JPH0837664A - Moving picture encoding/decoding device - Google Patents

Abstract

PURPOSE:To improve encoding efficiency by using reconstitution picture signals which are the picture signals of a previous frame, deciding a feature point in a picture as a representative point and performing motion correction to the reconstitution picture signals by the motion vector of the representative point. CONSTITUTION:The reconstitution picture signals of the previous frame are stored in a frame memory 110 and the picture signals 111 of the previous frame read from the memory 110 are supplied to a representative point decision part 112 and a motion compensation circuit 114. The information 113 of the representative point is inputted to the circuit 114 and a deformation processing is performed to the signals 111 based on the decided representative point and a triangle by the circuit 114. The deformed picture signals are deformed so as to be most similar to the signals 101, the deformed signals are outputted as predictive picture signals 115 and they are outputted as motion vector information 116. The signals 115 are inputted to a subtractor 102, predictive error signals 103 for indicating the error of the signals 115 to the signals 101 are outputted and thus, encoding predictive signals 105 are outputted from a predictive error encoding part 104.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding apparatus for compressing and coding a moving picture signal and a moving picture decoding apparatus for decoding the coded signal.

[0002]

2. Description of the Related Art A moving picture coding technique for compressing and coding a moving picture signal with high efficiency is used in fields such as image communication, broadcasting and storage. For example, moving image encoding is performed for the purpose of transmitting at a narrow band or a low bit rate in moving image communication such as a video conference and a videophone, and at high speed in still image transmission such as a facsimile. Furthermore, when the image information is stored in a disk, a memory, or the like, this moving image coding enables efficient storage of a large amount of image information for a long time.

A motion compensation predictive coding system is a typical system of such a moving image coding technique. This is to reduce information by utilizing the high correlation between consecutive frames of a moving image signal. Specifically, for example,
The image signal input on a frame-by-frame basis is divided into multiple blocks, a motion vector indicating the motion of the image signal of the previous frame is obtained for each block, and motion compensation is performed on the image signal of the previous frame using this motion vector. This is a method of encoding the difference (prediction error signal) between the predicted image signal obtained by performing the operation and the image signal of the current frame. Then, the coded prediction error signal and the motion vector information are transmitted as a coded output.

As an improved motion compensation technique in this motion compensation predictive coding system, a polygon patch motion compensation system is known. This is done by applying a polygonal (triangular in this case) patch to the image signal of the current frame as shown in FIG. 6A, and then applying a patch whose vertices are moved as shown in FIG. 6B. The difference between the corresponding pixels of the image signal of the current frame and the image signal of the previous frame is encoded as a prediction error signal and transmitted.
At this time, the positions of the corresponding pixels are determined by linear interpolation according to the deformation of the corresponding patch. The positions of the vertices of the patch applied on the image signal of the previous frame are obtained by minimizing the information amount of the predicted image signal. In this case, the information necessary as the encoded output to be transmitted is the motion vector information of each vertex and the prediction error signal.

This polygon patch motion compensation is effective when the motion vector in the patch changes smoothly because the motion vector in the patch is calculated by linear interpolation. However, when the contour of the object in the image is inside the patch, the motion vector changes discontinuously at the contour, which causes a problem of image quality deterioration. Therefore, a method of applying a polygonal patch according to the feature points of the current frame (IEICE 1993 Spring National Convention D-295) has been proposed. However, in this method, although the motion vector itself is improved, the shape of the patch applied on the current frame depends on the image, and therefore it is necessary to transmit the information of the shape as well, which causes a problem of lowering the coding efficiency. .

In the conventional motion compensation predictive coding system, the image signal of the current frame is basically predicted only from the image signal of the previous frame. Therefore, for example, when an image that was visible in the past image is hidden and then reappears, or when it is projected outside the screen and returned to the screen again, the image is predicted from the image of the previous frame. However, the prediction error increases and the coding efficiency decreases.

To solve this problem, a plurality of frame memories are prepared, past image signals for a plurality of frames are stored, and a frame memory in which an image signal closest to the current frame is stored for each block is used. A method of selecting and using the image signal accumulated in the block as a reference image signal for motion compensation is also conceivable. However, this method requires a large number of frame memories, which not only increases the hardware scale but also requires additional transmission of additional information indicating which frame memory has been selected, so that the coding efficiency is not always required. Does not improve.

By the way, the information to be transmitted as the encoded output in the motion compensation predictive encoding method is basically the information of the prediction error signal and the motion vector as described above. To improve the coding efficiency, it is necessary to reduce the amount of these two types of information.

In order to reduce the information amount of the prediction error signal, in other words, to reduce the prediction error of the prediction image signal obtained by motion compensation, it is sufficient to increase the accuracy of motion vector detection. Specifically, for example, the detection accuracy of the motion vector with 1 pixel accuracy may be changed to 1/2 pixel accuracy. But,
When the detection accuracy of the motion vector is increased, the information amount of the prediction error signal is reduced, but the information of the motion vector is increased.

As a method of reducing the information amount of the motion vector, it is possible to quantize the motion vector and increase the quantization step size. However, when the quantization step size is increased, the prediction error of the prediction image signal obtained by motion compensation increases, and the information amount of the prediction error signal also increases.

On the other hand, as in motion compensation in the above-described motion compensation predictive coding system, morphing of video effect, image warping, or weaning, all or part of a reference image is transformed to generate a new image. The technology is also known. Among them, there is a method in which a polygonal mesh is stretched over the whole or part of the screen, and the mesh is deformed, and at the same time, the image inside the mesh is also deformed to generate a new image. In this case, the deformation of the image inside the area delimited by the mesh conforms to the deformation parameter of the polygon surrounding the area. However, when this method is used, discontinuity occurs between mesh regions when the direction of deformation and the amount of movement are significantly different between adjacent meshes. Therefore, there is a problem that when this method is used for motion compensation, a prediction error increases, and when it is used for a video effect, a visual discomfort occurs.

In the field of video effects, in order to eliminate such discontinuity between regions of a mesh, one or more representative straight lines are defined in the image instead of the mesh, and the straight lines are moved. There is a method called field morphing that creates a new image based on the relative positional relationship of all the pixels that make up the image from a straight line (for example, ThaddeusBe
“Feature-Based Image Metam” by ier, Shawn, Neely et al.
orphosis ”ComputerGraphics, Vol.26, No.2, pp.35-42,
See 1992). Using this method, a very smoothly deformed image can be obtained. However, with this method, for each straight line defined for all pixels that make up the screen,
Since it is necessary to calculate the relative positional relationship, there is a disadvantage that a very large amount of calculation is required. Also, in the case of complex deformation in terms of image effects, this method requires the definition of very fine meshes and the definition of many straight lines and their movements, which requires a huge amount of labor. is there.

[0013]

As described above, the conventional motion compensation predictive coding system has a problem that should be improved in terms of coding efficiency. In other words, the polygon patch motion compensation method has a problem that the motion vector changes discontinuously when the contour is included in the patch. The method of applying the rectangular patch has a problem that the coding efficiency is lowered because it is also necessary to transmit the shape information of the patch.

Further, in order to reduce a prediction error in the motion compensation predictive coding, past image signals for a plurality of frames are stored in a plurality of frame memories, and a frame memory in which an image signal close to the present frame is stored is used. In the method of selecting and obtaining the reference image signal for motion compensation, since a large number of frame memories are required, the hardware scale increases, and it is necessary to newly transmit the selection information of the frame memory as additional information. However, there is a problem that the coding efficiency is not necessarily improved.

