) for each pixel of the coded small block, and the predicted vector is set as vp = ( _xvp , _yvp ). Here, the representative motion vector is calculated from the average value of the motion vector vi _-1 (m, n) for each pixel.

[0066]

(Equation 5)

Then, the process proceeds to a step S7, wherein the prediction error dx is calculated from the motion vector v _i = (x _i , y _i ) of the small block to be encoded and the prediction vector vp according to the following equation. Calculate _i and dy _i . _{That, dx i = x i -x vp} ...... (12); dy i = y i -y vp ...... (13)

If it is determined in step S2 that the coding mode of the coded small block is the intra-frame coding mode, in step S3 the prediction vector vp = (0,
0) and since the prediction error dx _i, dy _{i is, dx i = x i -0 ......} (14); dy i = y i -0 becomes ... (15).

If it is determined in step S4 that the motion compensation mode of the coded small block is LMC, the motion vector v _{i-1 of} the coded small block = (x _i−1 , y
_i-1 ) is used as the prediction vector vp, so that the prediction error d
x _i, dy _{i is, dx i = x i -x i} -1 ...... (16); and become _{dy i = y i -y i-} 1 ...... (17).

Next, the process proceeds to step S8, in which the prediction errors dx _i and dy _i calculated in step S7 are losslessly encoded (for example, Huffman encoding), and the motion vector encoding process is terminated.

(1-2) Motion vector decoding method:
A method of decoding a motion vector encoded by the above-described encoding method will be described with reference to FIG.

First, in step S11, the motion compensation mode of the small block to be decoded is determined. Here, when the motion compensation mode of the small block to be decoded is determined to be GMC, it is determined that the prediction encoding of the motion vector has not been performed, and the decoding process ends.

[0073] On the other hand, when the motion compensation mode of the decoded sub-block is determined to LMC, the process proceeds to step S12, decodes the prediction error dx _i and dy _i is lossless coding.

Then, the process proceeds to a step S13, wherein the coding mode of the decoded small block is determined. Here, when it is determined that the encoding mode of the decoded small block is the intra-frame encoding mode, in step S14, the prediction vector is set to vp = (0, 0), and the process proceeds to step S18 (described later). .

If it is determined in step S13 that the coding mode of the decoded small block is the inter-frame coding mode, the flow advances to step S15 to determine the motion compensation mode of the decoded small block. Then, when the motion compensation mode of the decoded small block is determined to be LMC, the motion vector v _i-1 of the decoded small block is calculated.
The process proceeds to step S18, using (x _i−1 , y _i−1 ) as the prediction vector vp.

On the other hand, if it is determined in step S15 that the motion compensation mode of the decoded small block is GMC, the process proceeds to step S16, where the global motion vector GMV =
From (X, Y, Z), the motion vector v _i-1 (m, n) = of the parallel motion model for each pixel of the decoded small block
(X _i-1 (m, n), y _i-1 (m, n)) is obtained. Here, m and n represent the positions of the pixels in the small block.

Then, the process proceeds to a step S17, wherein a representative motion vector is calculated from the motion vector V _i-1 (m, n) for each pixel of the decoded small block, and the predicted vector vp = (x
_vp , _yvp ). Here, the representative motion vector is calculated from the average value of the motion vector vi _-1 (m, n) for each pixel.

[0078]

(Equation 6)

[0079] then proceeds to step S18, the prediction and each element x _vp and y _vp vector vp, the prediction error dx _i and dy _i and were respectively added, the motion vector of the decoded small block v _i = (x _i y _i ) is calculated. That is, x _i = x _vp + d x _i (20); y _i = y _vp + d _{i i} (21)

If it is determined in step S13 that the encoding mode of the decoded small block is the intra-frame encoding mode, the prediction vector vp =
(0, 0), the motion vector v _i = (x _i y _i ) of the small block to be decoded is x _i = 0 + dx _i (22); y _i = 0 + dy _i (23) .

If it is determined in step S15 that the motion compensation mode of the decoded small block is LMC, the motion vector v _{i-1 of the} decoded small block = (x _i−1 ,
y _i−1 ) is used as the prediction vector vp, so that the motion vector v _i = (x _i y _i ) of the small block to be decoded is x _i = x _i−1 + dxi _i (24); y _i = y _i-1 + dy _i (25)

When the processing in step S18 ends, the motion vector decoding processing ends.

In the above-described embodiment, in step S6 of FIG. 1 and step S17 of FIG. 2, the calculation of the representative motion vector is calculated by the average value. However, the following statistic may be used. Here, a set of motion vector components x _ij (m, n) and y _i-1 (m, n) for each pixel is defined as X _i-1 and Y _i-1 , respectively.

The maximum value The representative motion vector is calculated from the following equation. _xvp = max (Xi _-1 ) (26); _yvp = max ( _{Yi -1} ) (27)

The minimum value The representative motion vector is calculated from the following equation. x _vp = min (X _i-1 ) (28); y _vp = min (Y _i-1 ) (29)

Intermediate value The representative motion vector is calculated from the following equation. _xvp = {max (Xi _-1 ) + min (Xi _-1 )} / 2 (30); _yvp = {max ( _{Yi -1} ) + min (Yi _-1 )} / 2 ... (31)

The frequency distribution is examined for each component of the mode motion vector, and the value having the highest frequency is determined as the component of the representative motion vector.

Median First, the components of the motion vector are arranged in descending order. xs ₁ ≤xs ₂ ≤ ... ≤xs _n-1 ≤xs _n ... (32); ys ₁ ≤ys ₂ ≤ ......... ≤ys _n-1 ≤ys _n ... (33) where n is The number of pixels in the small block, xs and ys, are obtained by rearranging the motion vector components Xi _-1 and Yi _-1 for each pixel in descending order. Then, the representative motion vector is calculated from the order of the components of the rearranged motion vectors. That is,

X _vp = xs _{(n + 1) / 2} (34); y _vp = ys _{(n + 1) / 2} (35)-Pixel number n is even _Xvp = (xsn _{/ 2} + xsn _{/ 2 + 1} ) / 2 (36); _yvp = (ysn _{/ 2} + ysn _{/ 2 + 1} ) / 2 (37)

In the above-described embodiment, the motion vector of the coded small block may be a small block coded before the current small block, including the motion vector of the previously coded small block. For example, the motion vector can be used. Further, it is also possible to calculate a prediction vector from motion vectors of a plurality of encoded small blocks.

(2) Motion vector predictive coding apparatus and decoding apparatus: Next, referring to FIGS. 3 and 4, a code for performing coding and decoding of a motion vector according to the above-described motion vector coding method and decoding method. The decoding device and the decoding device will be described.

