JP2003116141A - Moving picture prediction encoding method and its decoding method, and apparatus for the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a moving picture prediction encoding method and its decoding method, and an apparatus for the same capable of eliminating in-frame and inter-frame correlation of a moving picture and performing efficient encoding and decoding without increasing an attached information quantity. SOLUTION: A block unit motion estimate means 31 estimate a motion of an object in the unit of small blocks in the same way as that by a conventional method to obtain a motion vector. Thus, an attached information quantity required for transmission of motion information is not increased. A correlation coefficient R calculation means 35 calculates an inter-frame correlation coefficient R on the basis of a current frame and a reference frame shifted by the motion vector. A first discrimination branch means compares the coefficient R with a threshold value, selects a motion estimate three-dimensional prediction device 34 when the inter-frame correlation is strong or selects a motion estimate two-dimensional prediction device 33 when the inter-frame correlation is weak. Each prediction device includes a plurality of prediction units and selects the prediction units by each pixel. An addition/subtraction means 37 or 38 generates a prediction residual signal on the basis of the prediction signal and the current frame. The encoded prediction residual signal and motion vector are transmitted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoding method for efficiently compressing a moving image, a decoding method for the same, and an apparatus for them.

[0002]

2. Description of the Related Art J, which is a standard method for lossless encoding of still images
In PEG-LS, a prediction signal is calculated using a predictor,
The difference between the original signal and the predicted signal is encoded. As the predictor, three kinds of two-dimensional (intra-frame) predictors are prepared, and the predictor is switched for each pixel according to the characteristics of the image signal to perform the prediction. By switching the predictor for each pixel, the local properties of the image can be tracked, and compression is performed efficiently.

On the other hand, in a moving image, when there is a motion of the target object, it is necessary to estimate the motion of the target object in order to improve the prediction accuracy. In MPEG1 and MPEG2, which are standard methods of moving image coding, motion compensation, which is one of the inter-frame prediction methods, is used to remove the correlation between frames of image signals. In motion compensation, L
Estimate the movement of the target object in units of small blocks of × L pixels,
A residual signal is generated in block units. At that time, it is necessary to transmit the motion information (called a motion vector) of the object for each block, but since it is a block unit, the amount of additional information does not increase so much.

It should be noted that the prediction accuracy is improved by switching the predictor for each pixel as in the case of a still image, but when the motion of the target object is transmitted for each pixel, the amount of additional information increases and the coding efficiency decreases. .

[0005]

In order to efficiently code a moving image, it is desirable to use the correlation of image signals within and between frames. Predictive coding has a small amount of calculation, and can be adapted to lossless coding by setting the quantization step of the prediction residual signal to 1.

According to the present invention, a predictor is switched for each pixel without increasing the amount of additional information, and the intra-frame and inter-frame correlations are removed from the moving image to perform efficient coding. The problem is to propose a method and its decoding method and their apparatus.

[0007]

In order to solve the above-mentioned problems, the present invention, in an encoding method for moving images, estimates the motion of a target object for each block, and based on the motion estimation result. A plurality of three-dimensional predictors are switched for each pixel to calculate a prediction residual signal, and the prediction residual signal is coded together with the motion estimation result. .

Alternatively, in a decoding method for outputting a moving image, a plurality of three-dimensional predictors are switched for each pixel based on a prediction residual signal encoded by the above moving image predictive encoding method and a motion estimation result. A moving picture predictive decoding method is characterized in that a predictive signal is calculated by the above, the predictive residual signal is added to the predictive signal, and the target signal is decoded.

Alternatively, in a coding method for moving images, two-dimensional prediction or three-dimensional prediction is switched for each pixel depending on the characteristics of the image signal, and when the two-dimensional prediction is switched, the image signal of the image signal is switched. From the characteristics, a plurality of two-dimensional predictors are switched for each pixel to calculate a prediction residual signal, and when switched to the three-dimensional prediction, the motion of the target object in the image signal is motion-estimated for each block. A moving image prediction characterized in that a plurality of three-dimensional predictors are switched for each pixel based on the estimation result to calculate a prediction residual signal, and the calculated prediction residual signal is encoded together with the motion estimation result. The encoding method is used as a means.