Further, in order to reduce the amount of transmitted information and improve the coding efficiency, the method of increasing the motion vector detection accuracy to reduce the information amount of the prediction error signal increases the information of the motion vector. Therefore, the method of quantizing the motion vector and increasing the quantization step size to reduce the information amount of the motion vector has a problem that the prediction error of the prediction image signal increases and the information amount of the prediction error signal increases. there were.

On the other hand, as an image deformation technique used for motion compensation and image effects, a method of dividing an image into meshes and deforming the meshes causes discontinuity between meshes, which causes an increase in prediction error and image effects. However, the field morphing method to improve this problem requires a huge amount of calculation and labor.

An object of the present invention is to provide a moving picture coding / decoding device having high coding efficiency. Another object of the present invention is to provide an image transforming device that can obtain a smooth transformed image with a small amount of calculation and labor.

[0018]

[Means for Solving the Problems]

(1) A moving picture coding apparatus according to a first aspect of the present invention includes a prediction error generating unit that generates a prediction error signal indicating an error of a predicted image signal with respect to an input image signal, and a prediction prediction signal by coding the prediction error signal. Prediction error coding means for generating an error signal, image reconstruction means for generating a reconstructed image signal from the coded prediction error signal and the predicted image signal, and feature points in the image for the reconstructed image signal Representative point determining means for determining a representative point and a motion vector of the representative point are obtained, motion compensation is performed on the reconstructed image signal using this motion vector, and a predicted image signal for the next input image signal is generated. And a motion compensation means.

(2) A moving picture decoding apparatus according to a second aspect of the present invention receives at least a coded prediction error signal and motion vector information transmitted from the moving picture encoding apparatus according to the first aspect of the invention as a moving image. A moving picture decoding apparatus for decoding a signal, comprising a prediction error decoding means for decoding the coded prediction error signal to generate a prediction error signal, and a reconstructed image from the prediction error signal and the prediction image signal. Image reconstructing means for generating a signal, representative point determining means for deciding a characteristic point in the image as the representative point for the reconstructed image signal, and the reconstructed image signal using the representative point and the motion vector Motion compensation means for performing motion compensation and generating a predicted image signal for the next input image signal.

(3) A moving picture coding apparatus according to a third aspect of the present invention includes a prediction error generating means for generating a prediction error signal indicating an error of a predicted image signal with respect to an input image signal, and coding the prediction error signal. Prediction error coding means for generating a coded prediction error signal, image reconstructing means for generating a reconstructed image signal from the coded prediction error signal and the predicted image signal, and the reconstructed image signal in the image Representative point determining means for determining a characteristic point as a representative point, and a pattern analysis for analyzing a pattern around the representative point for the reconstructed image signal and obtaining motion vector resolution information indicating a resolution necessary for a motion vector based on the analysis. Means and a motion vector of the representative point at a resolution specified by the motion vector resolution information, and using this motion vector for the reconstructed image signal. Performed can compensate, characterized by comprising a motion compensation unit that generates a predicted image signal for the next input image signal.

(4) A moving picture coding apparatus according to a fourth aspect of the invention comprises a prediction error generating means for generating a prediction error signal indicating an error of the predicted image signal with respect to the input image signal, and coding the prediction error signal. Prediction error coding means for generating a coded prediction error signal, image reconstruction means for generating a reconstructed image signal from the coded prediction error signal and the predicted image signal, and motion compensation for the reconstructed image signal. First motion compensating means for generating a first motion-compensated image signal, characteristic image storage means for storing a characteristic image signal which is a signal of a characteristic region of the reconstructed image signal, and the characteristic image A second motion compensation means for performing motion compensation on the signal to generate a second motion compensated image signal, and an error of each small region between the first and second motion compensated image signals and the input image signal. Criticism Characterized by comprising a selection means for selecting whichever is smaller value as the prediction image signal for the next input image signal.

(4-1) In the moving picture coding apparatus according to the fourth aspect of the invention, the selecting means includes an error of each small area between the first and second motion compensation image signals and the input image signal. When all the evaluation values are equal to or greater than a predetermined value, no signal is selected as the predicted image signal, and the characteristic image storage means is a small area of the reconstructed image signal when the selection means selects the no signal. Is stored as the characteristic image signal.

(4-2) In the moving picture coding apparatus according to the fourth aspect of the invention, the characteristic image storage means stores the reconstructed image signal and the characteristic image signal read from the characteristic image signal storage means. A feature is that a reconstructed image signal when a state in which the evaluation value of the difference for each small region is larger than the set value continues for a predetermined period or longer is stored as a characteristic image signal.

(4-3) In the moving picture coding apparatus according to the fourth invention, the characteristic image storage means has a characteristic image having an image size larger than the image sizes of the input image signal, the motion vector and the reconstructed image signal. It is characterized by storing a signal.

(5) The moving picture decoding apparatus according to the fifth aspect of the present invention receives the moving picture encoding apparatus according to the fourth aspect of the invention as at least the encoded prediction error signal and the motion vector information. A moving picture decoding apparatus for decoding a signal, comprising a prediction error decoding means for decoding the coded prediction error signal to generate a prediction error signal, and a reconstructed image from the prediction error signal and the prediction image signal. An image reconstruction means for generating a signal, a first motion compensation means for performing motion compensation on the reconstructed image signal to generate a first motion compensated image signal, and a characteristic region of the reconstructed image signal Characteristic image storage means for storing a characteristic image signal which is a signal of
Second motion compensation means for performing motion compensation on the characteristic image signal to generate a second motion compensated image signal, and the first and second motion compensated image signals transmitted from the moving image encoding apparatus. Selection means for selecting the one designated by the selection information as the predicted image signal.

(5-1) In the moving picture decoding apparatus according to the fifth invention, the characteristic image storage means is an image signal of a small area of the reconstructed image signal when the selecting means selects no signal. Is stored as the characteristic image signal.

(5-2) In the moving picture decoding apparatus according to the fifth invention, the characteristic image storing means stores a characteristic image signal having an image size larger than the image size of the reconstructed image signal. And

(6) A moving picture coding apparatus according to a sixth aspect of the present invention includes image transforming means for transforming an input moving image signal according to an image transforming parameter to generate a transformed image signal, and the modified image signal. Prediction error generation means for generating a prediction error signal indicating an error of a prediction image signal, prediction error coding means for coding the prediction error signal to generate a coded prediction error signal, the coded prediction error signal, and An image reconstructing unit that generates a reconstructed image signal from the predicted image signal, and a motion analysis unit that analyzes the motion between the image signals from the input image signal and the reconstructed image signal to extract a motion parameter. Quantising the motion parameters,
Quantization means for generating a quantized motion parameter and the image transformation parameter, dequantization means for dequantizing the motion parameter quantized by the quantization means to obtain a motion vector, and the reconstructed image signal And a motion compensation unit that performs motion compensation using the motion vector to generate a predicted image signal for the next input image signal.

(7) A moving picture coding apparatus according to a seventh aspect of the present invention includes image transforming means for transforming an input moving image signal according to an image transforming parameter to generate a transformed image signal, and the modified image signal. Prediction error generation means for generating a prediction error signal indicating an error of a prediction image signal, prediction error coding means for coding the prediction error signal to generate a coded prediction error signal, the coded prediction error signal, and An image reconstructing means for generating a reconstructed image signal from a predicted image signal, a motion between the image signals is analyzed from the input image signal and the reconstructed image signal, and the image deformation parameter and motion compensation parameter Motion analyzing means for extracting the motion compensation parameter, quantizing means for quantizing the motion compensation parameter, and dequantizing the motion parameter quantized by the quantizing means to perform motion And inverse quantizing means for obtaining a vector,
And a motion compensating unit that performs motion compensation on the reconstructed image signal using the motion vector to generate the predicted image signal.