(2-1) Motion Vector Prediction Coding Apparatus FIG. 3 shows the configuration of a coding apparatus that performs motion vector coding according to the above-described motion vector coding method. Here, in this figure, a global motion vector is input once per screen, and a local motion vector is input only when the small block to be coded is local compensation.

First, the motion vector memory 1 stores the input local motion vector mv of the encoding target small block, and then stores the input local motion vector mv of the encoding target small block. The local motion vector is output as the local motion vector mv _t−1 of the coded small block.

The representative motion vector calculation unit 2 has a memory therein, the local motion compensation is used for the input small block to be coded, and the global motion compensation is used for the small coded block. At this time, the already coded global motion vector gmv is read from the internal memory, a representative motion vector is calculated for each of the small blocks from the global motion vector gmv, and then the local motion vector mv of the coding target small block is calculated. When input, it is output as a representative motion vector of the encoded small block.

The selection unit 3 outputs the data from the motion vector memory 1 according to the motion compensation mode of the small block to be coded, the coding mode of the coded small block, and the motion compensation mode of the coded small block. Either the motion vector mv _{t- 1} of the encoded small block or the representative motion vector output from the representative motion vector calculation unit 2 is output, or both vectors are not selected ( Hereinafter, it will be called neutral state)
Do not output both motion vectors.

[0096] Further, the subtractor 4, the local motion vector mv _t of the encoding target small block subtracts the output of the selector 3, and outputs a prediction error dmv _t. The motion vector encoding unit 5 calculates the prediction error dmv _t output from the subtractor 4
Is variable-length coded and output as local motion vector information. The motion vector coding unit 6 performs variable length coding on the global motion vector gmv, and outputs it as global motion vector information.

The operation of the coding apparatus having the above configuration will be described. First, when a global motion vector is supplied from outside, the supplied global motion vector gm
v is input to the representative motion vector calculator 2 and the motion vector encoder 6, respectively. Then, after being subjected to variable-length coding in the motion vector coding unit 6, the data is output to the outside. In the representative motion vector calculation unit 2, the global motion vector gmv is stored in an internal memory.

At this time, no local motion vector is supplied from the outside, and the selecting unit 3 recognizes that the supplied motion vector is a global motion vector and enters a neutral state. No local motion vector information is output from the unit 5.

Next, after the above processing is performed, when the local motion vector mv _t is supplied from the outside, the supplied local motion vector mv _t is input to the motion vector memory 1 and the subtractor 4 respectively. . In this case, the motion compensation method for the small block to be coded is local motion compensation, and the motion compensation method for the coded small block is global motion compensation. From the global motion vector gmv stored in the parallel motion model for each pixel of the coded small block.
_{t-1 (m, n)} = (x t -1 (m, n), y t-1 (m,
n)) is obtained (corresponding to the processing of step S5 in FIG. 1). Here, m and n represent the positions of the pixels in the small block.

Then, the representative motion vector calculation section 2 calculates a representative motion vector from the obtained motion vector v _t-1 (m, n) and outputs it as a motion vector of the coded small block (FIG. 1). , Step S6). Here, the representative motion vector is calculated from the average value of the motion vector v _t-1 (m, n) for each pixel. Also,
The selecting unit 3 determines the coding mode of the coded small block, and if the coding mode is the intra-frame coding mode, the selection unit 3 enters a neutral state and does not output any signal (corresponding to the processing in step S3 in FIG. 1). supplied local motion vector mv _t is directly passed through the subtracter 4 (FIG. 1, in the processing of step S7, the aforementioned (14), corresponds to the arithmetic processing by the equation (15)), the motion vector coding After being subjected to the variable-length encoding in the unit 5 (corresponding to the processing in step S8 in FIG. 1),
Output to the outside.

On the other hand, when the selecting unit 3 determines that the coding mode of the coded small block is the inter-frame coding mode, it determines the motion compensation mode of the coded small block (FIG. 1, step (Corresponds to the processing of S4). Here, since the motion compensation mode of the encoded small block is the global motion compensation mode, the selection unit 3 selects the representative motion vector output from the representative motion vector calculation unit 2 and outputs the selected motion vector to the subtractor 4. .

As a result, the subtractor 4 subtracts the representative motion vector from the small block to be encoded (FIG. 1, in the process of step S7, the expression (12), (1)
3) corresponds to the arithmetic processing by formula), the subtraction result is outputted to the motion vector coding unit 5 as a prediction error dmv _t,
After being subjected to variable-length coding (corresponding to the processing in step S8 in FIG. 1), the data is output to the outside.

Next, after the above-described processing is performed,
And the local motion vector mv _t Where was supplied
If the supplied local motion vector mv_t Is a movement
Input to the vector memory 1 and the subtractor 4, respectively.
In addition, the motion vector memory 1 stores the previously input local
Motion vector mv_t-1 Is output.

The selecting section 3 determines the coding mode of the coded small block. If the coding mode is the intra-frame coding mode, the selection section 3 enters a neutral state and does not output any signal (FIG. 1, step S3). The supplied local motion vector mv _t passes through the subtractor 4 as it is (FIG. 1, FIG. 1).
In the processing of step S7, the above-described equation (14),
After being subjected to variable-length coding in the motion vector coding unit 5 (corresponding to the processing of step S8 in FIG. 1), it is output to the outside.

On the other hand, when the selecting unit 3 determines that the coding mode of the coded small block is the inter-frame coding mode, it determines the motion compensation mode of the coded small block (FIG. 1, step (Corresponds to the processing of S4). Here, since the motion compensation mode of the coded small block is the local motion compensation mode, the selecting unit 3 subtracts the coded local motion vector mv _t−1 output from the motion vector memory ₁ by the subtractor Output to 4.

Thus, the subtracter 4 subtracts the encoded local motion vector mv _t−1 from the motion vector mv _{t of the} small block to be encoded (FIG. 1, step S
7, the subtraction result is the prediction error dmv
It is output to the motion vector encoding unit 5 as _t , and after being subjected to variable length encoding (corresponding to the processing in step S8 in FIG. 1), it is output to the outside.

(2-2) Motion vector decoding device: FIG. 4 shows the configuration of a decoding device for decoding a motion vector according to the above-described motion vector decoding method. In this figure, the motion vector decoding unit 10 decodes the global motion vector information output from the motion vector predictive coding apparatus shown in FIG. 3, and outputs it as a global motion vector gmv to the representative motion vector calculation unit 12 and the outside. . Motion vector decoding unit 11 decodes the local motion vector information outputted from the motion vector prediction encoding device of FIG. 3, and outputs to the adder 15 as a prediction error dmv _t.