Alternatively, in the above moving picture predictive coding method, in the process of switching between two-dimensional prediction and three-dimensional prediction, a correlation coefficient between frames is calculated, and the correlation coefficient between frames is compared with a threshold value. , If the interframe correlation is weak, switch to two-dimensional prediction, and if the interframe correlation is strong, 3
A moving image predictive coding method characterized by switching to dimensional prediction.

Alternatively, in a decoding method for outputting a moving image, two-dimensional prediction is performed for each pixel according to the characteristics of the image signal based on the prediction residual signal encoded by the above-described moving image predictive encoding method and the motion estimation result. Mode or three-dimensional prediction is switched, and when switched to the two-dimensional prediction, a plurality of two-dimensional predictors are switched for each pixel from the characteristics of the image signal to calculate a prediction signal, and the mode is switched to the three-dimensional prediction. In that case, a plurality of three-dimensional predictors are switched for each pixel to calculate a prediction signal,
The moving picture predictive decoding method is characterized by adding the predictive residual signal to the predictive signal to decode the target signal.

Alternatively, in the above-described moving picture predictive decoding method, in the process of switching between two-dimensional prediction and three-dimensional prediction, a correlation coefficient between frames is calculated, and the correlation coefficient between frames is compared with a threshold value. , If the interframe correlation is weak, switch to two-dimensional prediction, and if the interframe correlation is strong, 3
A moving image predictive decoding method characterized by switching to dimensional prediction.

Further, the present invention is also a coding device for a moving picture, wherein the motion estimation of a target object is performed block by block, and a plurality of three-dimensional predictions are made for each pixel based on the motion estimation result. A moving picture predictive coding apparatus characterized in that it has means for calculating a predictive residual signal by switching the units and means for coding the predictive residual signal together with the motion estimation result.

Alternatively, in a decoding device for outputting a moving image, a plurality of three-dimensional predictors are switched for each pixel on the basis of the prediction residual signal encoded by the moving image predictive encoding device and the motion estimation result. And means for calculating the prediction signal,
Means for adding the prediction residual signal to the prediction signal and decoding the target signal.

Alternatively, in an encoding device for a moving image, means for switching between two-dimensional prediction and three-dimensional prediction for each pixel depending on the characteristics of the image signal, and the image when switched to the two-dimensional prediction. Means for calculating a prediction residual signal by switching a plurality of two-dimensional predictors for each pixel based on the characteristics of the signal, and the movement of the target object in the image signal for each block when switched to the three-dimensional prediction. A means for calculating a prediction residual signal by switching a plurality of three-dimensional predictors for each pixel based on the motion estimation result of the motion estimation; and a means for encoding the calculated prediction residual signal together with the motion estimation result. And a moving picture predictive coding device.

Alternatively, in the above moving picture predictive coding device, means for switching between two-dimensional prediction and three-dimensional prediction is
The inter-frame correlation coefficient is calculated, the inter-frame correlation coefficient is compared with a threshold value, and when the inter-frame correlation is weak, the mode is switched to two-dimensional prediction, and when the inter-frame correlation is strong, the mode is switched to three-dimensional prediction. Means is a moving picture predictive coding device characterized by the above.

Alternatively, in a decoding device that outputs a moving image, two-dimensional prediction is performed for each pixel according to the characteristics of the image signal based on the prediction residual signal encoded by the above-described moving image predictive encoding device and the motion estimation result. Means for switching between three-dimensional prediction and three-dimensional prediction, and means for calculating a prediction signal by switching a plurality of two-dimensional predictors for each pixel from the characteristics of the image signal when the two-dimensional prediction is switched. When the mode is switched to prediction, a means for switching a plurality of three-dimensional predictors for each pixel to calculate a prediction signal and a means for adding the prediction residual signal to the prediction signal and decoding the target signal. , And a moving picture predictive decoding device characterized by having.

Alternatively, in the above moving picture predictive decoding apparatus, the means for switching between two-dimensional prediction and three-dimensional prediction is
The inter-frame correlation coefficient is calculated, the inter-frame correlation coefficient is compared with a threshold value, and when the inter-frame correlation is weak, the mode is switched to two-dimensional prediction, and when the inter-frame correlation is strong, the mode is switched to three-dimensional prediction. The moving picture predictive decoding device is characterized by being a means.

