JPH0974567A - Moving image encoding/decoding method and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve encoding efficiency of arbitrary shaped data contained in the inter-frame predicted error signal of a moving image. SOLUTION: When the boundary 105 of a signal level is extracted by using the image 103 of a previous frame, an intra-frame predicted image 123 is generated by using a motion compensation, the location of the moved boundary is detected in the predicted image, and the inter-frame predicted error 112 with the present frame 101 is encoded on the side of an encoding, the part in which the predicted error is large is divided into an arbitrary shape, a picture element width is designated from a starting point, a terminating point and a picture element width or the starting point and the location of detected boundary, the picture element contained in the image width is encoded, and information necessary for the motion compensation, the information on the starting point, the information on the picture element width, encoded data and the information on the terminating point if necessary are transmitted to the side of a decoding. On the side of the decoding, encoded data 119 is decoded from the image 208 of the previous frame by the entirely same technique as that of the side of the encoding.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving image encoding / decoding method.

[0002]

2. Description of the Related Art In a conventional motion-compensated interframe coding method, which is a typical high-efficiency coding method for moving pictures, when coding a prediction error, the picture is divided into small blocks of a certain size, and Two-dimensional discrete cosine transform (Di
screte Cosine Transform, DCT)
The transform coefficient is quantized, and the quantization number is entropy coded. At this time, all the quantization information of the block is 0.
It is determined whether or not there is no effective quantized coefficient in block units (or in macroblock units) in a block (or macroblock in which several blocks are combined into one) in which all quantum coefficients are 0. The information given was given.

[0003]

In the conventional coding method, the motion-compensated inter-frame prediction error signal is quantized to a level other than level 0 in units of blocks to be transformed or macroblock units in which several of them are collected. There is no method for efficiently indicating the information indicating the area to be coded and transmitted when the motion compensation inter-frame prediction error signal is divided into arbitrary shapes and coded. .

It is an object of the present invention to provide a moving picture coding / decoding method and apparatus for improving the coding efficiency of data of an arbitrarily shaped area included in an inter-frame prediction error signal of a moving picture.

[0005]

MEANS FOR SOLVING THE PROBLEMS Video coding according to the present invention
On the encoding side, the decoding method uses the decoded frame at the previous time to extract the boundary of the signal level or texture,
An inter-frame prediction image is generated by using motion compensation, a position of a boundary moved in the prediction image is detected, and a part with a large prediction error is detected when encoding an inter-frame prediction error with the current frame. Divide into arbitrary shapes, specify the pixel width from the start point and end point and pixel width, or the position of the boundary detected as the start point, encode the pixels included in those widths, information necessary for motion compensation and start point information The pixel width information, encoded data, and end point information, if necessary, are transmitted to the decoding side, and the decoding side uses the same method as the encoding side from the decoded frame at the previous time to demarcate the signal level or texture boundary. , The information necessary for motion compensation is decoded, the inter-frame prediction image is generated using the motion compensation, the position of the boundary moved in the prediction image is detected, and the inter-frame prediction error of an arbitrary shape is calculated. When decrypting Information and leave decode the information of the end point as required, characterized by decoding the encoded data based on the detected position and the decoded pixel width information of the border.