The representative motion vector calculation unit 12 has a memory inside, and when the motion compensation method of the small block to be decoded is local motion compensation and the motion compensation method of the decoded small block is global motion compensation, the motion is already decoded. The global motion vector gmv thus read out is read out from an internal memory, a representative motion vector is calculated for each of the small blocks from the global motion vector gmv, and then the local motion vector information is decoded by the motion vector decoding unit 11, and the prediction error dmv _t Is output as the representative motion vector of the decoded small block.

The selecting section 13 is output from the representative motion vector calculating section 12 according to the motion compensation mode of the small block to be decoded, the coding mode of the decoded small block, and the motion compensation mode of the decoded small block. Either the representative motion vector or the motion vector mv _t−1 of the decoded small block output from the motion vector memory 14 is output, or the state becomes neutral, and no motion vector is output.

The motion vector memory 14 stores the local motion vector mv of the decoded small block output from the adder 15. Then, the motion vector decoding unit 1
When the prediction error dmv _t is output from 1, the stored local motion vector is output as the local motion vector mv _t−1 of the decoded small block. The adder 15, the prediction error dmv _t output from the motion vector decoding unit 11 adds the output of the selector 13, and outputs it to the motion vector memory 14 and the external as a local motion vector mv _t.

Next, the operation of the motion vector decoding device having the above configuration will be described. First, when global motion vector information is supplied from the motion vector predictive coding apparatus shown in FIG. 3, the supplied global motion vector information is decoded into a global motion vector gmv by the motion vector decoding unit 10, and the representative motion vector A calculating unit 12,
It is output to the outside. As a result, the global motion vector gmv is stored in the internal memory in the representative motion vector calculation unit 12.

At this time, the local motion vector information is not supplied from the motion vector predictive coding apparatus of FIG. 3, and the selecting unit 13 determines that the supplied motion vector information is global motion vector information. since the recognition to a neutral state, never local motion vector mv _t is output from the adder 15.

Next, after the above-described processing is performed, the operation shown in FIG.
Local motion vector information from
Information, the supplied local motion vector information
The report is a prediction error dmv in the motion vector decoding unit 11.
_t (Corresponding to the process of step S12 in FIG. 2),
It is output to the adder 15. In this case, the decryption target small
Block motion compensation is local motion compensation, and
Global motion compensation
In the representative motion vector calculation unit 12,
Is the global motion vector gmv stored in the
Then, for each pixel of the decoded small block,
Motion vector v_t-1(M, n) = (x _t-1(M,
n), y_t-1(M, n)) (FIG. 2, step
Step S16). Here, m and n are small blocks.
Represents the position of the pixel within the box.

Then, the obtained motion vector v
_A representative motion vector is calculated from _t-1 (m, n) and output as a motion vector of the decoded small block (corresponding to the process of step S17 in FIG. 2). Here, the representative motion vector is calculated from the average value of the motion vector v _t-1 (m, n) for each pixel.

At this time, the selecting unit 13 determines the coding mode of the decoded small block, and if the coding mode is the intra-frame coding mode, the selecting unit 13 enters a neutral state and does not output any signal (FIG. 2, step S14). ). Therefore, the decoded prediction error dmv _t is directly added to the local motion vector mvt without any addition in the adder 15.
_t , which is output to the motion vector memory 14 and the outside (in the process of FIG. 2, step S18,
(Equivalent to the arithmetic processing by the aforementioned equations (22) and (23)).

On the other hand, when the selecting section 13 determines that the coding mode of the decoded small block is the inter-frame coding mode, it determines the motion compensation mode of the decoded small block (FIG. 2, step S15). Processing equivalent). Here, since the motion compensation mode of the decoded small block is the global motion compensation mode, the selector 13 selects the representative motion vector output from the representative motion vector calculator 12 and outputs the selected motion vector to the adder 15. .

[0117] Thus, the adder 15, the prediction error dmv _t decoded by the motion vector decoding unit 11, a representative motion vector and is added (Fig. 2, in the process of step S12, described above (20), (21) corresponds to the arithmetic processing by the formula), to result of the addition as a local motion vector mv _t, the motion vector memory 14 and the outside,
Each is output.

Next, after the above-described processing is performed, when the local motion vector information is further supplied from the motion vector predictive coding apparatus of FIG.
Is decoded into a prediction error dmv _t and output to the adder 15. Then, the selecting unit 13 again determines the coding mode of the decoded small block, and if the coding mode is the intra-frame coding mode, the selecting unit 13 enters a neutral state and does not output any signal (FIG. 2, step S14). equivalent), the decoded prediction error dmv _t, without being added any way in the adder 15, as it is the local motion vector mv _t, to a motion vector memory 14 and external, are outputted (Fig. 2, step S18 (Corresponds to the arithmetic processing by the above-described equations (22) and (23)).

On the other hand, when the selecting unit 13 determines that the coding mode of the decoded small block is the inter-frame coding mode, the motion compensation mode of the decoded small block is determined (step S15 in FIG. 2). ). Here, since the motion compensation mode of the decoded small block is the local motion compensation mode, the selecting unit 13 adds the local motion vector mv _t-1 of the decoded small block output from the motion vector memory 14 to the adder 15 is output.

As a result, the adder 15 adds the local motion vector mv _{t−1 of} the decoded small block output from the motion vector memory 14 and the prediction error mv _t (FIG. 2, processing in step 28). (2)
4), (25) corresponding to the arithmetic processing by the formula), to result of the addition to the motion vector memory 14 and the external as a local motion vector mv _t, are output.

In the motion vector predictive coding apparatus and decoding apparatus shown in FIGS. 3 and 4, the representative motion vector calculated by each of the representative vector calculation units 2 and 12 is, for each pixel of the decoded small block, , As described in (1-1) Encoding method and (1-2) Decoding method, the maximum value, minimum value, intermediate value, mode value, median value, etc. Statistics may be used.

[Second Embodiment]

(1) Motion vector prediction encoding method and decoding method: Next, a second embodiment of the motion vector prediction encoding method and decoding method according to the present invention will be described. The difference between the motion vector prediction encoding method and the decoding method in the second embodiment is that the prediction vector obtained from the global motion parameters exceeds the range of the local motion vector.
The point is that processing for clipping the prediction vector to the minimum value or the maximum value of the range is added.

(1-1) Motion vector predictive encoding method:
Hereinafter, the motion vector predictive encoding method according to the second embodiment will be specifically described with reference to the flowchart shown in FIG. FIG. 5 is a flowchart illustrating a motion vector predictive encoding method according to the second embodiment.
In this figure, steps for performing the same processing as the motion vector predictive encoding method shown in FIG. 1 are denoted by the same reference numerals, and detailed description thereof will be omitted.

