JPH08172631A - Image prediction coding method - Google Patents

Abstract

PURPOSE: To prevent deterioration in a predicted image when a large deformation or a large luminance change takes place in an object in the low bit rate image coding. CONSTITUTION: An image series comprising a luminance signal 321 representing a picture element value and a transmission signal 33 representing a transparent state of the picture element is received and stored in an input frame memory 30. A representative image is stored in a representative image frame memory 42. A prediction device 44 predicts an image being a coding object from the representative image representing the image signal to form a 1st predicted image, predicts an image being a coding object from the image displayed early timewise from the image of the coding object to form a 2nd predicted image, obtains a weight mean of the luminance signal and the transmission signal corresponding to the 1st predicted image and the 2nd predicted image respectively to form a 3rd predicted image. An optimum predicted image is selected among the 1st to 3rd predicted images and the selected image is given to an adder 32. A coder 34 encodes the image from the adder 32.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a predictive coding method for storing or transmitting digital image data.

[0002]

[Prior Art] J. Wang and E. Ade
Lson) proposed a method to decompose and encode a moving image into different layers for each object for the purpose of efficiently transmitting and recording the moving image.
and E. Adelson, "Layered Representation for Image
Sequence Coding ", Proc. IEEE Int. Conf. AcousticS
peech Signal Processing, 1993, pp.V221 ~ V224; J.
Wang and E. Adelson, "Layered Representation for Mo
Analysis Analysis ", Proc. Computer Vision and Pattern R
ecognition, pp. 361--366, 1993.

This method will be briefly described below. An example of an image composed of trees, buildings, and the sky, with the trees in the foreground, the buildings behind the trees, and the sky as the background will be described. When the camera moves, trees may obscure buildings and the sky behind them. Wan et al. Have proposed a method that analyzes this series of moving images, extracts trees, buildings, and sky, separates them into three layers, and compresses and encodes each layer. According to this method, a representative image (tree, building, sky) is defined for each layer in order to improve compression efficiency. The representative image is the one closest to all the frames in the series of images.

Wan et al. Obtain the representative image by the following method. First, determine one reference image, detect the parameter indicating how much the other images are displaced and deformed from this reference image, then deform and displace the image using the detected displacement parameters, and fit to the reference image. To do. That is, all fitted objects have the same shape at the same spatial position. Finally, a median filter is applied to all the fitted images to select a median value and form a representative image.

The representative image is compression-coded in advance, and the object in each frame is approximated by displacement or deformation from the reproduced representative image. Specifically, affine transformation is used. Since the affine transformation consists of 6 parameters, the object in one layer can be approximated by 6 parameters.
It is possible to realize a very high compression rate with a small number of bits. The decoding unit reproduces the layer corresponding to each frame, superimposes the objects in each layer in a predetermined order, and synthesizes and displays the image. It is necessary to have a transparency signal that indicates the rate at which the objects in each frame are superimposed.
It is also encoded and transmitted or recorded.

On the other hand, there is a motion-compensated prediction method by cross-reference of image series such as the MPEG1 or MPEG2 standard of the international standardization organization. This is shown in FIG. For example, frame 214 predicts from previous frame 212. Also, like the frame 220, the frame 21 two frames before
It may be predicted from 6. Further, there is a prediction method from the frame 216 that is before the frame 218 and the frame 220 that is after the frame 218, and the frame 218 may be predicted by the average of the frames 216 and 220.

[0007]

However, in the method of approximating an object using the representative image as described above, when the object undergoes a large deformation or a large change in luminance, the error becomes large and the approximation becomes poor. On the other hand, MPEG1 / MP
In the motion-compensated prediction based on the EG2 standard cross-reference, errors are accumulated, and the prediction error increases with the passage of time, leading to deterioration in image quality. In order to reduce this, there is a method to prevent error accumulation by performing intra-frame coding without prediction under a high number of bits, but intra-frame coding requires a high number of bits. It is not suitable for low bit rate transmission and recording.