The encoding apparatus of the present invention inputs a frame memory for storing the image of the previous frame, an image of the current frame and the image of the previous frame stored in the frame memory,
A motion compensation unit that detects a motion vector and generates an inter-frame predicted image, and a boundary information extraction unit that detects a signal level or texture boundary from the image of the previous frame stored in the frame memory and outputs it as boundary information. , The boundary information is motion compensated using the motion vector,
A boundary prediction unit that outputs boundary prediction data, a subtracter that inputs the current frame image and the interframe prediction image, and outputs an interframe prediction error, and the interframe prediction error and the boundary prediction data Input, collect a large portion of the inter-frame prediction error, if the region includes the prediction data of the boundary, the starting point of the boundary and the pixel width, if the region does not start from the end point of the boundary, A coding region determination unit that outputs a start point and an end point of a region and a pixel width as arbitrary shape designation information, and outputs a prediction error signal of an arbitrary shape, and inputs the arbitrary shape designation information and the prediction error signal of the arbitrary shape. , An arbitrary shape analysis filter bank processing unit that outputs an arbitrary shape analysis filter bank coefficient, and quantizes and encodes the arbitrary shape analysis filter bank coefficient, and outputs encoded data Quantization coding unit, coding region decoding unit that inputs the arbitrary shape designation information and the prediction data of the boundary and reproduces arbitrary shape information, and dequantization decoding unit that returns the coded data to a quantized value An arbitrary shape synthesis filter bank processing unit that performs, on the quantized value, an arbitrary shape synthesis filter bank process for each shape based on the arbitrary shape information, and outputs a quantized inter-frame prediction error image;
It has an adder which adds the inter-frame prediction image to the quantized inter-frame prediction error image and stores it in the frame memory as a locally decoded image.

The encoding area determination unit compares the inter-frame prediction error with a threshold value, and outputs data of a portion having a large error equal to or larger than the threshold value, and a portion having the large error. Data is checked against the predicted data of the boundary, and the latter is checked to see if it contains the former. If the latter is included, it is distributed as the data containing the boundary, and if it is not included, it is distributed as the data containing the boundary and output. A collation unit for inputting data of a portion including the boundary, outputting prediction error information and a starting point and a pixel width of the portion including the boundary, a processing unit of the portion including the boundary, and data of a portion not including the boundary , And outputs the prediction error information, the start point / end point, and the pixel width of the part that does not include the boundary, the processing part of the part that does not include the boundary, the prediction error information of the part that includes the boundary, and the A first integration unit that integrates prediction error information of a part and outputs it as the prediction error signal of the arbitrary shape, integrates the start point, the pixel width, the start point / end point, and the pixel width, and outputs as the arbitrary shape designation information. And a second integration unit.

The decoding apparatus of the present invention comprises an inverse quantization decoding unit for returning the encoded data of the encoding apparatus to a quantized value, and an arbitrary shape synthesis filter bank for each shape based on the quantized value and arbitrary shape information. An arbitrary shape synthesis filter bank processing unit that performs processing and outputs a quantized inter-frame prediction error image, a frame memory that stores the image of the previous frame, and an image of the previous frame that is stored in the frame memory are input, A boundary information extraction unit that detects a signal level or texture boundary and outputs it as boundary information; and a boundary prediction unit that performs motion compensation on the boundary information using the motion vector of claim 2 to obtain boundary prediction data, 3. The arbitrary shape designation information according to claim 2 and the prediction data of the boundary are input, and the starting point of the arbitrary shape or the boundary start point and the boundary obtained from the prediction data of the boundary Using the shape end point and the information of the pixel width, the coding area decoding unit that outputs the arbitrary shape information describing the area where the coded data is to be arranged, and the image of the previous frame stored in the frame memory, A motion compensation unit that performs motion compensation using the motion vector and outputs an interframe prediction image, and adds the interframe prediction image to the quantized interframe prediction error image, and outputs the decoded image as well as the frame. It has an adder that stores it in memory.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention detects a sharp boundary of image data by using a technique such as an edge filter or texture analysis when encoding a motion-compensated inter-frame prediction error signal, and detects a detected boundary. A part to be coded based on the position and a part to be coded irrespective of the boundary are separated, and a part having a large prediction error is the object of coding.

In order to correctly decode the encoded data,
It is necessary for the decoding side to perform exactly the same processing as the encoding side. For that purpose, it is usually necessary to transmit boundary information to the decoding side. However, since the amount of this information is huge, the coding efficiency decreases. Therefore, the data of the decoded image of the previous frame used for motion-compensated interframe prediction is used. For the decoded image of the previous frame, a sharp change in the image data is detected by using a technique such as an edge filter or texture analysis, and this is used as a boundary (FIG. 3). Data for predicting the boundary of the current frame is created by motion-compensating the image (FIG. 4).