The difference between the motion vector predictive encoding method shown in FIG. 5 and FIG. 1 is that after the representative motion vector is calculated in step 16, whether the calculated representative motion vector value is within a predetermined range. It is checked whether or not the value is outside the range, and the value of the representative motion vector is clipped.

That is, as described in the first embodiment, in step S1, the motion compensation mode of the current block is determined to be local motion compensation, and in step S2, the coding mode of the coded small block is And the motion compensation mode is set to G in step S4.
If it is determined to be MC, a motion vector of a parallel motion model is calculated for each pixel of the coded small block from the global motion vector GMV in step S5.

Then, in step S6, the average value of the motion vector for each pixel of the coded small block calculated in step S5 is obtained, and the obtained average value is used as a representative motion vector. Here, the representative motion vector is the same as the motion vector predictive encoding method in the first embodiment.
Not only the average value of the motion vector for each pixel of the encoded small block but also a statistic such as a maximum value, a minimum value, an intermediate value, a mode value, and a median value may be used.

Then, the process proceeds to a step S20, wherein a step S6 is performed.
It is determined whether or not the value of the representative motion vector obtained in is within a predetermined range. If it has exceeded, the process proceeds to step S21, and the above-described representative motion vector is clipped within a predetermined range.

Here, the expressible range of the motion vector (for example, fc in MPEG-4 shown in [Table 1])
mode) is MV _min or more and MV _max
If it is less than or equal to MV _min , if the representative motion vector is less than MV _min , and if it exceeds MV _max , MV _max
And For example, when the range of the motion vector is fcode = 3 (-64 to +63.5; see [Table 1]) defined by MPEG-4, the prediction vector vp obtained from the global motion parameters is (x _PV , y _PV ) = (+ 102, +
75), these values are forcibly applied (63.
5,63.5).

On the other hand, if it is determined in step S20 that the value of the representative motion vector is within the predetermined range, the representative motion vector calculated in step S6 is used as it is as the prediction vector vp.

Then, the process proceeds to a step S7, wherein a difference (prediction error) between the motion vector of the small block to be encoded and the prediction vector is calculated. Here, in MPEG-4, FIG.
As shown by, three blocks on the left, right above, and diagonally right above the current block are referred to. Therefore, for each of these blocks, steps S2 to S6, S20, S2
After individually performing the processes of S1 and S7, the median of the three candidate vectors is set as the prediction error.

Next, the process proceeds to step S8, where
The prediction error obtained in step 7 is encoded, and the encoding process in the second embodiment ends.

(1-2) Motion vector decoding method:
The motion vector decoding method according to the second embodiment will be specifically described with reference to the flowchart shown in FIG. FIG. 6 is a flowchart illustrating a motion vector decoding method according to the second embodiment. In this figure, steps for performing the same processing as the motion vector decoding method shown in FIG. 2 are denoted by the same reference numerals, and detailed description thereof will be omitted.

The motion vector decoding method shown in FIG.
The difference is that after calculating the representative motion vector in step 27, it is checked whether or not the calculated representative motion vector value is within a predetermined range. The point is to clip

That is, as described in the first embodiment, when it is determined in step S11 that the motion compensation mode of the current block is local motion compensation, step S12 is performed.
Decodes the prediction error. Then, in step S13, it is determined that the coding mode of the decoded small block is the inter-frame coding, and in step S15, it is determined that the motion compensation mode is GMC, the decoded small block is decoded from the global motion vector GMV in step S16. A motion vector of the parallel motion model is calculated for each pixel of the block.

In step S17, the average value of the motion vector for each pixel of the decoded small block calculated in step S16 is obtained, and the obtained average value is set as a representative motion vector. Here, the representative motion vector is the first motion vector.
Similarly to the decoding method of the embodiment, not only the average value of the motion vector for each pixel of the decoded small block but also a statistic such as a maximum value, a minimum value, an intermediate value, a mode value, and a median value may be used. .

Then, the process proceeds to a step S22, wherein a step S1 is executed.
It is determined whether or not the representative motion vector obtained in step 7 is within a predetermined range. If it has exceeded, the process proceeds to step S23, and the above-described representative motion vector is clipped within a predetermined range.

Here, the expressible range of the motion vector (for example, fc in MPEG-4 shown in [Table 1])
mode) is MV _min or more and MV _max
If it is less than or equal to MV _min , if the representative motion vector is less than MV _min , and if it exceeds MV _max , MV _max
And For example, when the range of the motion vector is fcode = 3 (-64 to +63.5; see [Table 1]) defined by MPEG-4, the prediction vector vp obtained from the global motion parameters is (x _PV , y _PV ) = (+ 102, +
75), these values are forcibly applied (63.
5,63.5).

On the other hand, when it is determined in step S22 that the value of the representative motion vector is within the predetermined range, the representative motion vector calculated in step S17 is used as it is as the prediction vector vp.

Then, the process proceeds to a step S18, and the step S1
The prediction error of the small block to be decoded obtained in step 2 and the prediction vector are added. Here, in MPEG-4, as shown in FIG. 14, three blocks on the left, right above, and diagonally right above the current block are referred to. Therefore, in each of these blocks, steps S12 to S17, S22, S
23, after performing the processes of S18 individually, the median value of the three candidate vectors is set as the prediction vector. Then, the decoding process in the second embodiment ends.

In the above-described motion vector predictive encoding method and decoding method, the clipping process performed in step S21 of FIG. 5 and step S23 of FIG. 6 is performed by clipping at a maximum value or a minimum value in a predetermined range. Instead, it may be clipped at 0.

That is, when the range in which a motion vector can be expressed is, for example, not less than MV _{min and not} more than MV _max , if the representative motion vector is less than MV _min , it is set to 0.
It is set to 0 even when it exceeds _max . For example, the range in which a motion vector can be expressed is fco defined by MPEG-4.
When de = 3 (−64 to +63.5; see [Table 1]), the prediction vector v obtained from the global motion parameter
When p is (x _PV , y _PV ) = (+ 102, +75), these are forcibly set to (0, 0).

(2) Motion vector predictive coding apparatus and decoding apparatus:

(2-1) Motion vector predictive coding apparatus:
Next, a motion vector predictive coding apparatus that performs predictive coding of a motion vector according to the motion vector predictive coding method according to the second embodiment (see FIG. 5) will be described with reference to FIG. In FIG. 7, the same components as those of the motion vector predictive coding apparatus shown in FIG. 3 are denoted by the same reference numerals, and detailed description thereof will be omitted.