Further, the contour portion of the predicted image and the target image often do not match, and the difference value increases due to the mismatch of the contour portions, so that the contour portion cannot be efficiently coded.

[0009]

In order to solve the above problems, the image predictive coding method of the present invention uses at least one representative image and an encoding target for an image to be encoded. The image displayed before or after the image temporally is converted under a predetermined conversion method to form a predicted image, and the predicted image with the smallest error is used as the optimal predicted image for the image to be encoded. .

Further, a new predicted image formed by averaging a plurality of predicted images is added as a candidate, and the predicted image with the smallest error in the image to be encoded is used as the optimum predicted image.

Further, the optimum predicted image is divided into a plurality of predicted small regions, the image to be coded is divided into a plurality of target small regions, and at least one target small region containing a pixel value which should not be coded corresponds to the target small region. For the prediction small area, the pixel value to be encoded included in the prediction small area is calculated by a predetermined function, and the calculated value is substituted for the pixel value that should not be encoded included in the target small area and the prediction area, The difference between the target small area and the predicted small area is calculated and encoded.

[0012]

The present invention adaptively selects at least one representative image and an image displayed before or after the image to be coded, thereby causing a large deformation or a large luminance change in the object. In this case, it is possible to prevent the deterioration of the approximation and reduce the accumulation of the prediction error over time. Further, since it is not necessary to periodically insert the intra-frame coded frame without prediction, it is possible to perform transmission and recording at a low bit rate. Furthermore, by averaging the two predicted images, the distortion signals included in the predicted images are also averaged, and the distortion can be reduced.

Further, before dividing the optimum predicted image and the image to be coded into a plurality of prediction small areas and a target small area, respectively, and obtaining a difference between the corresponding predicted small area and the target small area, at least one of them is calculated. For a predicted small area corresponding to a target small area containing a pixel value that should not be coded, the pixel value to be coded included in the predicted small area is calculated by a predetermined function, and the calculated value is predicted as the target small area. By substituting the pixel values that should not be encoded included in the area, the difference value becomes small and the number of encoded bits can be reduced.

[0014]

EXAMPLES The present invention will be described below with reference to specific examples.

The input image of the present invention is composed of a luminance signal representing a pixel value and a transparency signal representing a transmissive state of each pixel value. The same applies when a color difference signal is included in addition to the luminance signal. The transparency signal is a signal indicating at what ratio the pixel values of each image are combined when a plurality of images are combined. The transparency of 0% means that there is an opaque object and the background cannot be seen, and the pixel value should be encoded. A transparency of 100% means that the background is visible as it is without any object, and the pixel value does not need to be encoded. The intermediate level value appears
This is the case where a translucent object such as glass exists and the boundary between the objects. FIG. 1 shows an example of the luminance signal 31 and the transparency signal 33. Hereinafter, unless otherwise specified, an image is composed of a luminance signal and a transparency signal, and the pixel value means the value of the luminance signal.

1 to 3 are block diagrams of embodiments of the image predictive coding method of claims 1 to 6, and schematic diagrams of embodiments of the image predictive coding method of the present invention. As shown in FIG.

As shown in FIG. 1, first, the luminance signal 31 and the transparency signal 33 are input from the terminal 46, and these signals are stored in the input frame memory 30. The luminance signal and the transparency signal input to the input frame memory 30 are input to the predictor 44. Based on these signals, the predictor 44 displays the reproduced image stored in the reproduced image frame memory 40 and the representative image. An optimum predicted image is generated from the representative image stored in the image frame memory 42. Next, the input image output from the input frame memory 30 and the predictor 4
The difference image is obtained by the adder 32 from the prediction image output from No. 4, and this image is encoded by the encoder 34. In the encoder 34, the difference image is divided into blocks, and each block is transformed by discrete cosine transform (DCT) or other orthogonal transform and quantized. The output of the encoder 34 is terminal 4
8 and at the same time the local decoder 36
Inverse quantization and inverse DCT are performed. This inversely transformed difference image is added by the adder 38 to the predicted image from the predictor 44 to generate a reproduced image, and the reproduced image frame memory 40 is generated.
To be stored. Note that we encode not only the luminance signal but also the transparency signal here. In this embodiment, the transparency signal is DCT-converted like the luminance signal, but may be converted by a different conversion method. Note that the encoder 34 may perform wavelet transform or vector quantization instead of DCT. However, the local decoder 36 must perform the corresponding inverse transform.