(8) A moving picture decoding apparatus according to an eighth aspect of the present invention includes at least an encoded prediction error signal transmitted from the moving picture encoding apparatus according to the sixth or seventh aspect and a quantized motion parameter. A prediction error decoding unit that decodes the coded prediction error signal to generate a prediction error signal, and the prediction error signal and the prediction. Image reconstructing means for generating a reconstructed image signal from an image signal, dequantizing means for dequantizing the quantized motion parameter to obtain a motion vector, and using the motion vector for the reconstructed image signal Motion compensation is performed to generate the predicted image signal.

(9) An image transforming device according to the ninth invention is
Based on the coordinate information of a representative point / line or area in the input image and the motion vector information indicating the movement of the point / line or area, coordinates other than the coordinate position specified by the coordinate information in the input image. Fluid analysis means for obtaining the unknown motion vector information of the position by fluid analysis and outputting the motion vector information of the entire area of the input image, and the pixels of the input image based on the motion vector information output from the fluid analysis means. And a mapping unit that obtains an output image by transforming the input image by mapping.

(10) An image transformation apparatus according to a tenth aspect of the present invention comprises a motion vector detecting means for detecting discrete motion vector information indicating a motion between an input image and a reference image, and a feature amount in the input image. A feature amount detecting unit for detecting and determining a coordinate position where the motion vector is not detected by the motion vector detecting unit, and a feature amount detecting unit for determining the coordinate position based on the motion vector information detected by the motion vector detecting unit. Fluid analysis means for outputting unknown motion vector information of the coordinate position by fluid analysis and outputting motion vector information indicating the movement of the entire area between the input image and the reference image, and the motion vector information output from this fluid analysis means. Mapping means for obtaining the output image by deforming the input image by mapping the pixels of the input image based on And wherein the Rukoto.

(11) An image transforming apparatus according to an eleventh aspect of the present invention comprises a motion vector detecting means for detecting discrete motion vector information indicating a motion between an input image and a reference image, and a feature amount in the input image. A feature amount detecting unit for detecting and determining a coordinate position where the motion vector is not detected by the motion vector detecting unit, and a feature amount detecting unit for determining the coordinate position based on the motion vector information detected by the motion vector detecting unit. Fluid analysis means for outputting unknown motion vector information of the coordinate position by fluid analysis and outputting motion vector information indicating the movement of the entire area between the input image and the reference image, and the motion vector information output from this fluid analysis means. Mapping means for obtaining a modified output image by mapping the pixels of the input image based on the mapping means; Error detecting means for detecting an error between the output image obtained by the ping means and the reference image, and motion vector information detected by the motion vector detecting means when the error detected by the error detecting means is equal to or more than a predetermined value. And a motion vector updating means for updating.

[0034]

In the moving picture coding apparatus according to the first aspect of the invention, the feature point in the image of the reconstructed image signal which is the image signal of the previous frame is determined as the representative point, and the motion vector of this representative point is used. Since motion compensation is performed on the reconstructed image signal to generate a predicted image signal, by performing motion compensation adapted to the characteristics of the image signal of the previous frame,
The information amount of the prediction error signal is reduced and the coding efficiency is improved. In this case, in the moving picture decoding apparatus, a representative point is determined in the same manner as the moving picture coding apparatus, and motion compensation is performed using the motion vector of the representative point to generate a predicted image signal, It is not necessary to transmit extra additional information such as patch shape information from the encoding device side as in the conventional case, and the encoding efficiency is not reduced.

In the moving picture decoding apparatus according to the second invention, the coding prediction error signal and the motion vector information transmitted from the moving picture coding apparatus according to the first invention are received and decoded, Reproduce the original video signal.

In the moving picture coding apparatus according to the third aspect of the present invention, the pattern around the representative point is analyzed for the reconstructed image signal, and the motion vector resolution information is obtained based on the analysis.
Since the motion vector of the representative point is obtained at the resolution specified by this motion vector resolution information, in the case of a complicated pattern, the motion vector is accurately detected with high resolution so that the prediction error signal becomes small, and the flat part of the image is detected. The resolution of the motion vector is reduced in the simple part of the pattern such that the accuracy is not higher than necessary, thus improving the coding efficiency by efficiently reducing the information amount of the motion vector and the motion vector search. Speed can be improved.

In the moving picture coding apparatus according to the fourth aspect of the invention, the first motion-compensated picture signal obtained by performing motion compensation on the reconstructed picture signal and the motion compensation on the characteristic picture signal are obtained. Of the second motion-compensated image signals, the one having the smaller evaluation value of the error for each small area from the input image signal is selectively used as the predicted image signal, so that the reconstructed image signal, that is, the image signal of the previous frame is always generated. The prediction error becomes smaller and the coding efficiency is improved as compared with the conventional method in which only the motion-compensated signal is used as the prediction image signal.

In the moving picture decoding apparatus according to the fifth invention, the coded prediction error signal, the motion vector information and the first motion compensated picture signal transmitted from the moving picture coding apparatus according to the fourth invention. The original moving image signal is reproduced by receiving the selection information indicating which of the second and the second motion-compensated image signal has been selected as the predicted image signal and performing decoding.

In the moving picture coding apparatus according to the sixth and seventh inventions, motion parameters (motion compensation parameters) are quantized in order to reduce the amount of transmission information, and dequantized motion parameters are motion-compensated. A motion compensation is performed on the reconstructed image signal as a motion vector for the motion vector to obtain a predicted image signal. In this case, due to the quantization error when quantizing the motion parameter, the motion vector for motion compensation deviates slightly from the optimum value in the sense that the prediction error is the smallest, and as a result, the information amount of the prediction error signal is reduced. It acts to increase.

Therefore, the input image signal is deformed by the image deformation parameter by the motion parameter corresponding to the quantization error, and the prediction error of the predicted image signal with respect to the obtained modified image signal is obtained. Since the prediction error obtained in this way has a small value equivalent to the prediction error before quantization of the motion parameter, the increase in the information amount of the prediction error signal can be suppressed low. Further, the reconstructed image signal in this case is an image signal obtained by modifying the input image signal by the quantization error, but the quantization error is less than 1/2 of the quantization step size, and Does not occur, the transmission information amount is reduced with almost no deterioration in image quality, and the coding efficiency is improved.

In the moving picture decoding apparatus according to the eighth invention, the coding prediction error signal and the motion vector information transmitted from the moving picture coding apparatus according to the sixth or seventh invention are received and decoded. To reproduce the original moving image signal.

In the image transformation apparatus according to the ninth to eleventh inventions, when obtaining an image obtained by transforming the input image,
The unknown motion vector information in the input image is obtained by fluid analysis from the motion vector information of a representative point / line or area in the input image or the discrete motion vector information between the input image and the reference image. By generating the motion vector information of the entire area and mapping the pixels of the input image based on this motion vector information, an output image obtained by transforming the input image with a small amount of calculation can be obtained.