The difference between the motion vector predictive coding apparatus shown in FIG. 7 and FIG. 3 is that a representative motion vector clipping section 20 is provided between the representative motion vector calculating section 2 and one input end of the selecting section 3. That is the point. The representative motion vector clipping unit 20 checks whether or not the value of the representative motion vector output from the representative motion vector calculation unit 2 is within a predetermined range. Clips the value of a vector to the maximum or minimum value in the range.

According to the motion vector predictive coding apparatus having the above-described configuration, when the representative motion vector is calculated by the representative motion vector calculating section 2, that is, in the representative motion vector calculating section 2, FIG. S5, S
6 is performed, the representative motion vector clipping unit 20 causes the representative motion vector calculation unit 2
It is determined whether or not the representative motion vector value calculated in is within a predetermined range (FIG. 5, step S20).
).

If the value of the representative motion vector is outside the predetermined range, the value of the representative motion vector is clipped within the predetermined range (FIG. 5, corresponding to the processing in step S21). Output to the selection unit 3. Also,
If the representative motion vector value is within a predetermined range, the representative motion vector calculated by the representative motion vector calculation unit 2 is output to the selection unit 3 as it is.

That is, a range in which a motion vector can be expressed (for example, fc in MPEG-4 shown in [Table 1])
mode) is MV _min or more and MV _max
If the it was the following representative motion vector in the MV _min when less than MV _min, also when more than MV _{ma x} is the MV _max. For example, when the range of the motion vector is fcode = 3 (-64 to +63.5; see [Table 1]) defined by MPEG-4, the prediction vector vp obtained from the global motion parameters is (x _PV , y _PV ) = (+ 102, +
75), these values are forcibly applied (63.
5,63.5).

It should be noted that the representative motion vector clipping unit 2
In the clipping process performed at 0, clipping may be performed at 0 instead of clipping at a maximum or minimum value in a predetermined range. That is, when the range in which the motion vector can be expressed is, for example, MV _min or more and MV _max or less, the value is set to 0 when the representative motion vector is less than MV _min , and is set to 0 when the representative motion vector exceeds MV _max . For example, the range in which a motion vector can be represented is fcode = 3 defined by MPEG-4.
(−64 to +63.5; see [Table 1]), the prediction vector vp obtained from the global motion parameters is (x
_PV, y _PV) = (+ 102, if the + 75), forcing the (0,0) them.

As a result, it is possible to realize a motion vector predictive coding apparatus which performs coding according to the motion vector predictive coding method shown in the flowchart of FIG.

(2-2) Motion vector decoding device:
A motion vector decoding device that decodes a motion vector according to the motion vector decoding method according to the second embodiment (see FIG. 6) will be described with reference to FIG. FIG.
, The same components as those of the motion vector decoding device shown in FIG. 4 are denoted by the same reference numerals, and detailed description thereof will be omitted.

The difference between the motion vector decoding apparatus shown in FIG. 8 and FIG.
The representative motion vector clipping unit 21 is provided between the input terminal and the one input terminal. This representative motion vector clipping unit 21
2 is checked to see if the value of the representative motion vector is within a predetermined range. If the value exceeds the range, the value of the representative motion vector is clipped to the maximum value or the minimum value of the range. I do.

According to the motion vector decoding apparatus having the above-described configuration, when the representative motion vector is calculated by the representative motion vector calculating unit 12, that is, in the representative motion vector calculating unit 12, the process shown in FIGS. S
17 is performed, the representative motion vector clipping unit 21 determines whether or not the representative motion vector value calculated by the representative motion vector calculation unit 12 is within a predetermined range (FIG. 6, Step S
22).

If the value of the representative motion vector exceeds the predetermined range, the value of the representative motion vector is clipped within the predetermined range (corresponding to the processing in step S22 in FIG. 6). Output to the selection unit 13. If the representative motion vector value is within a predetermined range, the representative motion vector calculated by the representative motion vector calculation unit 12 is output to the selection unit 13 as it is.

That is, the expressible range of the motion vector (for example, fc in MPEG-4 shown in [Table 1])
mode) is MV _min or more and MV _max
If the it was the following representative motion vector in the MV _min when less than MV _min, also when more than MV _{ma x} is the MV _max. For example, when the range of the motion vector is fcode = 3 (-64 to +63.5; see [Table 1]) defined by MPEG-4, the prediction vector vp obtained from the global motion parameters is (x _PV , y _PV ) = (+ 102, +
75), these values are forcibly applied (63.
5,63.5).

The representative motion vector clipping unit 2
The clipping process performed at 1 may be clipped at 0 instead of clipping at the maximum or minimum value in a predetermined range. That is, when the range in which the motion vector can be expressed is, for example, MV _min or more and MV _max or less, the value is set to 0 when the representative motion vector is less than MV _min , and is set to 0 when the representative motion vector exceeds MV _max . For example, the range in which a motion vector can be represented is fcode = 3 defined by MPEG-4.
(−64 to +63.5; see [Table 1]), the prediction vector vp obtained from the global motion parameters is (x
_PV, y _PV) = (+ 102, if the + 75), forcing the (0,0) them.

Thus, it is possible to realize a decoding device that performs decoding according to the motion vector decoding method shown in the flowchart of FIG.

In the first and second embodiments, the motion vector predictive coding procedure shown in the flowchart of FIG. 1 or FIG. 5, or the motion vector memory 1 shown in the block diagram of FIG. Motion vector calculator 2, selector 3, subtractor 4, motion vector encoder 5
6. Further, a program for realizing the operation of the representative motion vector clipping unit 20 shown in the block diagram of FIG. 7 is recorded on a computer-readable recording medium such as a CD-ROM or a floppy disk, and May be read by the computer and executed to perform predictive coding of the motion vector.

Similarly, the motion vector decoding procedure shown in the flowchart of FIG. 2 or FIG. 6, or the motion vector decoding sections 10 and 11, the representative motion vector calculation section 12, and the selection section 13 shown in the block diagram of FIG. , A motion vector memory 14, an adder 15, and a program for realizing the operation of the representative motion vector clipping unit 21 shown in the block diagram of FIG. 8 are recorded on a computer-readable recording medium, and are recorded on this recording medium. The computer may load and execute the program so as to decode the motion vector.

The present invention is not limited to the above embodiments, but can be variously modified and applied within the scope of the claims.

[0161]

As described above, according to the motion vector prediction encoding method and decoding method, the prediction encoding device and the decoding device, and the motion vector prediction encoding program and the decoding program of the present invention, a motion model Can be predicted between motion vectors different from each other, and the prediction of a motion vector can be performed for a motion vector between global motion compensation and local motion compensation, so that the amount of generated codes of a motion vector can be reduced. . Further, an effect of improving the efficiency of motion vector prediction can be obtained.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a motion vector predictive encoding method according to a first embodiment of the present invention.