In the present invention, motion estimation of the target object is performed in small block units in order to switch the predictor for each pixel without increasing the amount of additional information. The additional information required for transmitting the motion information does not increase because it is performed in small block units as in the conventional method, but since the three-dimensional predictor is used to switch the predictor for each pixel, The coding efficiency is improved.

The three-dimensional (interframe) predictor is effective when the interframe correlation is strong, but on the contrary, when the interframe correlation is weak, the residual signal may be large. Therefore, in the present invention, when the inter-frame correlation is weak, the two-dimensional predictor is used.

[0021]

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

[First Embodiment] FIG. 1 shows the basic configuration of a motion estimation three-dimensional prediction encoder. In FIG. 1, 11 is a block unit motion estimation means, 12 is a shift means, and 13 is 3
The dimension predictor 14 is an adder / subtractor.

A motion estimation three-dimensional prediction method is shown as an example of the operation of FIG. First, the block unit motion estimation means 11
In order to estimate the motion of the target object, motion estimation is performed in small block units of L × L pixels from the current frame and the reference frame. Next, the shift means 12 shifts the signal of the reference frame by the amount of the motion vector obtained as a result of the motion estimation. Next, the three-dimensional predictor 13 performs the three-dimensional prediction by switching the three-dimensional prediction for each pixel using the shifted reference frame and the signal value in the vicinity of the target pixel to be coded of the current frame. Next, the addition / subtraction means 14 calculates a prediction residual signal from the signal of the current frame and the three-dimensional prediction signal. According to this method, since it is possible to follow the local property change of the signal, the prediction accuracy is improved.

A method of encoding using only this motion estimation three-dimensional predictive encoder is also possible. In that case, the prediction residual signal and the motion vector as the additional information may be encoded using an encoding means. Good.

Since the prediction is performed using the shifted signal in the reference frame when performing the three-dimensional prediction, the estimation can be performed more accurately even when the target object is moving. Further, since the motion is estimated for each block, the amount of additional information is the same as in the case of motion compensation used in MPEG1 and MPEG2. A specific method of switching the three-dimensional predictor will be shown in the third embodiment.

[Second Embodiment] FIG. 2 shows the basic arrangement of a motion estimation three-dimensional prediction decoder. In FIG. 2, 21 is a shift means, 22 is a three-dimensional predictor, and 23 is an addition means.

A motion estimation three-dimensional prediction decoding method as an operation example of FIG. 2 is shown. First, the shift unit 21 shifts the reference frame using the motion vector transmitted as the additional information. Next, the three-dimensional predictor 22 generates a prediction signal using the shifted reference frame signal and the decoded signal of the current frame signal (output signal from the adding means 23). Next, the adding means 23 adds the prediction residual signal to this prediction signal and decodes the target signal of the current frame.

It is possible to decode the data coded in the first embodiment by using only this motion estimation three-dimensional prediction decoder. In that case, the coded and transmitted prediction residual signal and motion vector are decoded and used.

[Third Embodiment] FIG. 4 shows a basic configuration for realizing the moving picture predictive coding method according to the present embodiment. In FIG. 4, 31 is a block unit motion estimation means, 32 is a shift means, 33 is a two-dimensional predictor, 34 is a motion estimation three-dimensional predictor, 35 is a correlation coefficient R calculation means, 36.
Is a first judgment branching unit, 37 is an adding / subtracting unit, 38 is an adding / subtracting unit, 39 is a quantizing unit, and 39 is an entropy coding unit.

In the present embodiment, three kinds of two-dimensional predictors are prepared as the two-dimensional predictor 33, and seven kinds of three-dimensional predictors are prepared as the motion estimation three-dimensional predictor 34 therein. Then, the second judgment branching means is built in each of them, and each predictor is switched depending on the property of the image signal. The motion estimation three-dimensional predictor 34 is effective when the interframe correlation is strong, but on the contrary, the residual signal may be large when the interframe correlation is weak. Therefore, in the present embodiment, a method of switching to the two-dimensional predictor 33 when the inter-frame correlation is weak in the first judgment branching means 36 is adopted.