[0043]

Embodiments of the present invention will be described below with reference to the drawings. (Embodiment 1) FIG. 1 is a block diagram showing an embodiment of a moving picture coding apparatus according to the first invention. In the figure,
The input image signal 101 is, for example, a moving image signal in units of frames, and is input to the subtractor 102 and the motion compensation circuit 114.

On the other hand, the frame memory 110 stores the reconstructed image signal of the previous frame. The image signal 11 of the previous frame read from the frame memory 110
1 is input to the representative point determination unit 112 and the motion compensation circuit 114. The representative point determination unit 112 uses the frame memory 110.
Motion compensation circuit 1 based on the image signal 111 read from
At 14, a representative point for motion compensation is determined.
As the representative points, corners of the object in the image and other characteristic points are selected. At this time, not only the representative points but also lines connecting the representative points (triangle sides) may be determined depending on the motion compensation method. FIG. 2 shows an example of determining the representative points and the triangles when the motion compensation circuit 114 uses the triangle patch compensation method. The circles represent the representative points and the broken lines represent the sides of the triangles.

The image signal 111 read from the frame memory 110 and the representative point information 113 determined by the representative point determination unit 112 are input to the motion compensation circuit 114. The motion compensation circuit 114 transforms the image signal 111 based on the representative points and the triangles determined by the representative point determination unit 112, and transforms the modified image signal so that it is most similar to the input image signal 101. Then, the transformed image signal is output as the predicted image signal 115, and the motion vector information 116 of the representative point at that time is output. The representative point motion vector information 116 is information indicating the motion vector of the representative point portion between the image signal 111 before deformation and the image signal (prediction image signal) 115 after deformation.

The prediction error signal 115 is input to the subtractor 102. The subtractor 102 is a prediction error signal 103 indicating an error of the prediction image signal 115 with respect to the input image signal 101.
Is output. The prediction error signal 103 is coded by the prediction error coding unit 104. Prediction error coding unit 104
Is, for example, a DCT (discrete cosine transform) circuit and this D
It is composed of a quantization circuit that quantizes the DCT coefficient obtained by the CT circuit, and outputs a coded prediction error signal 105.

Any coding method can be used in the prediction error coding unit 104 as long as it can efficiently represent the waveform. Instead of the DCT, a vector quantum that collectively represents several pixels with one code is used. It is also possible to use encoding, sub-band encoding, a conditional pixel replenishment method for sending a non-zero pixel value with an address, and a method for encoding the difference between adjacent pixels.

The coded prediction error signal 105 is input to the prediction error decoding unit 106, and the prediction error signal is decoded.
The prediction error decoding unit 106 includes, for example, an inverse quantization circuit that inversely quantizes the coded prediction error signal 105, and an inverse DCT circuit that performs an inverse process on the dequantized signal as compared with the DCT circuit. , And outputs the decoded prediction error signal 107. The decoded prediction error signal 107 is added to the prediction image signal 115 by the adder 108 to become a reconstructed image signal 109. The reconstructed image signal 109 is stored in the frame memory 110 for one frame in preparation for encoding the input image signal of the next frame.

The coded prediction error signal 105 output from the prediction error coding unit 104 and the motion vector information 116 output from the motion compensation circuit 114 are variable length coded by the variable length coding units 117 and 118, respectively. Further, after being multiplexed by the multiplexing unit 119, the encoded output 120
Is transmitted to the transmission line or the storage system.

Hereinafter, the same operation is repeated every time an image signal of a new frame is input as the input image signal 101. As described above, according to the present embodiment, since the motion compensation that matches the characteristics of the image signal 111 of the previous frame is performed to generate the predicted image signal 115, the prediction error signal 103 becomes smaller and the encoded prediction error signal 105 becomes smaller. The amount of information is reduced. Therefore, there is an advantage that the coding efficiency is improved in combination with the fact that additional information such as the conventional patch shape information need not be newly transmitted.

(Embodiment 2) FIG. 3 is a block diagram showing an embodiment of a moving picture decoding apparatus according to the second invention. In the figure, an input signal 201 is a signal transmitted from the moving picture coding apparatus of FIG. 1 via a transmission system or a storage system, and includes a coding prediction error signal and motion vector information. This input signal 201 is input to the separation unit 202,
The encoded prediction error signal 203 and the side information 204 including the motion vector information are separated and output. The encoded prediction error signal 203 and the side information 204 are variable-length decoded by the variable-length decoding units 205 and 206, respectively.

The coded prediction error signal 207 output from the variable length decoding unit 205 is decoded by the prediction error decoding unit 209. The prediction error decoding unit 209 performs a process reverse to that of the prediction error encoding unit 104 in FIG. 1, and, for example, an inverse quantization circuit that inversely quantizes the encoded prediction error signal 207 and an inverse quantization unit. It is composed of an inverse DCT circuit that performs a process opposite to that of the DCT circuit on the signal, and outputs a prediction error signal 210. The prediction error signal 210 is added to the adder 21.
Input to 1.

On the other hand, the frame memory 213 stores the reconstructed image signal of the previous frame. The image signal 21 of the previous frame read from the frame memory 213
4 is input to the representative point determination unit 215 and the motion compensation circuit 217. The representative point determination unit 215 uses the frame memory 110.
Motion compensation circuit 2 based on the image signal 214 read from
At 17, a representative point for motion compensation is determined.
As the representative points, corners of the object in the image and other characteristic points are selected. That is, the representative point determining unit 215 obtains the same representative point as the representative point obtained by the representative point determining unit 112, by the same operation as that of the representative point determining unit 112 in the moving picture coding apparatus of FIG. At this time, not only the representative points but also the line connecting the representative points (the side of the triangle) may be determined depending on the motion compensation method, as in the representative point determining unit 112.

The image signal 214 read from the frame memory 213 and the representative point information 216 determined by the representative point determining unit 215 are input to the motion compensation circuit 217. The motion compensation circuit 217 also receives the motion vector information 208 decoded from the side information 204 by the variable length decoding unit 206, and the motion vector information based on the representative point and the triangle determined by the representative point determining unit 215. The image signal 214 is modified using 208, and the modified image signal is output as the predicted image signal 218.

The predicted image signal 218 is input to the adder 211. The adder 211 adds the prediction error signal 210 and the prediction image signal 218 and outputs the reconstructed image signal 212. The reconstructed image signal 212 is sent as an output image signal to a circuit in the next stage (not shown) and is stored in the frame memory 213 in preparation for decoding of the next frame.

As described above, according to this embodiment, the original moving picture signal can be decoded from the coded prediction error signal and the motion vector information transmitted from the moving picture coding apparatus of FIG. .

(Embodiment 3) FIG. 4 is a block diagram showing an embodiment of a moving picture coding apparatus according to the third invention. The parts corresponding to those in FIG. 1 will be assigned the same reference numerals, and only the differences from the first embodiment will be described. In this embodiment, a pattern analysis unit 121 is added to the configuration of FIG. This pattern analysis unit 1
The image signal 111 of the previous frame read from the frame memory 110 and the representative point information 113 from the representative point determination unit 112 are input to 21.

The pattern analysis unit 121 includes a representative point determination unit 112.
The surrounding pattern is analyzed for each of the representative points determined in step 1, and the motion vector resolution information 122 indicating the resolution required for the motion vector is output to the motion compensation circuit 114 based on the analysis. As a specific method of realizing the pattern analysis unit 121, for example, DCT is performed on the image around the representative point,
There is a method of classifying a picture into several patterns by the obtained DCT coefficient and determining the resolution of the motion vector according to the pattern.