The representative image is the one closest to all the frames in the series of images. For one image series,
A plurality of representative images may be provided. Further, the representative image may be obtained by the same method as the one described in the prior art. The representative image also has a luminance signal and a transparency signal, and is compression encoded in advance and transmitted, and the reproduced representative image is stored in the representative image frame memory 42. Further, the representative image does not have to be a complete image, and may be a part (eg, ear, eye, mouth) in the image. Further, the representative image may be stored in the receiving unit in advance without being transmitted.

The operation of the predictor 44 will be described below with reference to FIGS. 2 and 4. First, consider the case of predicting the frame 14 shown in FIG. 4 (corresponding to claim 1). The frame 14 is an image to be encoded, the frame 10 is a representative image, and the frame 12 is an image displayed temporally before the frame 14. These images are input to the predictor of FIG. That is, the frame 14 (luminance signal and transparency signal) that is the image to be encoded is input to the terminal 72, the representative image 10 is input to the terminal 68, and the frame 12 that is the reproduced image is input to the terminal 70. The frame 12 is the reproduced image frame memory 4 of FIG.
It is a decompressed reproduction stored in 0.

Target image (frame 14) and representative image 10
Is input to the deformation / displacement calculator 50. The deformation / displacement calculator 50 obtains a deformation / displacement parameter that allows the representative image to most approximate the target image. Although the affine transformation is preferably used to obtain the deformation / displacement parameter, a transformation including a quadratic term may be used. Although the affine transformation is known, it is disclosed in, for example, the above-mentioned Wan et al. The affine coefficient calculated by the deformation / displacement calculator 50 is input to the predicted image generator 52 together with the representative image, and the first predicted image (luminance signal and transparency signal) is generated by affine transformation. The first predicted image and the target image are input to the error calculator 54, the sum of the first squared error is calculated only for the pixel value to be encoded, and the sum is output to the comparator 66.

Similarly, the deformation / displacement calculator 56 calculates the affine coefficient from the reproduced image (frame 12) and the target image (frame 14), and the predicted image generator 58 generates a second predicted image. The error calculator 60 calculates the sum of the second squared errors and outputs the sum to the comparator 66. The comparator 66 compares the sum of the first and second squared errors and controls the switch 76 to output the predicted image corresponding to the smaller one.
That is, for example, when the error of the first predicted image is small, the switch 76 connects to the terminal 82 and outputs the first predicted image to the terminal 74.

Further, the first predicted image and the second predicted image are input to the averaging unit 62, and the weighted averages of the corresponding luminance signal and transmission signal are respectively obtained to generate the third predicted image. The error calculator 64 calculates the sum of the third squared error from the third predicted image and the target image, and the comparator 6
6 is output. The comparator 66 compares the sums of the first, second and third squared errors and controls the switch 76 so as to output the predicted image corresponding to the smaller one (corresponding to claim 2). Although the error calculator calculated the sum of squared errors,
You may obtain the sum of the absolute value of an error.

Next, consider the case of predicting the frame 18 shown in FIG. 4 (corresponding to claim 3). The frame 18 is an image to be encoded, the frame 10 is a representative image, and the frame 20 is an image displayed temporally after the frame 18. These images are input to the predictor of FIG. That is, the frame 18 (luminance signal and transparency signal), which is the image to be encoded, is connected to the terminal 72, the representative image 10 is connected to the terminal 68,
The frame 20 which is a reproduced image is input to the terminal 70. The frame 20 is a decompressed image stored in the reproduced image frame memory 40 of FIG. 1, and is a compressed image that is obtained by adaptively selecting a predictive image from the frame 16 and the representative image 10 by the method described above.