Of the motion-compensated inter-frame prediction error signals,
A part with a large prediction error is divided into arbitrary shapes (FIG. 5).
Image signals of arbitrary shape are classified into those that are clustered around the boundary and those that are isolated regardless of the boundary. Furthermore, after intra-frame prediction such as DPCM or processing by an arbitrary shape DCT is performed on these regions, the obtained data is quantized. When the level obtained as a result of the quantization is not 0, the starting point and the pixel width are specified for the data around the boundary (FIG. 6 (B)), and the starting point is specified for the part not related to the boundary. The end point and the pixel width are designated (FIG. 6 (A)). The position information is transmitted to the decoding side together with the encoded data of the inter-frame prediction error signal of arbitrary shape.

On the decoding side, the signal level or texture boundary is extracted from the decoded frame at the previous time by the same method as on the encoding side, an inter-frame predicted image is generated using motion compensation, and The position of the boundary moved by is detected.

On the decoding side, the coded data of the inter-frame prediction error signal of arbitrary shape is decoded and placed in the prediction error image based on the information of the start point and the pixel width or the position information of the start point, the end point and the pixel width.

Here, the motion compensation can be combined from the simplest block matching type to the complex one such as projection transformation and affine transformation.

The arbitrary shape encoding may be not only DPCM or arbitrary shape DCT, but also conversion such as generating a base from only shape information or a filter bank. Also, including arbitrary shape transformation,
An analysis / synthesis filter bank of any shape, which is an extension of the above, can be used.

According to the present invention, the boundary information extracted from the previous frame is motion-compensated to specify an arbitrary shape of the motion-compensated inter-frame prediction error that includes a portion having a large prediction error. Is predicting. In order to specify the arbitrary shape, the starting point of the arbitrary shape is restricted and specified on the boundary line, and further the pixel width is specified. This allows
It is not necessary to encode and transmit the prediction error in the entire area of one frame. In addition, the prediction error of a region having no starting point on the boundary may be large. For this, the shape can be specified by specifying the coordinates of the start point and the end point and the pixel width.

[0017]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1A is a block diagram of an encoding apparatus according to an embodiment of the present invention.

The frame memory 102 stores the image 103 of the previous frame. The motion compensation unit 106 inputs the image 101 of the current frame and the image 103 of the previous frame stored in the frame memory 102, detects the motion vector 107,
The inter-frame predicted image 108 is generated. The boundary information extraction unit 104 detects the boundary of the signal level from the image 103 of the previous frame stored in the frame memory 102, and outputs it as boundary information 105. The boundary prediction unit 109 performs motion compensation on the boundary information 105 using the motion vector 107, and outputs boundary prediction data 110. The subtractor 111 inputs the image 101 of the current frame and the inter-frame prediction image 108, and outputs the inter-frame prediction error 112. The coding area determination unit 113 inputs the inter-frame prediction error 112 and the boundary prediction data 110, and inputs the inter-frame prediction error 11
If the area includes the prediction data 110 of the boundary, the start point and the pixel width of the boundary are set. If the area does not start from the end point of the boundary, the start point and the end point and the pixel of the area are set. The width is output as the arbitrary shape designation information 115, and the prediction error signal 114 of the arbitrary shape is output. The arbitrary shape DCT unit 116 inputs the arbitrary shape designation information 115 and the arbitrary shape prediction error signal 114, and outputs an arbitrary shape DCT coefficient 117. The quantization coding unit 118 quantizes and codes the arbitrarily shaped DCT coefficient 117 to generate the coded data 1
19 is output. The coding area decoding unit 126 inputs the arbitrary shape designation information 115 and the boundary prediction data 110, and reproduces the arbitrary shape information 127. The inverse quantization decoding unit 120 returns the encoded data 119 to the quantized value 121. Arbitrary shape I
The DCT unit 122 converts the quantized value 121 into arbitrary shape information 127.
An arbitrary shape IDCT is performed for each shape based on the above, and a quantized inter-frame prediction error image 123 is output. The adder 124 adds the inter-frame prediction image 108 to the quantized inter-frame prediction error image 123, and outputs the locally decoded image 12
5 is stored in the frame memory 102.