The motion compensation circuit 114 searches for and obtains the motion vector of the representative point at the resolution specified by the motion vector resolution information 122, and uses this motion vector to motion-compensate the image signal 111 from the frame memory 110. By performing the above, the predicted image signal 115 is generated.

FIG. 5 shows the motion compensation circuit 1 according to this embodiment.
14 shows a search point of the motion vector at 14. These motion vector search points are determined according to the motion vector resolution information 122 from the pattern analysis unit 121. Now, assuming that the motion vector search points in the case of a flat pattern are shown in FIG. 5A, the motion vector search points in the case of a complicated pattern are shown in FIG.
The interval is smaller than that in the case of FIG. 5A as shown in FIG. On the other hand, in the case of a pattern in which the frequency component in the vertical direction is lower than the frequency component in the horizontal direction, the motion vector search points have large intervals in the vertical direction and small intervals in the horizontal direction, as shown in FIG.

As described above, in this embodiment, the resolution of the motion vector, that is, the search accuracy of the motion vector is determined according to the fineness of the pattern of the image signal 111 of the previous frame, and the motion compensation is performed using the motion vector. Predicted image signal 11
5 is generated, the resolution of the motion vector is increased so that the prediction error signal 103 becomes smaller in the case of a complicated pattern, and the resolution of the motion vector is decreased in the case of a simple pattern such as a flat portion of the image. As a result, it is possible to reduce the amount of generated information and speed up the motion vector search.

(Embodiment 4) FIG. 7 is a block diagram showing an embodiment of a moving picture coding apparatus according to the fourth invention. In the figure, the input image signal 101 is, for example, a moving image signal in units of frames, and is input to the subtractor 102 and the first and second motion compensation circuits 114 and 133.

On the other hand, the frame memory 110 stores the reconstructed image signal of the previous frame, and the characteristic image memory 131 stores the characteristic image signal of the previous frame. The image signal 111 of the previous frame read from the frame memory 110 is input to the first motion compensation circuit 114, and the characteristic image signal 132 read from the characteristic image memory 131 in synchronization with this is the second image signal 132. Motion compensation circuit 1
33 is input. The first motion compensation circuit 114 performs motion compensation on the image signal 111 from the frame memory 110 so as to be closest to the input image signal 101 of the current frame, and outputs the first motion-compensated image signal 115. Output.

Here, as the motion compensation method, the screen is divided into a plurality of blocks, the motion vector is obtained for each block, and the image signal 111 of the previous frame is shifted to obtain the input image signal 101 of the current frame. The matching block matching method may be used, or any of the methods using polygonal patches as described in the first embodiment may be used. Similarly to the first motion compensation circuit 114, the second motion compensation circuit 133 also performs motion compensation so that the characteristic image signal 132 from the characteristic image memory 131 is closest to the input image signal 101. The second motion compensation image signal 134 is output.

First and second motion-compensated image signals 11
5,134 are input to the error evaluation unit 136 and the predicted image signal selection unit 139. The error evaluator 136 calculates the motion-compensated image signals 115 and 134 and the input image signal 10 of the current frame.
The error with respect to 1 is evaluated, and the evaluation result is sent to the control unit 137. The control unit 137 supplies the selection control signal 138 to the predicted image signal selection unit 139 based on the evaluation result from the error evaluation unit 136.

That is, the control unit 137 controls the input image signal 10 of the current frame among the motion compensation image signals 115 and 134.
The signal with the smaller error from 1 is the predicted image signal selection unit 13
9 as the predicted image signal 140, and the motion-compensated image signals 115 and 134 and the input image signal 10 of the current frame.
When the evaluation results of the error with respect to 1 are both large and are equal to or more than a certain threshold value, the selection control signal 138 is supplied to the prediction image signal selection unit 139 so as to select the no-signal “0” as the prediction image signal 140. To do.

The predicted image signal 115 is input to the subtractor 102. The subtractor 102 is a prediction error signal 103 indicating an error of the prediction image signal 115 with respect to the input image signal 101.
Is output. The prediction error signal 103 is coded by the prediction error coding unit 104. Prediction error coding unit 104
Is, for example, a DCT (discrete cosine transform) circuit and this D
It is composed of a quantization circuit that quantizes the DCT coefficient obtained by the CT circuit, and outputs a coded prediction error signal 105.

Any coding method may be used as the coding method in the prediction error coding unit 104 as long as the waveform can be efficiently expressed. Instead of the DCT, a vector quantum expressing several pixels collectively by one code is used. It is also possible to use encoding, sub-band encoding, a conditional pixel replenishment method for sending a non-zero pixel value with an address, and a method for encoding the difference between adjacent pixels.

The coded prediction error signal 105 is input to the prediction error decoding unit 106, and the prediction error signal is decoded.
The prediction error decoding unit 106 includes, for example, an inverse quantization circuit that inversely quantizes the coded prediction error signal 105, and an inverse DCT circuit that performs an inverse process on the dequantized signal as compared with the DCT circuit. , And outputs the decoded prediction error signal 107. The decoded predicted image signal 107 is added to the predicted image signal 115 by the adder 108 to become the reconstructed image signal 109. The reconstructed image signal 109 is stored in the frame memory 110 for one frame in preparation for encoding the input image signal of the next frame.

The reconstructed image signal 109 from the adder 108 is also input to the characteristic image memory 131, and the characteristic image signal which is the signal of the characteristic region is stored therein. A specific method of realizing the characteristic image memory 131 will be described later.

The coded prediction error signal 105 output from the prediction error coding unit 104 is variable length coded by the variable length coding unit 117. Also, the motion vector information 116 and 135 used for motion compensation in the first and second motion compensation circuits 114 and 133, and the selection control signal 1 supplied from the control unit 137 to the predicted image signal selection unit 139.
38 is variable-length coded by the variable-length coding unit 118. The variable-length codes output from the variable-length coding circuits 117 and 118 are multiplexed by the multiplexing unit 119 and then sent to the transmission line or the storage system as the coded output 120.

Thereafter, the same operation is repeated every time an image signal of a new frame is input as the input image signal 101. Next, a specific configuration example of the characteristic image memory 131 in FIG. 7 will be described with reference to FIGS. 8 and 9.

The characteristic image memory 131 shown in FIG. 8 is input from an image signal when no signal "0" is selected in the predicted image signal selecting section 139, that is, the first and second motion compensation image signals 115 and 134. The image signal of a portion where the image signal 101 cannot be predicted is stored as a characteristic image signal, and is configured by the memory unit 141 and the write control unit 142. The memory unit 141 has an adder 108 shown in FIG.
The reconstructed image signal 109 from is input. The write control unit 142 receives the selection control signal 138 from the control unit 137.
, And the prediction image signal selection unit 139 outputs no signal “0” by the selection control signal 138.
During the period selected as, the write signal is supplied to the memory unit 141. As a result, the reconstructed image signal 109 during this period is written in the memory unit 141 as a characteristic image signal.

The characteristic image memory 131 shown in FIG. 9 checks the difference between the reconstructed image signal 109 and the characteristic image signal 132 read from the characteristic image memory 131 for each small area, and the difference is larger than the set value. The reconstructed image signal 109 is stored as a characteristic image signal for a predetermined period or more, and is composed of a memory unit 151, a write control unit 152, a subtractor 153, and a time measuring unit 154.