In the same manner as described above, the first predicted image (luminance signal and transparency signal) is generated from the target image and the representative image via the deformation / displacement calculator 50 and the predicted image generator 52. It The first predictive image and the target image are input to the error calculator 54, and the sum of the first squared error is calculated only for the pixel value to be encoded and output to the comparator 66. Similarly, a second predicted image is generated from the frame 20 and the target image via the deformation / displacement calculator 56 and the predicted image generator 58, and the error calculator 60 calculates the second squared error of the second squared error. The sum is calculated and output to the comparator 66. The comparator 66 compares the sum of the first and second squared errors and controls the switch 76 to output the predicted image corresponding to the smaller one.
That is, for example, when the error of the first predicted image is small, the switch 76 connects to the terminal 82 and outputs the first predicted image to the terminal 74.

Further, the first predicted image and the second predicted image are input to the averaging unit 62, and the weighted averages of the corresponding luminance signal and transmission signal are respectively obtained to generate the third predicted image. The error calculator 64 calculates the sum of the third squared error from the third predicted image and the target image, and outputs the sum to the comparator 66. The comparator 66 compares the sum of the first, second and third squared errors, and controls the switch 76 so as to output the predicted image corresponding to the smaller one (corresponding to claim 4). Although the error calculator calculated the sum of squared errors,
You may obtain the sum of the absolute value of an error.

Next, consider the case of predicting the frame 22 shown in FIG. 4 (corresponding to claim 5). The frame 22 is an image to be encoded, the frame 10 is a representative image, the frame 20 is an image displayed temporally before the frame 22, and the frame 24 is an image displayed temporally after the frame 22. These images are input to the predictor 44 of FIG. That is, the frame 22 (luminance signal and transparency signal) is input to the terminal 72, the representative image 10 is input to the terminal 68, the frame 20 is input to the terminal 70, and the frame 24 is input to the terminal 86. The frame 20 is the frame 16 and the frame 10.
The prediction image is adaptively selected, compression coded, and reproduced, and the frame 24 is a prediction image adaptively selected, compressed, coded, and reproduced from the frame 20 and the frame 10. Similarly to the above, the first predicted image (luminance signal and transparency signal) is generated from the target image and the representative image via the deformation / displacement calculator 50 and the predicted image generator 52. This first predicted image and the target image are the error calculator 5
4, the sum of the first squared errors is calculated only for the pixel value to be encoded, and the sum is output to the comparator 66. Similarly, a second predicted image is generated from the frame 20 and the target image via the deformation / displacement calculator 56 and the predicted image generator 58, and the error calculator 60 calculates the second squared error. The sum is calculated and output to the comparator 66. Further, a third predicted image is generated from the frame 24 and the target image via the deformation / displacement calculator 88 and the predicted image generator 90, and the error calculator 92 calculates the sum of the third squared errors. Obtained and output to the comparator 66. The comparator 66 compares the sum of the first, second and third squared errors, and controls the switch 76 so as to output the predicted image corresponding to the smaller one.

Further, the second predicted image and the third predicted image are input to the averaging unit 62, and the weighted average of the corresponding luminance signal and transmission signal is obtained to generate the fourth predicted image. The error calculator 64 calculates the sum of the fourth squared errors from the fourth predicted image and the target image, and outputs the sum to the comparator 66. The comparator 66 compares the sums of the first, second, third and fourth squared errors, and controls the switch 76 so as to output the predicted image corresponding to the smaller one (corresponding to claim 6). ). Although the error calculator calculated the sum of squared errors,
You may obtain the sum of the absolute value of an error.