FIG. 1B is a block diagram of a decoding apparatus according to an embodiment of the present invention.

The inverse quantization decoding unit 201 uses the encoded data 11
9 is returned to the quantized value 202. Arbitrary shape IDCT unit 203
Is the quantized value 202 based on the arbitrary shape information 216,
Arbitrary shape IDCT is performed for each shape, and the quantized inter-frame prediction error image 204 is output. Frame memory 2
07 stores the image 208 of the previous frame. The boundary information extraction unit 211 inputs the image 208 of the previous frame stored in the frame memory 207, detects the boundary of the signal level,
It is output as the boundary information 212. The boundary prediction unit 213
The boundary information 212 is motion-compensated using the motion vector 107 to obtain boundary prediction data 214. The coding area decoding unit 215 inputs the arbitrary shape designation information 115 and the prediction data 214 of the boundary, and inputs the start point of the arbitrary shape or the end point of the boundary and the information of the pixel width obtained from the prediction data 214 of the boundary. Is used to output the arbitrary shape information 216 describing the area in which the encoded data is to be arranged. The motion compensation unit 209 motion-compensates the image 208 of the previous frame stored in the frame memory 207 using the motion vector 107, and outputs an inter-frame predicted image 210.
The adder 205 adds the inter-frame predicted image 210 to the quantized inter-frame predicted error image 204, displays it as a decoded image 206, and stores it in the frame memory 207.

Next, the operation of this embodiment will be described.

The image 101 of the current frame and the image 103 of the previous frame stored in the frame memory 102 are input to the motion compensation section 106, and the motion vector 107 is detected.
The inter-frame predicted image 108 is generated and the motion vector 107 is output to the transmission path. The inter-frame predicted image 108 is used not only for prediction but also for generating a locally decoded image 125 necessary for prediction of the next frame.

The image 103 of the previous frame stored in the frame memory 102 is processed by the boundary information extraction unit 104.
By edge filter such as Sobel operator,
Signal level boundaries are detected. Obtained boundary information 10
5 is the motion vector 107 in the boundary prediction unit 109.
Is used to obtain the boundary prediction data 110.

The inter-frame prediction error 112 between the current frame image 101 and the inter-frame predicted image 108 is subtracted by the subtractor 111.
Is output from the output terminal and is input to the coding area determination unit 113 together with the boundary prediction data 110. Boundary prediction data 11
0 is used as data for determining an arbitrary shape to be encoded in the encoding area determination unit 113. If the areas with large prediction errors are grouped together and the area includes boundary prediction data, the boundary start point and pixel width are used.If the area does not start from the boundary end points, the area start and end points are used. The pixel width is output as the arbitrary shape designation information 115 and transmitted to the decoding side. Further, the coding region determination unit 113 also outputs a prediction error signal 114 having an arbitrary shape.

The arbitrary shape prediction error signal 114 and the arbitrary shape designation information 115 are input to the arbitrary shape DCT unit 116,
The DCT coefficient 117 is output. The DCT coefficient 117 is further quantized and encoded by the quantization encoding unit 118, and the encoded data 119 is transmitted.

On the other hand, the arbitrary shape designation information 115 is input to the coding area decoding unit 126 together with the boundary prediction data 110, and the arbitrary shape information 127 is reproduced from this. This arbitrary shape information 127 is used by the arbitrary shape IDCT unit 122.

The encoded data 119 is the inverse quantization decoding unit 12
At 0, the quantized value 121 is restored. Quantized value 12
In the arbitrary shape IDCT unit 122, the arbitrary shape IDCT 1 is applied to each shape based on the arbitrary shape information 127. As a result, the quantized inter-frame prediction error image 123 is obtained.