A moving picture predictive coding method is shown as an example of the operation of FIG. First, the block unit motion estimation means 31 estimates the motion of the target object in units of small blocks of L × L pixels from the current frame and the reference frame. Next, the shift means 32 shifts the signal of the reference frame by the amount of the motion vector obtained as a result of the motion estimation. Next, the correlation coefficient R calculation means 35 uses the two-dimensional predictor 3
In order to switch between 3 and the motion estimation 3D predictor 34, the correlation coefficient R of the decoded signal in the vicinity of the target pixel to be coded of the current frame and the reference frame is calculated. First judgment branching means 3
6 indicates that the prediction accuracy is improved when the correlation coefficient R is large, that is, when the waveforms of the current intra-frame signal and the reference intra-frame signal are similar, the motion estimation three-dimensional predictor 34 side And the three-dimensional prediction is performed. In other cases, the two-dimensional predictor 33 is switched to the two-dimensional predictor.

In the case of performing three-dimensional prediction, the motion estimation three-dimensional predictor 34 is internally operated by the second decision branching means incorporated using the shifted reference frame and the signal value in the vicinity of the pixel to be coded of the current frame. The three types of three-dimensional predictors prepared in Section 3 are switched for each pixel to perform three-dimensional prediction. The adder / subtractor 37 calculates a prediction residual signal from the current frame signal and the three-dimensional prediction signal, and outputs the prediction residual signal to the quantizer 39. On the other hand, 2
When performing dimensional prediction, three types of two-dimensional predictors prepared internally are switched for each pixel by the second decision branching means incorporated using the signal value in the vicinity of the pixel to be coded of the current frame. Two-dimensional prediction is performed. The adder / subtractor 38 calculates a prediction residual signal from the current frame signal and the three-dimensional prediction signal, and outputs the prediction residual signal to the quantizer 39.

The prediction method of each predictor and the concrete switching method of the predictors will be described below. The prediction method of each predictor is shown below.

Two-dimensional predictor 0 prediction signal y = min (a, b) (1) Two-dimensional predictor 1 prediction signal y = max (a, b) (2) Two-dimensional predictor 2 prediction signal y = a + bc (3) Three-dimensional predictor 3 Prediction signal y = min (a, x ') (4) Three-dimensional predictor 4 Prediction signal y = max (a, x') (5) Three-dimensional prediction Prediction signal y = a + x'-a (6) Three-dimensional prediction device 6 Prediction signal y = min (b, x ') (7) Three-dimensional prediction device 7 Prediction signal y = max (b, x') (8) Three-dimensional predictor 8 Prediction signal y = b + x'-b (9) Three-dimensional predictor 9 Prediction signal y = (a + b + x ') / 3 (10) a, b, c used here are , As shown in FIG.
It is the decoded value of the upper, left, and upper right pixels adjacent to the pixel x to be encoded. a ′, b ′, x ′ are the decoded values of the pixels of the reference frame, and the motion vector v (iv, jv) between the frame to be encoded and the reference frame (iv: horizontal direction,
jv: vertical direction) to determine the pixel position. If the position of the pixel x to be encoded is (ix, jx), the position of x ′ is (ix + iv, jx + jv). In addition, FIG.
As shown in FIG. 3, a ′ and b ′ are the upper and left pixels adjacent to x ′. As the reference frame, either a temporally forward frame, a backward frame, or both a forward and backward frame is used, but the reference frame needs to be decoded first on the decoder side.

When R≤T0 and c≥max (a, b), the two-dimensional predictor 0 is selected. When R≤T0 and c≤min (a, b), the two-dimensional predictor 1 is selected. R≤T0 and min. (A, b) <c <max (a, b)
In the case of, the two-dimensional predictor 2 is selected When R> T0 and S> T1 and a ′ ≧ max (a, x ′) The three-dimensional predictor 3 is selected R> T0 and S> T1 and a ′ ≦ min (a , X ′) Select the three-dimensional predictor 4 R> T0 and S> T1 and min (a, x ′) <a ′ <
In the case of max (a, x ') Select the three-dimensional predictor 5 R> T0 and S <-T1 and b'≥max (b, x')
In the case of, the three-dimensional predictor 6 is selected R> T0 and S <-T1 and b ′ ≦ min (b, x ′)
In the case of, the three-dimensional predictor 7 is selected R> T0 and S <−T1 and min (b, x ′) <b ′
In the case of <max (b, x '), the three-dimensional predictor 8 is selected. In the case of R> T0 and -T1 <S <T1, the three-dimensional predictor 9 is selected.
However, R = aa '+ bb' + cc '+ dd'-(a + b + c + d) (a '+ b' + c '+ d') (11) S = | x'-b '|-| x'-a' | (12) T0, T1: threshold values a, b, c, d, a ', b', c ', d', x ': FIG.
It is the decoded value of the pixel shown in.