That is, the difference between the reconstructed image signal 109 and the characteristic image signal 132 read from the memory unit 151 is obtained by the subtractor 153, and this difference is evaluated by the evaluation unit 154 in a small area (block or a size different from the block). Area) is evaluated. Then, the time measuring unit 154 measures the time length of the period in which the error evaluation value is larger than the set value, and when it exceeds the predetermined period, the write control unit 1
A write signal is supplied from 52 to the memory unit 151. As a result, the reconstructed image signal 109 during this period is written in the memory unit 151 as a characteristic image signal.

The size of the characteristic image memory 131 (image size of the characteristic image signal) may be the same as the size of the frame memory 110 (image size of the reconstructed image signal), but as shown in FIG. May be larger than the size. In the figure, the broken line indicates the size of the frame memory 110, and the solid line indicates the characteristic image memory 1
31 shows the size. Then, if the entire reconstructed image is moving in parallel, that is, moving in the same direction,
Similarly, the characteristic image is also moved in parallel, and the portion protruding from the screen is also saved as the characteristic image.

By doing so, when the portion protruding from the screen returns to the inside of the screen again, the characteristic image memory 13
The characteristic image signal 132 from the first motion compensation circuit 13
The second motion-compensated image signal 134 that has been motion-compensated in 3 can be used as the predicted image signal 140. For example, when the user holds the camera by hand and shoots, the part outside the screen often returns to the inside of the screen again. According to the present embodiment, the prediction error is reduced even in such a case. It is possible to improve the coding efficiency.

As described above, according to this embodiment, the first motion-compensated image signal obtained by performing the motion compensation on the reconstructed image signal 111 read from the frame memory 110 by the first motion compensation circuit 114. 115 and the input image signal 101 of the second motion-compensated image signal 134 obtained by performing motion compensation on the characteristic image signal 132 read from the characteristic image memory 131 by the second motion compensation circuit 133. The predicted image signal selection unit 139 selects one having a smaller evaluation value of the error for each small region and uses it as the predicted image signal 140, so that the image signal 11 of the previous frame is always generated.
The prediction error becomes smaller and the coding efficiency is improved as compared with the conventional method in which a signal in which only 1 is simply motion-compensated is used as the prediction image signal.

(Embodiment 5) FIG. 11 is a block diagram showing an embodiment of a moving picture decoding apparatus according to the fifth invention.
In the figure, an input signal 201 is a signal transmitted from the moving picture coding apparatus of FIG. 7 via a transmission system or a storage system, and includes a coding prediction error signal and side information such as motion vector information and a selection control signal. Is included. The input signal 201 is input to the separation unit 202, and the encoded prediction error signal 203 and the side information 204 including the motion vector information and the selection control signal are separated and output. The encoded prediction error signal 203 and the side information 204 are variable-length decoded by the variable-length decoding units 205 and 206, respectively.

The prediction error signal 207 output from the variable length decoding unit 205 is decoded by the prediction error decoding unit 209. The prediction error decoding unit 209 performs a process reverse to that of the prediction error encoding unit 104 in FIG. 1, and for example, an inverse quantization circuit that inversely quantizes the prediction error signal 207 and an inversely quantized signal. On the other hand, the decoded prediction error signal 210 is composed of an inverse DCT circuit that performs processing opposite to that of the DCT circuit.
Is output. The decoded prediction error signal 210 is added to the adder 21.
Input to 1.

On the other hand, the frame memory 213 stores the reconstructed image signal of the previous frame, and the characteristic image memory 231 stores the characteristic image signal of the previous frame. The image signal 214 of the previous frame read from the frame memory 213 is input to the first motion compensation circuit 217, and in synchronization with this, the characteristic image signal 232 read from the characteristic image memory 231 is the second image signal 232. It is input to the motion compensation circuit 233. The first motion compensation circuit 214 performs motion compensation on the image signal 214 from the frame memory 213 using the first motion vector information 208 decoded from the side information 204 by the variable length decoding circuit 206, The first motion compensation image signal 218 is output.

Similarly to the first motion compensation circuit 217, the second motion compensation circuit 233 also decodes the characteristic image signal 232 from the characteristic image memory 231 from the side information 204 by the variable length decoding circuit 206. Motion compensation is performed using the second motion vector information 222, and a second motion-compensated image signal 234 is output.

First and second motion-compensated image signals 21
8, 234 are input to the predicted image signal selection unit 139. The predicted image signal selection unit 139 includes the variable length decoding circuit 2
A selection control signal 221 decoded from the side information 204 according to 06 selects either the motion-compensated image signals 218, 234 or no signal “0” as the predicted image signal 236, and supplies it to the adder 211.

The adder 211 outputs the decoded prediction error signal 210.
And the predicted image signal 236 from the predicted image signal selection unit 235
And are added and the reconstructed image signal 212 is output. This reconstructed image signal 212 is sent as an output image signal to a circuit in the next stage (not shown), and is stored in the frame memory 213 and the characteristic image memory 231 in preparation for the decoding of the next frame.

As described above, according to the present embodiment, the original moving picture signal is decoded from the coded prediction error signal, the motion vector information and the selection control signal transmitted from the moving picture coding apparatus of FIG. can do.

(Embodiment 6) FIG. 12 is a block diagram showing an embodiment of a moving picture coding apparatus according to the sixth invention.
In the figure, the input image signal 101 is, for example, a moving image signal in units of frames, and is input to the image transformation unit 161 and the motion analysis unit 164.

On the other hand, the frame memory 110 stores the reconstructed image signal of the previous frame. Reconstructed image signal 1 of the previous frame read from the frame memory 110
11 is input to the motion compensation circuit 114. The motion compensation circuit 114 performs motion compensation on the reconstructed image signal 111 from the frame memory 110 using motion vector information 170 generated as described below, and the predicted image signal 11
5 is output.

The image signal 111 of the previous frame read from the frame memory 110 is also input to the motion analysis unit 164. The motion analysis unit 164 analyzes the motion between the input image signal 101 and the reconstructed image signal 111 to extract a motion parameter 165. As the motion parameter 165, for example, a parameter expressing the motion of each pixel, each block, each motion region or the motion of the vertex of a polygon patch is used. The extraction method of the motion parameter 165 is, for example, a gradient method for each pixel, a block matching method for each block,
Affine parameter estimation (eg,
Yamada et al., "Motion Extraction of Color Video by Affine Matching", Technical Report of the Television Society of Japan, October 1992, etc.), and hexagonal matching method (Nakaya, Naemura et al.) 16 kb / s encoding of moving image ", PCSJ92) and the like can be used.

The motion parameter 165 extracted by the motion analysis unit 164 is quantized by the quantization unit 166. The quantizer 166 uses the quantized value 167 and the quantized error information 168.
Is output. Quantized value 1 out of the output of the quantizer 166
67 is input to the dequantization unit 169, dequantized, and supplied to the motion compensation circuit 114 as motion vector information 170 for motion compensation.

On the other hand, of the output of the quantizing unit 166, the quantizing error information 168 is sent to the image transforming unit 161 as an image transforming parameter 162 having a value corresponding to the quantizing error of the motion parameter 165 in the quantizing unit 166. Supplied. The image transformation unit 161 performs motion compensation (deformation) on the input image signal 101 using the image transformation parameter 162 by the same operation as that of the motion compensation circuit 114.
63 is output.