Quantized inter-frame prediction error image 12
3, the inter-frame predicted image 108 is added by the adder 124, and the locally decoded image 125 is added to the frame memory 102.
Is stored in

The encoded data 119 is transferred to the inverse quantization decoding unit 20.
At 1, the quantization value 202 is restored. Quantization value 20
2, the arbitrary shape IDCT unit 203 performs an arbitrary shape IDCT for each shape based on the arbitrary shape information 216. As a result, the quantized inter-frame prediction error image 204 is obtained.

The image 208 of the previous frame stored in the frame memory 207 is processed by the boundary information extraction unit 211.
By edge filter such as Sobel operator,
Signal level boundaries are detected. Obtained boundary information 21
2 is the motion vector 107 in the boundary prediction unit 213.
Is used to obtain the boundary prediction data 214.

The arbitrary shape designation information 115 and the boundary prediction data 214 are input to the coding area decoding unit 215. The coding area decoding unit 215 describes the area in which the coded data is to be arranged, using the starting point of the arbitrary shape or the boundary starting point obtained from the prediction data 214 of the boundary, the ending point of the arbitrary shape, and the pixel width information. The arbitrary shape information 216 is output.

The image 208 of the previous frame stored in the frame memory 207 is motion-compensated by the motion compensator 209 using the motion vector 107, and the inter-frame predicted image 210 is output.

Quantized inter-frame prediction error image 20
The inter-frame predicted image 210 is added to 4 by the adder 205 and displayed as the decoded image 206 and stored in the frame memory 207.

FIG. 2 shows a coding area determining unit 1 in FIG. 1 (A).
13 is a block diagram of FIG.

The threshold processing unit 301 compares the inter-frame prediction error 112 with a threshold and outputs the data 302 of a portion having a larger error than the threshold. Collating unit 303
Compares the data 302 having a large error with the boundary prediction data 110, checks whether the latter includes the former, and if the latter includes the boundary, the data 304 including the boundary is included. If not, the boundary is included. Data of the missing part 30
It is distributed as 5 and output. The processing unit 306 of the part including the boundary inputs the data 304 of the part including the boundary, and outputs the prediction error information 308 of the part including the boundary, the starting point, and the pixel width 311. The processing unit 307 that does not include the boundary
Inputs the data 305 of the portion not including the boundary, and outputs the prediction error information 309, the start point / end point and the pixel width 310 of the portion not including the boundary. The first integration unit 312 integrates the prediction error information 308 of the portion including the boundary and the prediction error information 309 of the portion not including the boundary, and outputs the prediction error signal 114 having an arbitrary shape. The second integration unit 313 integrates the start point and the pixel width 311, the start point / end point, and the pixel width 310,
It is output as the arbitrary shape designation information 115.

Next, the operation of the coding area determining unit 113 will be described.

First, the motion-compensated inter-frame prediction error 112 is compared with a threshold value in the threshold value processing unit 301, and the data 302 of a portion having a larger error than the threshold value is obtained.
Is output. The data 302 having a large error and the boundary prediction data 110 are collated by the collation unit 303,
It is checked if the latter contains the former. If it includes the boundary, the data 304 of the portion including the boundary, and if it does not include the boundary, the data 305 of the portion not including the boundary is distributed to the processing unit 306 including the boundary and the processing unit 307 including the boundary. . From the processing unit 306 of the part including the boundary, the prediction error information 30 of the part including the boundary is obtained.
8 and the starting point and the pixel width 311 are output. The prediction error information 309, the start point / end point, and the pixel width 310 of the portion that does not include the boundary are output from the processing unit 307 that does not include the boundary. The prediction error information 308 of the portion including the boundary and the prediction error information 309 of the portion not including the boundary are integrated by the first integration unit 312, and the prediction error signal 11 having an arbitrary shape is obtained.
It is output as 3. The start point / pixel width 311, the start point / end point, and the pixel width 310 are integrated by the second integration unit 313 and output as the arbitrary shape designation information 114.