Here, R is a correlation coefficient, and R is a threshold value T
If it is larger than 0, a three-dimensional predictor is selected. Otherwise, the two-dimensional predictor is selected. This selection is actually done in two stages. In other words, the first decision branching means 36 in FIG. 4 makes a first decision branch according to the R value, and first switches between the two-dimensional predictor and the three-dimensional predictor, and the two-dimensional predictor 33 or the three-dimensional predictor 34. A second decision branch that remains in is performed to select one of the predictors 1-9.

The method of switching the three predictors of the two-dimensional predictor has already been proposed and is the same as the method adopted in the international standard JPEG-LS for lossless compression of still images. When it is determined that there is an edge in the vertical direction and the horizontal method, prediction is performed using one pixel adjacent to each edge direction, and in other cases, prediction is performed using three adjacent pixels.

In the case of three-dimensional prediction as well, similar to the two-dimensional predictor, adaptive prediction is performed by switching the predictors depending on the state of the neighboring signal value of the pixel to be coded. When it is determined that there is an edge in the vertical direction or the horizontal direction, prediction is performed using signals adjacent to the respective edge directions of the current frame and the reference frame. The direction of the edge is the absolute difference | x′−a ′ | in the vertical direction of the reference frame and the absolute difference │x′− in the horizontal direction.
b ′ | is compared, and a direction larger than the threshold T1 is determined to be an edge. S in equation (12) is their difference value,
It is a parameter for determining the edge direction. When it is determined that the edge is a vertical edge for determining the edge direction, a prediction signal is selected in the same manner as the two-dimensional predictor with respect to vertical signals of the current frame and the reference frame. The prediction signal is selected in the same manner also in the case of horizontal edges. When it is determined that the pixel is not an edge, the average value of three neighboring pixels is set as the predicted value.

The quantizer 39 quantizes the difference between the prediction signal output from the predictor and the original signal, that is, the prediction residual signal. The quantized residual signal and the motion vector used in the three-dimensional predictor are input to the entropy encoder 40, and the encoded bit stream is output. When the quantization step of the quantizer 39 is set to 1, this coding method is a lossless coding method.

[Fourth Embodiment] FIG. 5 shows a basic configuration for realizing the moving picture predictive decoding method according to the present embodiment. In FIG. 5, 41 is an entropy decoding means,
42 is a dequantization / decoding unit, 43 is a shift unit, 44 is a two-dimensional predictor, 45 is an addition unit, 46 is a motion estimation three-dimensional predictor, 47 is an addition unit, 48 is a correlation coefficient R calculation unit,
49 is a first judgment branching means.

The moving picture predictive decoding method as the operation example of FIG. 5 is as follows. First, in order to determine whether to use the two-dimensional predictor 44 or the three-dimensional predictor 46, the motion vector (decoded through the entropy decoding means 41 and the dequantization unit / decoding unit 42) is used. Then, the correlation coefficient R calculation means 48 calculates the correlation coefficient R from the decoded signal of the reference intraframe signal and the current intraframe signal shifted by the shift means 43. When the correlation coefficient R is larger than the threshold value T0, the first decision branching means 49 switches to the motion estimation three-dimensional predictor 46 side to perform three-dimensional prediction, and in other cases, the two-dimensional predictor 44. Switch to the side and perform two-dimensional prediction. The configurations of the two-dimensional predictor 44 and motion estimation three-dimensional predictor 46 and switching of a plurality of predictors prepared inside are the same as those in the third embodiment. The motion estimation three-dimensional predictor 46 generates a prediction signal by using the decoded signal of the current intraframe signal and the reference intraframe signal shifted by using the motion vector, and the adding means 47 adds the residual signal to this prediction signal. Is added to restore the target signal of the current frame. In the two-dimensional predictor 44, a decoded signal of the signal in the current frame is used to generate a predicted signal, and the adding means 45 adds the residual signal to this predicted signal to restore the target signal of the current frame.