FIG. 2 is a flowchart illustrating a motion vector decoding method according to the first embodiment of the present invention.

FIG. 3 is a block diagram illustrating a configuration of a motion vector predictive encoding device according to the first embodiment of the present invention.

FIG. 4 is a block diagram illustrating a configuration of a motion vector decoding device according to the first embodiment of the present invention.

FIG. 5 is a flowchart illustrating a motion vector predictive encoding method according to the second embodiment of the present invention.

FIG. 6 is a flowchart illustrating a motion vector decoding method according to the second embodiment of the present invention.

FIG. 7 is a block diagram illustrating a configuration of a motion vector predictive encoding device according to a second embodiment of the present invention.

FIG. 8 is a block diagram illustrating a configuration of a motion vector decoding device according to a second embodiment of the present invention.

9A and 9B are diagrams illustrating an example of a motion model, in which FIG. 9A illustrates a parallel motion model, and FIG. 9B illustrates a parallel motion + enlargement / reduction motion model.

FIG. 10 is a diagram showing a motion vector of a motion model.

FIG. 11 is a diagram illustrating an example (MPEG-4 encoder) of an encoder of a moving image encoding method.

FIG. 12 shows a decoder (MP) corresponding to the encoder of FIG.
It is a figure showing the composition of EG-4 decoder).

FIG. 13 is a diagram showing an arrangement of reference blocks in MPEG-4 motion vector prediction.

[Explanation of symbols]

 1,14 motion vector memory 2,12 representative motion vector calculation unit 3,13 selection unit 4 subtractor 5,6 motion vector encoding unit 10,11 motion vector decoding unit 15 adder 20,21 representative motion vector clipping unit

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Hiroshi Watanabe 3-19-2 Nishishinjuku, Shinjuku-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Inventor Jun Sagata 3-19, Nishishinjuku, Shinjuku-ku, Tokyo No. 2 Nippon Telegraph and Telephone Corporation (72) Inventor Masayuki Takamura 3-19-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo Nippon Telegraph and Telephone Corporation (56) References JP-A-10-136374 (JP, A) (58) Field surveyed (Int. Cl. ⁶ , DB name) H04N ^7/ 24-7/68

Claims (40)