Here, an example in which comparison by a threshold value is used to extract a portion having a large inter-frame prediction error has been described. However, after the actual conversion is performed and quantization is performed, the quantization level becomes 0. A method of extracting a portion having a non-quantized value can be used.

[0040]

As described above, according to the present invention, only a portion of the motion-compensated inter-frame prediction error signal that needs to be coded can be designated with a small amount of information, so that high coding efficiency can be achieved. There is an effect that can be achieved.

[Brief description of drawings]

FIG. 1 is an encoding device according to an embodiment of the present invention (FIG. 1 (A)).
2 is a block diagram of a decoding device (FIG. 1 (B)).

FIG. 2 is a block diagram of a coding area determination unit 113 in FIG.

FIG. 3 is a diagram illustrating an example of detecting a boundary line of an image.

4 is a diagram showing an example of a boundary line predicted based on the detected boundary line in FIG.

5 is a diagram showing an example in which a portion having a large prediction error in the motion compensation inter-frame prediction error signal of FIG. 4 is divided into arbitrary shapes.

FIG. 6 is a diagram showing an example in which a start point, an end point, and a pixel width are designated for the divided shape of FIG.

[Explanation of symbols]

 101 Input Pixel 102 Frame Memory 103 Image of Previous Frame 104 Boundary Information Extraction Unit 105 Boundary Information 106 Motion Compensation Unit 107 Motion Vector 108 Inter-frame Predicted Image 109 Boundary Prediction Unit 110 Boundary Prediction Data 111 Subtractor 112 Inter-frame Prediction Error 113 Code Coded region determination unit 114 Arbitrary shape prediction error signal 115 Arbitrary shape designation information 116 Arbitrary shape DCT unit 117 DCT coefficient 118 Quantization encoding unit 119 Encoded data 120 Inverse quantization decoding unit 121 Quantization value 122 Arbitrary shape IDCT unit 123 Quantized inter-frame prediction error image 124 Adder 125 Locally decoded image 126 Encoding area decoding unit 127 Arbitrary shape information 201 Inverse quantization decoding unit 202 Quantization value 203 Arbitrary shape IDCT unit 204 Quantized inter-frame prediction Measurement error image 205 Adder 206 Decoded image 207 Frame memory 208 Image of previous frame 209 Motion compensation unit 210 Inter-frame predicted image 211 Boundary information extraction unit 212 Boundary information 213 Boundary prediction unit 214 Boundary prediction data 215 Coding region decoding unit 216 Arbitrary shape information 301 Threshold processing unit 302 Data with large error 303 Collation unit 304 Data with portion including boundary 305 Data with portion without boundary 306 Processing portion with portion including boundary 307 Processing with portion not including boundary Part 308 Prediction error information of part including boundary 309 Prediction error information of part not including boundary 310 Start point / end point and pixel width 311 Start point and pixel width 312, 313 Integration section

Claims (4)