Here, a specific example of the case where the block matching method is used in the image transformation unit 161 will be described. Now, it is assumed that the motion vector for a certain block extracted by the motion analysis unit 164 is (2.3, 0), and the quantization step size in the quantization unit 166 is 1. At this time, since the quantization error in the quantization unit 166 is (0.3, 0), the transformed image signal 163 in the block is the transformed image signal 163.
1, the input image signal 101 can be obtained by motion compensation (deformation) using the motion parameter of (-0.3, 0).

By the way, when the image transforming unit 161 transforms an image using the block matching method as described above, block distortion occurs due to lack of matching between adjacent blocks, and the reconstructed image signal 109 is visually recognized. There is a possibility that the image quality may deteriorate significantly. A method using a polygon patch is effective as a transformation method that causes less visual deterioration in image quality. In the case of a polygon patch, the motion parameter represents the motion of the vertices of the polygon. As a specific example, the case where the motion parameter of a certain vertex is (2.3, 0) and the quantization step size is 1 is shown. At this time, since the quantization error is (0.3, 0), the motion parameter of this vertex is (-0.3, 0), and this is used as the image transformation parameter 16
By performing motion compensation (deformation) on the input image signal 101 as 2, the deformed image signal 163 can be obtained.・
The transformed image signal 163 obtained by the image transformation unit 161 is input to the subtractor 102. The subtractor 102 outputs a prediction error signal 103 indicating an error of the prediction image signal 115 with respect to the modified image signal 163. The prediction error signal 103 is coded by the prediction error coding unit 104. The prediction error encoding unit 104 is, for example, DCT (discrete cosine transform).
Circuit and a quantization circuit for quantizing the DCT coefficient obtained by this DCT circuit, and the coded prediction error signal 1
05 is output.

Any coding method in the prediction error coding unit 104 may be used as long as it is a method capable of efficiently expressing a waveform. Instead of the DCT, a vector quantum expressing several pixels collectively by one code is used. It is also possible to use encoding, sub-band encoding, a conditional pixel replenishment method for sending a non-zero pixel value with an address, and a method for encoding the difference between adjacent pixels.

The coded prediction error signal 105 is input to the prediction error decoding unit 106, and the prediction error signal is decoded.
The prediction error decoding unit 106 includes, for example, an inverse quantization circuit that inversely quantizes the coded prediction error signal 105, and an inverse DCT circuit that performs an inverse process on the dequantized signal as compared with the DCT circuit. , And outputs the decoded prediction error signal 107. The decoded predicted image signal 107 is added to the predicted image signal 115 by the adder 108 to become the reconstructed image signal 109. The reconstructed image signal 109 is stored in the frame memory 110 for one frame in preparation for encoding the input image signal of the next frame.

The coded prediction error signal 105 output from the prediction error coding unit 104 is variable length coded by the variable length coding unit 117. The motion parameter (quantized value 167) quantized by the quantization unit 167 is variable length coded by the variable length coding unit 118. The variable-length codes output from the variable-length coding circuits 117 and 118 are multiplexed by the multiplexing unit 119 and then sent to the transmission line or the storage system as the coded output 120.

Hereinafter, the same operation is repeated every time an image signal of a new frame is input as the input image signal 101. As described above, according to the present embodiment, the motion parameter 165 output from the motion analysis unit 164 is quantized by the quantization unit 166 and the inverse quantization unit 169 dequantizes the motion parameter 165 in order to reduce the amount of transmission information. A motion vector 170 for motion compensation is generated, and a motion compensation circuit 114 performs motion compensation on the reconstructed image signal 111 to obtain a predicted image signal 115.

In this case, due to the quantization error in quantizing the motion parameter 165, the motion vector for motion compensation is slightly deviated from the optimum value in the sense that the prediction error is the smallest, and as a result, the coding prediction is performed. Error signal 105
It acts to increase the amount of information. However, in this embodiment, the image transforming unit 161 uses the image transforming parameter 162 having a value corresponding to the quantization error to input the input image signal 1
01 is deformed, and the prediction error signal 103 indicating the error of the prediction image signal 115 with respect to the modified image signal 163 obtained thereby is generated, so that the prediction error becomes the prediction error before the quantization of the motion parameter. Since it can be set to an equivalent small value, the encoded prediction error signal 105
It is possible to suppress the increase in the information amount of Further, the reconstructed image signal 109 in this case is an image signal obtained by modifying the input image signal 101 by the quantization error, but the quantization error is 1/2 or less of the quantization step size. Since the accumulation of encoding errors does not occur, the amount of transmission information is reduced with almost no deterioration in image quality, and the encoding efficiency can be effectively improved.

(Embodiment 7) FIG. 13 is a block diagram showing an embodiment of a moving picture coding apparatus according to the seventh invention.
The same reference numerals are given to the portions corresponding to those in FIG. 12, and the difference from the seventh embodiment will be described. In this embodiment, the motion analysis unit 172.
The second embodiment is different from the seventh embodiment in that two parameters, that is, a motion compensation parameter 165 and an image deformation parameter 162 are extracted.

These two parameters are extracted as follows. That is, first, the predicted image signal obtained by motion-compensating the image signal of the previous frame using the first motion parameter with the quantization step size as the minimum unit, and the size equal to or smaller than 1/2 of the quantization step size. A prediction error, which is an error between the input image signal and the deformed image signal obtained by deforming the input image signal with the second motion parameter having a value as a value, is obtained. Then, of the various combinations of the values of the first and second motion parameters, the combination having the smallest prediction error is extracted, and the set of these parameter values is set as the motion compensation parameter 165 and the image deformation parameter 162, respectively. It is supplied to the quantization unit 166 and the image transformation unit 161.

The motion compensation parameter 165 is used in the seventh embodiment.
Similarly, the parameter 167 quantized by the quantizer 166 and quantized by the inverse quantizer 169 is inversely quantized and supplied to the motion compensation circuit 114 as motion vector information 170 for motion compensation.

Now, for example, the motion compensation circuit 114 performs motion compensation using a polygonal patch with a motion vector of one vertex as one pixel precision (quantization step size 1), and the input image signal 101 in the image transformation unit 161. Is assumed to be performed with a precision of ¼ pixel. In this case, the motion vector candidate values of the motion compensation parameter 165 are (0, 0), (1, 1),
(1, 0), (1, -1), ..., On the other hand, the motion vector candidate values of the image transformation parameter 162 are (0, 0), (0.2
5, 0.25), (0.25, 0), (0.25, -0.25), and so on. A combination that minimizes the prediction error between the modified image signal 163 and the predicted image signal 115 is searched for from the combination of both. Any search method may be used, a full search may be performed, or the first image transformation parameter 1
The motion vector candidate value of 62 may be fixed to (0, 0), the motion vector candidate value of the motion compensation parameter 165 may be roughly searched to limit the search range, and then the full search may be performed. Assuming that the combination of (3, -2) and (0.25, 0) minimizes the prediction error, (3, -2)
Motion compensation is performed with and the image is transformed with (0.25, 0).

(Embodiment 8) FIG. 14 is a block diagram showing an embodiment of a moving picture decoding apparatus according to the eighth invention.
In the figure, an input signal 201 is a signal transmitted from the moving picture coding apparatus of FIG. 12 or FIG. 13 via a transmission system or a storage system, and includes a side information such as a coding prediction error signal and motion parameter information. Contains. The input signal 201 is input to the separation unit 202, and the encoded prediction error signal 203 and the side information 20 including the motion parameter information.
4 is separated and output. Encoding prediction error signal 203
And the side information 204 are respectively received by the variable length decoding unit 20.
5, 206, variable length decoding is performed.