(57) [Claims] 1. A motion vector predictive encoding method which divides an encoding target frame into small blocks and switches between a global motion compensation method and a local motion compensation method for each encoding target small block. When predicting the motion vector of a small block from the motion vector of a coded small block, the motion compensation method of the small block to be coded is a local motion compensation method.
The motion compensation method of the coded small block to be used is global.
For motion compensation, encode from global motion vector
Calculate the local motion vector of the small block
Of the motion vector of the small block
Because, the motion vector predictive encoding method characterized by coding the prediction error of the motion vector. 2. The method according to claim 1, wherein said small block to be encoded is
The size of the prediction vector of the motion vector
If not, the magnitude of the obtained prediction vector
Is clipped within the specified range.
The motion vector predictive encoding method according to claim 1. [Claim 3] put before Symbol obtained coded small block
The size of the prediction vector of the motion vector
If not, the magnitude of the obtained prediction vector is
2. The motion vector predictive encoding method according to claim 1 , wherein 0 is set. 4. The motion vector prediction code according to claim 1,
Of decoding a motion vector encoded using the encoding method
In the vector decoding method, the motion compensation method for the decoding target small block is a local motion compensation method.
Compensation method for decoded small blocks used for prediction
If is global motion compensation, the global motion vector
Calculate local motion vector of decoded small block from
And the prediction vector of the motion vector of the decoding target small block
Is calculated by adding the prediction error to the prediction vector.
The motion vector decoding method comprising decoding the Le. 5. The size of the obtained prediction vector is
If the value is not within the predetermined range, the size of the prediction vector is large.
The method according to claim 4 , wherein the magnitude is clipped within the specified range . 6. The motion vector decoding method according to claim 4, wherein when the size of the obtained prediction vector is not within a predetermined range, the size of the prediction vector is set to 0. 7. A motion vector of an arbitrary motion model of a small block.
A vector with a smaller number of motion vector parameters.
When converting to the motion vector of the model
From the motion vector of the motion model obtained for each pixel,
And calculating a representative motion vector of the lock
The motion vector according to any one of claims 1 to 3.
Prediction encoding method . 8. The motion vector of an arbitrary motion model of a small block.
A vector with a smaller number of motion vector parameters.
When converting to the motion vector of the model
From the motion vector of the motion model obtained for each pixel,
Calculating a representative motion vector of a lock
The motion vector decoding method according to claim 4 . 9. The motion vector of an arbitrary motion model of a small block.
Convert a vector to a motion vector of a parallel motion model
At the same time, a translational motion model obtained for each pixel in a small block
From the motion vector of the small block,
4. The method according to claim 1, wherein
The motion vector predictive encoding method according to claim 1 . 10. The motion of an arbitrary motion model of a small block.
Convert vector to parallel motion model motion vector
When moving, the parallel motion model determined for each pixel in the small block
From the motion vector of the small block, the representative motion vector of the small block
7. The motion vector decoding method according to claim 4, wherein a motion vector is calculated . 11. When calculating a representative motion vector, a small
Motion vector of parallel motion model for each pixel in lock
For each component of, the value calculated for each component, the average
Value, median, median, mode, maximum, or minimum
10. The motion vector predictive encoding method according to claim 9 , wherein? Is set as a value of each component of the representative motion vector . 12. When calculating a representative motion vector, a small
Motion vector of parallel motion model for each pixel in lock
For each component of, the value calculated for each component, the average
Value, median, median, mode, maximum, or minimum
11. The method according to claim 10 , wherein? Is set as a value of each component of the representative motion vector . 13. A global motion or deformation of the entire frame.
Predict global motion compensation and block-by-block local
Use any of the local motion compensation to predict motion
A motion vector predictive coding method for performing inter-frame prediction.
I, step of determining the motion compensation mode of the encoding target small block
And the motion compensation mode of the small block to be coded is
For compensation, determine the coding mode of the coded small block.
Setting the encoding mode of the encoded small block in a frame.
In the case of encoding, the motion vector of the block is set to 0
And the encoding mode of the encoded small block is between frames.
In the case of encoding, the motion compensation mode of the encoded small block is used.
A step of determining de, motion compensation mode of the encoded small blocks global
In the case of motion compensation, the global motion parameters
Calculate the motion vector for each pixel of the encoded small block
In a predetermined manner from the motion vector for each pixel.
Calculate the representative motion vector of a coded small block
And determining whether the representative motion vector is within a specified range.
And when the representative motion vector is not within a specified range,
Clip the table vector within the specified range, or
And a motion vector and a prediction vector of the encoding target small block.
Phase and the dynamic-out vector prediction encoding method you; and a step of encoding the calculated prediction error to calculate a prediction error that is a difference Le. 14. A motion vector predictor according to claim 13, wherein
A motion vector decoding method corresponding to the measurement coding method.
Te, the step of determining the motion compensation mode of the current block, the motion compensation mode of the current block LMC
Decoding the prediction error of the motion vector, determining the coding mode of the decoded small block, and setting the coding mode of the decoded small block to the intra-frame code.
In the case of the encoding mode, the motion vector of the block is set to 0.
And the encoding mode of the decoded small block is an interframe code.
In the case of the encoding mode, the motion compensation mode of the decoded small block is used.
A step of determining over de, motion compensation mode of the decoded sub-block is a global
In the case of motion compensation, the reference
Calculating a motion vector for each pixel of the lock; and blocking the motion vector in a predetermined manner from the motion vector for each pixel.
Calculating a representative motion vector of the lock; and determining whether the representative motion vector is within a specified range.
And when the representative motion vector is not in the specified range,
Clip the representative motion vector within the specified range
Or adding a prediction error and a prediction vector of the decoded small block.
The motion vector decoding method with the steps of calculation. 15. An encoding target frame is divided into small blocks.
Global motion compensation method for each small block to be encoded.
Motion vector prediction with switchable local motion compensation method
In the encoding device, the local motion vector of the input small block to be encoded
And then the local motion of the small block to be encoded
When a vector is input, the station of the coded small block
Encoded small block output as local motion vector
Storage means for storing the input global motion vector for each of the small blocks;
Then, by converting to the form of the local motion vector,
Prediction vector of motion vector in encoding target small block
And then the local motion of the small block to be coded
When a vector is input , a small block
Prediction vector calculation means for outputting as a measurement vector, and the motion compensation
This is a compensation method that uses the small encoded block used for prediction.
If the motion compensation method is a global motion compensation method,
The prediction vector calculated by the measurement vector calculation means is selected.
And both motion compensation methods are local motion compensation methods.
Output from the encoded small block storage means.
Predict local motion vectors of encoded small blocks
A prediction vector selecting means for selecting as a vector, the prediction said selected by the prediction vector selecting means vector
And the motion vector of the input small block to be encoded
From the motion vector in the encoding target small block
Prediction error output means for outputting a prediction error of, a prediction error output from the prediction error output means,
And encodes the input global motion vector
; And a coding unit dynamic xylem vector predictive encoding apparatus. 16. A calculation by the prediction vector calculation means.
The size of the predicted vector is within a predetermined range.
If not, the magnitude of the prediction vector is
16. The motion vector predictive coding apparatus according to claim 15, further comprising clip means for clipping within a set range.
Place . 17. Calculation by the prediction vector calculation means.
The size of the predicted vector is within a predetermined range.
, The locality of the transformed coded small block
16. The motion vector predictive coding apparatus according to claim 15 , further comprising clip means for setting a dynamic motion vector magnitude to zero . 18. The motion vector prediction according to claim 15 ,
Decode the motion vector coded by the coding device
In the motion vector decoding device, the motion vector in the encoded small block to be decoded
Decoding method for decoding prediction error and global motion vector
And a decoded small decoded by the motion vector decoding device.
The local motion vector of the block is stored and then
Output when the measurement error is decoded by the decoding means.
A global motion vector is decoded by the decoded small block storage unit and the decoding unit.
In this case, the global motion vector is
Into a dynamic motion vector,
Calculate the predicted motion vector in
Output when the prediction error is decoded by the decoding means.
Prediction vector calculation means, and a motion compensation method for a prediction error decoded by the decoding means
Is the local motion compensation method, and the decoded small
Global motion compensation method for motion vectors in blocks
In the case of motion compensation, the prediction
The prediction vector to be applied, and both motion compensation methods
If the local motion compensation method, the decoded small block
Of the decoded small block output from the block storage means
Prediction vector that selects dynamic motion vectors as prediction vectors
And a prediction vector selected by the prediction vector selection means.
And the prediction error decoded by the decoding means.
Motion vector decoding apparatus characterized by having a motion vector decoding means for decoding motion vector Te. 19. A calculation by the prediction vector calculation means.
The size of the predicted vector is within a predetermined range.
If not, the magnitude of the prediction vector is