Next, an embodiment corresponding to claim 7 will be described with reference to FIGS. 5 and 6. FIG. 5 is almost the same as the encoding device of FIG. 1, but only the input side of the adder 32 is different. That is, the operations of blocking and pixel substitution are incorporated before the difference is obtained by the adder 32.

First, the output (luminance signal and transparency signal) of the input frame memory 30 is divided into small areas by the blocker 41. Here, the block is divided into blocks of 4 × 4 pixels, but the present invention is not limited to this. The output of the predictor 44 (luminance signal and transparency signal)
Is also divided into 4 × 4 blocks by the blocker 35. The output of the blocker 35 consists of a block of luminance signals and a block of transparency signals. The block of the luminance signal and the block of the transparency signal are input to the pixel value generator 37. In the pixel value generator 37, pixels having a transparency other than 100% (corresponding to the pixel to be encoded, that is, the transparency signal 33
(Corresponding to the black portion of) is calculated by a predetermined function to generate an assigned pixel value. Here, the average value of all pixels having a transparency other than 100% is used, but the present invention is not limited to this. This substitution pixel value is input to the pixel substitution devices 39 and 43, and the substitution pixel value is substituted for the pixel value (of the luminance signal) with the transparency of 100%, and the block of the luminance signal output from the pixel substitution devices 39 and 43. Is calculated by the adder 32. The difference signal is obtained without substituting the block value of the transparency signal.

FIGS. 6A to 6C show an example of one numerical value. FIG. 6A shows a block of the transparency signal.
Block 110 is the output of blocker 41 and block 112 is the output of blocker 35. Block 11
At 0 and block 112, pixels with a value of 1 indicate pixel values that should not be encoded (100% transparency). In order to store the transparency information, the blocks 110 and 112 are directly subtracted to generate a block 114, and the encoder 3
Send to 4. On the other hand, FIG. 6B shows a block of the luminance signal. Block 116 is the output of blocker 41 and block 118 is the output of blocker 35. A block 120 is obtained by taking the difference between these two blocks as they are. A large difference value occurs in the block 120 because the contours of the pixels to be encoded do not match. According to the present invention, the block 112 and the block 118 are input to the pixel value generator 37, and the average value of the pixel values of the block 118 whose value corresponds to zero is calculated in the block 112. In this numerical example, the average value is 48. The substitution pixel value 48 is input to the pixel substitution devices 43 and 39, and the numerical value 48 is substituted into the pixel corresponding to the value of the transparency signal of 1 is block 1 respectively.
22 and 124. When the difference between these two blocks is obtained, block 126 is obtained, and it can be seen that the difference value is smaller than that of block 120. When the block 126 is input to the encoder 34 instead of the block 120 and is encoded, the number of required bits is reduced. On the reproducing side, the substituted pixel value can be restored based on the transparency information.

An embodiment corresponding to claims 8 to 14 will be described below. In the above-described embodiment, the predicted image is adaptively selected and encoded for the entire area of the image. In this embodiment, instead of the entire area of the image, the area to be encoded is divided into a plurality of small areas, and the optimum predicted small area is adaptively selected and encoded for each small area. The small region is preferably a 16 × 16 or 8 × 8 block, but may be divided into small regions of arbitrary size and shape. The block diagram to which this embodiment can be applied is the same as that of FIG.
It is necessary to insert a blocker in front of and divide the input image into blocks before inputting. The predictor 44 is the same as in FIG. 2 or FIG. 3, but in this case the affine coefficient must be obtained for each block. It should be noted that conversion such as rotation may be stopped and replaced by motion detection and motion compensation based on simple parallel movement. The embodiment of claim 8 is the embodiment of claim 1, the embodiment of claim 9 is the embodiment of claim 2, the embodiment of claim 10 is the embodiment of claim 3, and the embodiment of claim 11. An example is the embodiment of claim 4, an embodiment of claim 12 is the embodiment of claim 5, an embodiment of claim 13 is the embodiment of claim 6, and an embodiment of claim 14 is the embodiment of claim 7. Corresponding to the example of
The details are the same except for the prediction in block units, and thus detailed description thereof will be omitted.