[Claims] 1. A moving picture coding / decoding method, wherein the coding side extracts a signal level or texture boundary using a decoded frame at a previous time and generates an inter-frame predicted image using motion compensation. , When detecting the position of the boundary that has moved in the predicted image and encoding the inter-frame prediction error with the current frame, divide the part with the large prediction error into an arbitrary shape, and start and end points and pixel width , Or the pixel width is specified from the position of the boundary detected as the start point, the pixels included in the pixel width are encoded, and the information necessary for motion compensation, the start point information, the pixel width information, and the encoded data are required. If there is, the end point information is transmitted to the decoding side, and the decoding side extracts the signal level or texture boundary from the decoded frame at the previous time using exactly the same method as the coding side, and decodes the information necessary for motion compensation. Then move When an inter-frame predicted image is generated using compensation, the position of a boundary that has moved in the predicted image is detected, and when decoding an inter-frame prediction error of an arbitrary shape, information on the start point and, if necessary, the end point A moving picture coding / decoding method characterized in that information is decoded and coded data is decoded based on the detected boundary position and decoded pixel width information. 2. A frame memory that stores an image of a previous frame, and a motion compensation that inputs a current frame image and an image of the previous frame stored in the frame memory, detects a motion vector, and generates an inter-frame prediction image. Unit, a boundary information extraction unit that detects a signal level or texture boundary from the image of the previous frame stored in the frame memory and outputs the boundary information as boundary information, and the boundary information is motion-compensated using the motion vector. A boundary prediction unit that outputs boundary prediction data, a subtracter that inputs the current frame image and the interframe prediction image, and outputs an interframe prediction error, the interframe prediction error, and the boundary prediction data , The large part of the inter-frame prediction error is collected, and if the area includes the prediction data of the boundary, And the pixel width, and if the area does not start from the end point of the boundary, the starting point and the end point of the area and the pixel width are output as arbitrary shape designation information, and a prediction error signal of an arbitrary shape is determined. A unit, an arbitrary shape analysis filter bank processing unit that inputs the arbitrary shape designation information and the prediction error signal of the arbitrary shape, and outputs an arbitrary shape analysis filter bank coefficient; A quantized coding unit that outputs the coded data, a coding region decoding unit that inputs the arbitrary shape designation information and the prediction data of the boundary, and reproduces the arbitrary shape information; and quantizes the coded data. An inverse quantization decoding unit for returning to a value, and the quantized inter-frame prediction error quantized by subjecting the quantized value to arbitrary shape synthesis filter bank processing for each shape based on the arbitrary shape information An encoding device that includes an arbitrary shape synthesis filter bank processing unit that outputs an image, and an adder that adds the inter-frame prediction image to the quantized inter-frame prediction error image and stores the image as a locally decoded image in the frame memory. 3. A threshold processing unit, wherein the coding area determination unit compares the inter-frame prediction error with a threshold value and outputs data of a portion having a large error equal to or larger than the threshold value; The data of a large part is compared with the predicted data of the boundary, and it is checked whether the latter includes the former. If the latter is included, it is classified as the data of the part including the boundary. And a collation unit that outputs the data including the boundary, and outputs prediction error information and a start point and a pixel width of the boundary, and a processing unit that includes the boundary and a part that does not include the boundary. Input data, and outputs the prediction error information of the part not including the boundary, the start point / end point, and the pixel width, the processing part of the part not including the boundary, the prediction error information of the part including the boundary, and the boundary. Integrating prediction error information portion without integrates the first integration unit which outputs as a prediction error signal of the arbitrary shape, the start point and the pixel width and the start / end and pixel width,
The encoding device according to claim 2, further comprising a second integration unit that outputs the arbitrary shape designation information. 4. An inverse-quantization decoding unit for returning the encoded data of claim 2 to a quantized value, and an arbitrary-shape synthesis filter bank process for each shape based on arbitrary shape information to the quantized value, Arbitrary shape synthesis filter bank processing unit that outputs a quantized inter-frame prediction error image, frame memory that stores the image of the previous frame, and input the image of the previous frame stored in the frame memory, and input the signal level or texture The boundary of
A boundary information extraction unit that outputs as boundary information, a boundary prediction unit that performs motion compensation on the boundary information using the motion vector of claim 2, and obtains prediction data of the boundary, and the arbitrary shape designation information of claim 2. By inputting the prediction data of the boundary, the start point of the arbitrary shape or the start point of the boundary and the end point of the arbitrary shape obtained from the prediction data of the boundary and the information of the pixel width are used to describe the area in which the encoded data is to be arranged. A coded area decoding unit that outputs the arbitrary shape information, and an image of the previous frame stored in the frame memory,
A motion compensation unit that performs motion compensation using the motion vector and outputs an inter-frame prediction image, and adds the inter-frame prediction image to the quantized inter-frame prediction error image, and outputs the decoded image as well as the frame. A decoding device having an adder for storing in a memory.