19. The motion vector decoding device according to claim 18, further comprising clip means for clipping within a set range . 20. If the magnitude of the prediction vector calculated by the prediction vector calculation means is not within a predetermined range, the magnitude of the local motion vector of the converted coded small block is calculated. 19. The motion vector decoding device according to claim 18, further comprising a clip unit for setting to 0. 21. The prediction vector calculating means calculates a motion vector of an arbitrary motion model of a small block as a motion vector.
Motion of motion model with less number of vector parameters
Calculated for each pixel in a small block when converting to a vector
From the motion vector of the motion model, the representative motion of the small block
From claim 15, characterized that you calculate the vector can
18. The motion vector predictive encoding device according to any one of items 17 to 17 . 22. The prediction vector calculating means calculates a motion vector of an arbitrary motion model of a small block as a motion vector.
Motion of motion model with less number of vector parameters
Calculated for each pixel in a small block when converting to a vector
From the motion vector of the motion model, the representative motion of the small block
From claim 18, wherein that you calculate the vector can
21. The motion vector decoding device according to any one of 20 . 23. The predictive vector calculating means performs parallel shift of a motion vector of an arbitrary motion model of a small block.
When converting to the motion vector of the motion model, small blocks
Motion vector of the parallel motion model obtained for each pixel in
From any one of claims 15, characterized that you calculate a representative motion vector of the small block 17
Motion vector predictive coding device . 24. The predictive vector calculating means performs parallel shift of a motion vector of an arbitrary motion model of a small block.
When converting to the motion vector of the motion model, small blocks
Motion vector of the parallel motion model obtained for each pixel in
From the motion vector decoding apparatus according to any one of claims 18, characterized that you calculate a representative motion vector of the small block 20. 25. The predictive vector calculating means, when calculating a predictive vector, calculates a flat vector for each pixel in a small block.
For each component of the motion vector of the line movement motion model,
Average, median, median,
Either the mode, the maximum, or the minimum is
24. The motion vector predictive coding apparatus according to claim 23, wherein a value of each component is used. 26. The predictive vector calculating means, when calculating a predictive vector, calculates the average of each pixel in a small block.
For each component of the motion vector of the line movement motion model,
Average, median, median,
One of the mode, maximum, or minimum value is a representative motion vector
Motion vector decoding apparatus according to claim 24, wherein the a value of each component of. 27. An encoding target frame is divided into small blocks.
Global motion compensation method for each small block to be encoded.
Motion vector prediction with switchable local motion compensation method
Read by the computer which recorded the encoding program
The recording medium is a recordable medium, and the motion vector of the encoding target small block is
When predicting from the lock motion vector,
The lock motion compensation method is a local motion compensation method, and
Motion compensation method for coded small blocks used for
In the case of local motion compensation, the code
Calculate local motion vector of encoded small block
Step, the local motion vector of the calculated encoded small block
The motion vector of the small block to be encoded.
Calculate motion vector and encode prediction error of motion vector
That step a computer readable friendly recording the motion-out vector predictive encoding program that is characterized by having a
Recording media . 28. The encoding-target small block obtained as described above
The size of the predicted motion vector
If not, the magnitude of the obtained prediction vector
28. The motion vector prediction code according to claim 27 , further comprising the step of:
Computer-readable record of the
Recording medium . 29. The obtained small block to be encoded is
The size of the predicted motion vector
If not, the magnitude of the obtained prediction vector
The characterized by having a step of 0. claim 2
The motion vector predictive coding program described in No. 7 was recorded.
Computer readable recording medium . 30. A motion vector predictor according to claim 27.
The behavior encoded by executing the encoding program
A motion vector decoding program that decodes
Computer-readable recording medium
Therefore, the motion compensation method for the decoding target small block is a local motion compensation method.
Compensation method for decoded small blocks used for prediction
If is global motion compensation, the global motion vector
Calculate local motion vector of decoded small block from
And the prediction vector of the motion vector of the decoding target small block
Determining a motion by adding the prediction error to the predicted vector obtained the vector
A computer which records the motion-out vector decoding program that is characterized in that a step of decoding the Le
A readable recording medium . 31. The magnitude of the obtained prediction vector
Is not within the predetermined range, the prediction vector
The size, the motion vector restored in claim 30, characterized in that it comprises a step of clipping within the prescribed range
Computer readable recording medium recording the degree programs. 32. The size of the previous SL predicted vector obtained
Is not within the predetermined range, the prediction vector
請 characterized in that it comprises a step of size 0
31. A computer-readable recording medium recording the motion vector decoding program according to claim 30 . 33. Motion of an arbitrary motion model of a small block
Vectors with fewer motion vector parameters
When converting to a motion model motion vector,
From the motion vector of the motion model obtained for each pixel of
A computer-readable recording medium storing a motion vector predictive coding program according to any one of claims 27 to 29, comprising a step of calculating a representative motion vector of a block . 34. Motion of an arbitrary motion model of a small block
Vectors with fewer motion vector parameters
When converting to a motion model motion vector,
From the motion vector of the motion model obtained for each pixel of
33. A computer-readable recording medium recording a motion vector decoding program according to claim 30, further comprising a step of calculating a representative motion vector of a block . 35. Motion of an arbitrary motion model of a small block
Convert vector to parallel motion model motion vector
When moving, the parallel motion model determined for each pixel in the small block
From the motion vector of the small block, the representative motion vector of the small block
Calculating the number of files.
30. A computer-readable recording medium recording the motion vector predictive encoding program according to any one of 27 to 29 . 36. Motion of an arbitrary motion model of a small block
Convert vector to parallel motion model motion vector
When moving, the parallel motion model determined for each pixel in the small block
From the motion vector of the small block, the representative motion vector of the small block
Calculating the number of files.
33. A computer-readable recording medium recording the motion vector decoding program according to any one of 30 to 32 . 37. When calculating a representative motion vector, a small
Motion vector of parallel motion model for each pixel in lock
For each component of, the value calculated for each component, the average
Value, median, median, mode, maximum, or minimum
36. The motion vector predictor according to claim 35, further comprising the step of: setting a value of each component of the representative motion vector to
A computer-readable recording medium on which an encoding measurement program is recorded. 38. When calculating a representative motion vector, a small
Motion vector of parallel motion model for each pixel in lock
For each component of, the value calculated for each component, the average
Value, median, median, mode, maximum, or minimum
37. A computer-readable recording medium recording a motion vector decoding program according to claim 36, further comprising the step of: setting the value of each component of the representative motion vector . 39. Global movement and deformation of the entire frame
Predict global motion compensation and block-by-block local
Use any of the local motion compensation to predict motion
A motion vector predictive coding program that performs inter-frame prediction.
Computer readable recording medium
The motion compensation mode of the encoding target small block.
And the motion compensation mode of the encoding target small block is local motion.
For compensation, determine the coding mode of the coded small block.
Setting the encoding mode of the encoded small block in a frame.
In the case of encoding, the motion vector of the block is set to 0
Step and the encoding mode of the encoded small block is between frames.
In the case of encoding, the motion compensation mode of the encoded small block is used.
Determining a de, motion compensation mode of the encoded small blocks global
In the case of motion compensation, the global motion parameters
Calculate the motion vector for each pixel of the encoded small block
And a previous step from the motion vector for each pixel in a predetermined manner.
Calculate the representative motion vector of a coded small block
A step, determining whether the representative motion vector is within the prescribed range
And when the representative motion vector is not within a specified range,
Clip the table vector within the specified range, or
And a motion vector and a prediction vector of the encoding target small block.
Steps and, the calculated computer readable recording medium recording a moving-out vector predictive encoding program that is characterized in that a step of encoding the prediction error calculating a prediction error which is the difference Le. 40. A motion vector predictor according to claim 39.
Decodes the motion vector encoded by the encoding program
Computer that records the motion vector decoding program
Determining a motion compensation mode of a decoding target block on a recording medium readable by a data
And the motion compensation mode of the current block is local motion compensation.
Decoding the prediction error of the motion vector if
Determining the encoding mode of the decoded small block
And the encoding mode of the decoded small block is an intra-frame code.
In the case of the encoding mode, the motion vector of the block is set to 0.
And the encoding mode of the decoded small block is an inter-frame code.
For-coding mode, the motion compensation mode of the decoded sub-block
Determining the over-de, motion compensation mode of the decoded sub-block is a global
In the case of motion compensation, the reference
Calculating a motion vector for each pixel of the lock; and blocking the motion vector in a predetermined manner from the motion vector for each pixel.
Calculating a representative motion vector of the lock; and determining whether the representative motion vector is within a specified range.
And when the representative motion vector is not within the specified range,
Clip the representative motion vector within the specified range
Or adding a prediction error and a prediction vector of the decoding target small block.
Calculation steps and moving-out computer-readable recording medium recording a vector decoding program that have a.