[0032]

As is clear from the above description, according to the image predictive coding method of the present invention, at least one representative image and an image displayed temporally before or after the image to be coded are displayed. By adaptively selecting, even when a large deformation or a large brightness change occurs in the object, it is possible to prevent the approximation from being deteriorated and reduce the accumulation of prediction errors over time. Further, since it is not necessary to periodically insert the intra-coded frame without prediction, it is possible to perform transmission and recording at a low bit rate. Furthermore, by averaging the two predicted images, the distortion signals included in the predicted images are also averaged, and the distortion can be reduced.

Further, at least one of the optimum predicted image and the image to be coded is divided into a plurality of prediction small areas and a target small area, and the difference between the corresponding predicted small area and the target small area is calculated. For a predicted small area corresponding to a target small area containing pixel values that should not be coded, the pixel value to be coded included in the predicted small area is calculated by a predetermined function, and the calculated value is set as the target small area. By substituting the pixels in the prediction region that should not be encoded, the difference value becomes smaller and the number of encoded bits can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of an apparatus to which an image coding method of the present invention can be applied.

FIG. 2 is a block diagram of an embodiment of a predictor according to the present invention.

FIG. 3 is a block diagram of another embodiment of the predictor according to the present invention.

FIG. 4 is a schematic diagram showing a prediction method of the present invention.

FIG. 5 is a block diagram of another embodiment of an apparatus to which the image coding method of the present invention can be applied.

FIG. 6 is a diagram for explaining the effect of difference value reduction by substituting pixel values according to the present invention.

FIG. 7 is a schematic diagram of a conventional prediction method.

[Explanation of symbols]

 30 Input Frame Memory 31 Luminance Signal 32 Adder 33 Transparency Signal 34 Encoder 36 Local Decoder 38 Adder 40 Playback Image Frame Memory 42 Representative Image Frame Memory 44 Predictor

Claims (14)