[0042]

As is apparent from the above, according to the present invention, the three-dimensional prediction is switched for each pixel or the three-dimensional prediction is switched to the two-dimensional prediction depending on the property of the image signal of each block. As a result, the predictive coding for efficiently compressing a moving image and its decoding can be realized. In the present invention, when the quantization step size is set to 1, the lossless encoding method is set to 1
If set to a larger value, lossy encoding can be realized.

[Brief description of drawings]

FIG. 1 is a diagram showing a basic configuration of a motion estimation three-dimensional prediction encoder according to a first embodiment example of the present invention.

FIG. 2 is a diagram showing a basic configuration of a motion estimation 3D prediction decoder according to a second exemplary embodiment of the present invention.

3A and 3B are diagrams illustrating a decoded value of a pixel to be encoded.

FIG. 4 is a diagram showing a basic configuration for realizing a moving picture predictive coding method according to a third embodiment of the present invention.

FIG. 5 is a diagram showing a basic configuration for realizing a moving picture predictive decoding method according to a fourth embodiment of the present invention.

[Explanation of symbols] 11 ... Block unit motion estimation means 12 ... Shift means 13 ... Three-dimensional predictor 14 ... Addition / subtraction means 21 ... Shift means 22 ... Three-dimensional predictor 23 ... Addition means 31 ... Block unit motion estimation means 32 ... Shift means 33 ... Two-dimensional predictor 34 ... Motion estimation three-dimensional predictor 35 ... Correlation coefficient R calculation means 36 ... First judgment branching means 37 ... Addition / subtraction means 38 ... Addition / subtraction means 39 ... Quantizer 39 ... Entropy code part 41 ... Entropy decoding means 42 ... Inverse quantizer / decoder 43 ... Shift means 44 ... Two-dimensional predictor 45 ... Addition means 46 ... Motion estimation three-dimensional predictor 47 ... Addition means 48 ... Correlation coefficient R calculation means 49 ... First judgment branching means

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Tomoko Sawabe             2-3-1, Otemachi, Chiyoda-ku, Tokyo             Inside Telegraph and Telephone Corporation F-term (reference) 5C059 KK15 MA04 MA05 MA45 MA47                       ME01 NN28 PP04 TA23 TB08                       TB10 UA02 UA05                 5J064 AA01 BA13 BB03 BC08 BC14                       BC16 BC27 BD02 BD03

Claims (12)