[Claims] 1. An image sequence composed of a luminance signal representing a pixel value and a transparency signal representing a transmissive state of the pixel value should be encoded by the transparency signal of an image to be encoded. For all regions, the image to be encoded is predicted from at least one representative image representing the image sequence to generate a first predicted image, and the image to be encoded is temporally preceded by the image. The image to be encoded is predicted from the displayed image to generate a second predicted image, and from the first and second predicted images, the one with a smaller error in the image to be encoded is selected. An image predictive coding method characterized by selecting as an optimum predicted image. 2. The first and second predicted images are weighted and averaged to generate a third predicted image, and the first and second predicted images are generated.
2. The image predictive coding method according to claim 1, wherein one of the second and third predicted images having a smaller error in the image to be coded is selected as the optimum predicted image. 3. An image sequence composed of a luminance signal representing a pixel value and a transparency signal representing a transparency state of the pixel value, to be encoded specified by the transparency signal of an image to be encoded. For all regions, the image to be coded is predicted from at least one representative image that represents the image sequence to generate a first predicted image, and is displayed later in time than the image to be coded. The image to be encoded is predicted from the image to be generated, and a second predicted image is generated. From the first and second predicted images, the image with the smaller error is optimal for the image to be encoded. An image predictive coding method characterized by selecting as a predicted image. 4. The first and second predicted images are weighted and averaged to generate a third predicted image, and the first and second predicted images are generated.
4. The image predictive coding method according to claim 3, wherein one of the second and third predicted images having a smaller error in the image to be coded is selected as the optimum predicted image. 5. An image sequence composed of a luminance signal representing a pixel value and a transparency signal representing a transmissive state of the pixel value should be encoded by the transparency signal of an image to be encoded. For all regions, the image to be encoded is predicted from at least one representative image representing the image sequence to generate a first predicted image, and the image to be encoded is temporally preceded by the image. The image to be coded is predicted from the displayed image to generate a second predicted image, and the image to be coded is selected from the image displayed after the image to be coded temporally. , A third predicted image is generated, and one of the first, second and third predicted images having a smaller error in the image to be coded is selected as the optimum predicted image. Image predictive coding method . 6. The second and third predicted images are weighted and averaged to generate a fourth predicted image, and the first and second predicted images are generated.
The image predictive coding method according to claim 5, wherein one of the second, third and fourth predicted images having a smaller error in the image to be coded is selected as an optimum predicted image. 7. An optimum predicted image is divided into a plurality of predicted small areas, an image to be coded is divided into a plurality of target small areas, and a target small area containing at least one pixel value that should not be coded. For a corresponding prediction small area, a pixel value to be coded included in the prediction small area is calculated by a predetermined function, and the calculated value is assigned to a pixel that should not be coded included in the target small area and the prediction area. The image predictive coding method according to any one of claims 1 to 6, wherein the difference between the target small area and the predicted small area is obtained and then coded. 8. An image sequence composed of a luminance signal representing a pixel value and a transparency signal representing a transparency state of the pixel value should be encoded by the transparency signal of an image to be encoded. The entire area is divided into a plurality of target small areas, the target small area is predicted from at least one representative image representing the image series, and a first predicted small area is generated, and the image to be encoded is the image. The target small area is predicted from an image displayed earlier in time, a second predicted small area is generated, and the target small area has a small error from the first and second predicted small areas. An image predictive coding method, characterized in that one is selected as the optimum prediction small area. 9. The first and second prediction sub-regions are averaged by weighting to generate a third prediction sub-region, and the target sub-regions are extracted from the first, second and third prediction sub-regions. 9. The image predictive coding method according to claim 8, wherein one of the regions having a smaller error is selected as the optimum prediction small region. 10. An image sequence composed of a luminance signal representing a pixel value and a transparency signal representing a transmissive state of the pixel value should be encoded by the transparency signal of an image to be encoded. The entire area is divided into a plurality of target small areas, the target small area is predicted from at least one representative image representing the image series, and a first predicted small area is generated, and the image to be encoded is the image. The target small area is predicted from an image displayed later in time, a second predicted small area is generated, and the one having a smaller error from the first and second predicted small areas to the target small area. An image predictive coding method, characterized in that is selected as the optimum prediction small area. 11. The first and second prediction sub-regions are weighted and averaged to generate a third prediction sub-region, and the target sub-regions are extracted from the first, second and third prediction sub-regions. 11. The image predictive coding method according to claim 10, wherein one having a smaller error in the area is selected as the optimum prediction small area. 12. An image sequence composed of a luminance signal representing a pixel value and a transparency signal representing a transparency state of the pixel value should be encoded by the transparency signal of an image to be encoded. The entire area is divided into a plurality of target small areas, the target small area is predicted from at least one representative image representing the image series, and a first predicted small area is generated, and the image to be encoded is the image. The target small area is predicted from an image displayed earlier in time, a second predicted small area is generated, and the target small area is displayed from an image displayed later in time than the image to be encoded. Predict a region, generate a third prediction small region, and select one of the first, second and third prediction small regions with a smaller error in the target small region as the optimum prediction small region. Characteristic image predictive coding method. 13. The second and third prediction sub-regions are averaged by weighting to generate a fourth prediction sub-region, and from the first, second, third and fourth prediction sub-regions, The image predictive coding method according to claim 12, wherein one of the target small areas having a smaller error is selected as the optimum prediction small area. 14. A pixel function to be coded included in the optimum prediction small area is defined by a predetermined function with respect to the optimum prediction small area corresponding to the target small area including at least one pixel value that should not be coded. The calculation is performed, the calculated value is substituted into a pixel value that should not be encoded included in the target small area and the optimum prediction area, and then the difference between the target small area and the optimum prediction small area is obtained and encoded. The image predictive coding method according to any one of claims 8 to 13.