[Claims] 1. A coding method for a moving image, wherein a motion of a target object is motion-estimated for each block, and a plurality of three-dimensional predictors are switched for each pixel based on the motion estimation result to predict residual error. A moving image predictive coding method, comprising: calculating a signal and coding the prediction residual signal together with the motion estimation result. 2. A decoding method for outputting a moving image,
A prediction signal is calculated by switching a plurality of three-dimensional predictors for each pixel based on a prediction residual signal coded by the moving picture predictive coding method according to claim 1 and a motion estimation result. A moving picture predictive decoding method characterized by adding a prediction residual signal and decoding a target signal. 3. A coding method for a moving image, wherein two-dimensional prediction or three-dimensional prediction is switched for each pixel depending on the characteristics of the image signal, and when the two-dimensional prediction is switched, the image signal of the image signal is changed. From the characteristics, a plurality of two-dimensional predictors are switched for each pixel to calculate a prediction residual signal, and when switched to the three-dimensional prediction, the motion of the target object in the image signal is motion-estimated for each block. A moving image prediction characterized in that a plurality of three-dimensional predictors are switched for each pixel based on the estimation result to calculate a prediction residual signal, and the calculated prediction residual signal is encoded together with the motion estimation result. Encoding method. 4. The moving picture predictive coding method according to claim 3, wherein in the process of switching between two-dimensional prediction and three-dimensional prediction, a correlation coefficient between frames is calculated, and the correlation coefficient between frames and the threshold value are calculated. By comparison, when the inter-frame correlation is weak, switching to two-dimensional prediction is performed, and when the inter-frame correlation is strong, 3
A video predictive coding method characterized by switching to dimensional prediction. 5. A decoding method for outputting a moving image,
A two-dimensional prediction or a three-dimensional prediction is switched for each pixel according to the characteristics of the image signal based on the prediction residual signal coded by the moving picture predictive coding method according to claim 3 or 4 and the motion estimation result. When switched to dimensional prediction, a plurality of two-dimensional predictors are switched for each pixel from the characteristics of the image signal to calculate a prediction signal, and when switched to three-dimensional prediction, a plurality of three-dimensional predictors are calculated for each pixel. A moving picture predictive decoding method, characterized in that a predictive signal is calculated by switching a dimensional predictor, and the predictive residual signal is added to the predictive signal to decode a target signal. 6. The moving picture predictive decoding method according to claim 5, wherein in the process of switching between two-dimensional prediction and three-dimensional prediction, a correlation coefficient between frames is calculated, and the correlation coefficient between frames and a threshold value are set. By comparison, when the inter-frame correlation is weak, switching to two-dimensional prediction is performed, and when the inter-frame correlation is strong, 3
A moving image predictive decoding method characterized by switching to dimensional prediction. 7. A coding device for a moving image, wherein a motion of a target object is estimated for each block, and a plurality of three-dimensional predictors are switched for each pixel based on the motion estimation result to perform prediction. A moving picture predictive coding apparatus comprising: a means for calculating a residual signal; and a means for coding the predicted residual signal together with the motion estimation result. 8. A decoding device for outputting a moving image,
Means for calculating a prediction signal by switching a plurality of three-dimensional predictors for each pixel based on a prediction residual signal coded by the moving picture predictive coding apparatus according to claim 7 and a motion estimation result,
And a unit for decoding the target signal by adding the prediction residual signal to the prediction signal. 9. An encoding device for a moving image, a means for switching between two-dimensional prediction and three-dimensional prediction for each pixel according to the characteristics of an image signal, and the image when switched to the two-dimensional prediction. Means for calculating a prediction residual signal by switching a plurality of two-dimensional predictors for each pixel from the characteristics of the signal;
When the mode is switched to the dimensional prediction, a plurality of three-dimensional predictors are switched for each pixel based on the motion estimation result in which the motion of the target object in the image signal is motion-estimated for each block to calculate a prediction residual signal. A moving picture predictive coding apparatus comprising: means and means for coding the calculated prediction residual signal together with the motion estimation result. 10. The moving image predictive coding apparatus according to claim 9, wherein the means for switching between two-dimensional prediction and three-dimensional prediction calculates a correlation coefficient between frames and sets the correlation coefficient between frames and a threshold value. By comparison, when the inter-frame correlation is weak, switching to two-dimensional prediction is performed, and when the inter-frame correlation is strong, 3
A moving picture predictive coding apparatus characterized by being switched to dimensional prediction. 11. A decoding device for outputting a moving image, wherein pixels are selected according to the characteristics of the image signal based on the prediction residual signal and the motion estimation result encoded by the moving image predictive encoding device according to claim 9 or 10. Means for switching between two-dimensional prediction and three-dimensional prediction for each, and means for calculating a prediction signal by switching a plurality of two-dimensional predictors for each pixel from the characteristics of the image signal when switched to the two-dimensional prediction And means for calculating a prediction signal by switching a plurality of three-dimensional predictors for each pixel when switched to the three-dimensional prediction, and decoding the target signal by adding the prediction residual signal to the prediction signal A moving image predictive decoding device, comprising: 12. The moving picture predictive decoding apparatus according to claim 11, wherein the means for switching between two-dimensional prediction and three-dimensional prediction calculates a correlation coefficient between frames and sets the correlation coefficient between frames and a threshold value. By comparison, when the inter-frame correlation is weak, switching to two-dimensional prediction is performed, and when the inter-frame correlation is strong, 3
A moving picture predictive decoding device characterized by switching to dimensional prediction.