The prediction error signal 207 output from the variable length decoding unit 205 is decoded by the prediction error decoding unit 209. The prediction error decoding unit 209 performs a process reverse to that of the prediction error encoding unit 104 in FIG. 1, and for example, an inverse quantization circuit that inversely quantizes the prediction error signal 207 and an inversely quantized signal. On the other hand, the decoded prediction error signal 210 is composed of an inverse DCT circuit that performs processing opposite to that of the DCT circuit.
Is output. The decoded prediction error signal 210 is added to the adder 21.
Input to 1.

On the other hand, the reconstructed image signal of the previous frame is stored in the frame memory 213. The image signal 21 of the previous frame read from the frame memory 213
4 is input to the motion compensation circuit 217. Further, the variable-length decoding circuit 206 decodes the motion parameter 241 from the side information 204, and further, the dequantization unit 242 dequantizes the motion parameter 241 to obtain motion vector information 243 for motion compensation. Motion compensation circuit 217
Performs motion compensation on the image signal 214 from the frame memory 213 using this motion vector information 243, and outputs a predicted image signal 218.

The predicted image signal 218 is input to the adder 211. The adder 211 adds the prediction error signal 210 and the prediction image signal 218 and outputs the reconstructed image signal 212. This reconstructed image signal 212 is sent as an output image signal to a circuit in the next stage (not shown), and is stored in the frame memory 213 and the characteristic image memory 231 in preparation for the decoding of the next frame.

As described above, according to the present embodiment, the original moving picture signal is decoded from the coded prediction error signal and the motion parameter information transmitted from the moving picture coding apparatus shown in FIG. 12 or 13. Can be converted.

Next, an embodiment of the image transformation apparatus according to the present invention will be described. (Embodiment 9) FIG. 15 is a block diagram showing an embodiment of an image transformation apparatus according to the ninth invention. This image transformation apparatus is composed of a fluid analysis unit 301 and a mapping unit 302. Coordinate information 311 of a representative point / line or area in the input image 313 and representative motion vector information 312 indicating the movement of the point / line or area indicated by the coordinate information 311 are input to the fluid analysis unit 301. .

The fluid analysis unit 301 uses the information 31
1, 312, the unknown motion vector information which is the motion vector information of the coordinate position other than the coordinate position specified by the coordinate information 311 in the input image 313 is obtained by the fluid analysis method described later, and the representative motion vector information 312 is obtained. Together with this, it outputs as motion vector information 314 of the entire area of the input image 313. In this embodiment, the coordinate position for obtaining the unknown motion vector information is defined in advance, and the resolution of the coordinate position of the unknown motion vector information is the representative motion vector information 31.
Sufficiently finer than that of 2.

The motion vector information 314 of the entire region obtained by the fluid analysis unit 301 is input to the mapping unit 302. The mapping unit 302 arranges (maps) the pixels of the input image 313 based on the motion vector information 314 of the entire region, and outputs the output image 315 obtained by transforming the input image 313.

In the fluid analysis unit 301, the input image 3
Regarding motion vector information discretely defined in all 13 areas, unknown motion vector information is obtained by numerical analysis based on known representative motion vector information 312. in this case,
The amount of calculation greatly depends on the method of numerical analysis. The fluid analysis method solves the basic equations of flow, such as continuity equation, motion equation, and energy equation. These equations are expressed as the following equations in the differential form.

[0111]

[Equation 1]

Where ρ is the density of the fluid, u is the velocity, f is the body force per unit mass, Φ is the stress tensor of the fluid, e is the internal energy per unit volume, k is the thermal conductivity, and T is the temperature. is there. Exactly solving these equations usually requires enormous calculations. For this reason, if a particularly complicated deformation effect is not required, unnecessary parameters are omitted from these to simplify the process. For example: (a) Eliminate the time partial differential term assuming a steady flow, (b) ignore the body force, (c) ignore the temperature gradient term, (d) ignore the body's stress tensor, ( e) Simplify the viscosity of the fluid contained in the fluid's stress tensor, and so on.

In the simplest case, assuming that the fluid is incompressible and inviscid and has no vortices, the velocity potential ψ can be defined, and the following Laplace equation holds in all regions of the flow.

[0114]

[Equation 2] This can be easily solved by numerical analysis by giving boundary conditions. Once ψ is obtained, the velocity u, which is the motion vector of the fluid, is given by the following equation.

[0115]

(Equation 3)

In the mapping unit 302, when the motion vector defined discretely is lower than the resolution in pixel units, the value of the motion vector in each pixel is determined by positioning one or more motion vector values in the vicinity of the target pixel with respect to each other. Interpolate considering the relationship. As the interpolation method, for example, PIC (Part
Nearest Grid Point method used in particle simulations such as icle-In-Cel) and Area Weighting
Method, and an interpolation method at an arbitrary position in an arbitrary shape by David Seldner, Thomas Westermann et al. (“Algolithms for In
terpolation and Localization in Irregular 2D Meshe
s ”, Journal of Computational Physics, Vol.79, No.
1, Nov. 1988) and the like can be used. This defines motion vectors for all pixels in the image.

(Embodiment 10) FIG. 16 is a block diagram showing another embodiment of the image transformation apparatus according to the ninth invention.
The same reference numerals are given to the portions corresponding to those in FIG. 15, and the difference will be mainly described. In the present embodiment, a feature amount detection unit 303 is newly added in addition to the configuration in FIG. The feature amount detection unit 303 checks the feature amount of the input image 313 based on the coordinate information 311 and the input image 313, and is motion vector information of a coordinate position other than the coordinate position designated by the coordinate information 311 in the input image 313. The coordinate position where the unknown motion vector information should be obtained is defined. As a method of determining this coordinate position,
For example, the method of using the edge of the image, SNAKES (M.Kass, A
Witkin, D. Terzopoulos, “SNAKES: Active Contour M
odels ”, International Journal of Computer Visio
n, Vol.1, No.4, pp.321-331, 1988) and other methods using representative points of contours, and analytical methods used in the fields of computer simulation of physical phenomena using pixel values as parameters. Lattice formation method and algebraic lattice formation method (For example, see “Numerric” by JF Tompson for analytical methods.
al Grid Genaration ", North-Holland, 1982), or even using grid points obtained by simply dividing the screen into equal parts, polygonal parts or random parts.

The fluid analysis unit 301 uses the position coordinates 316 of the unknown motion vector information defined by the feature amount detection unit 303 as described above, the representative motion vector information 312, and the position coordinates 311 thereof to implement the ninth embodiment. The unknown motion vector information is obtained by the method described in. Mapping unit 3
The process of 02 is also similar to that of the ninth embodiment.

(Embodiment 11) FIG. 17 is a block diagram showing an embodiment of an image transformation apparatus according to the tenth invention.
The same reference numerals are given to the parts corresponding to those in FIG. 16, and the difference will be mainly described. In this embodiment, a motion analysis unit 304 is further provided in addition to the configuration in FIG. The motion analysis unit 304 compares the input image 313 with the reference image 317, and uses the block matching method or the like to input image 313.
The motion vector information 312 discretely defined therein and the coordinate information 311 indicating the coordinate position thereof are obtained. As a method of calculating the motion vector, other methods such as the optical flow method may be used in addition to the block matching method.

In this case, the feature amount detection unit 303 has the input image 313 and the coordinate information 31 from the motion vector analysis unit 304.
Based on 1, the feature amount of the input image 313 is examined in the same manner as in the tenth embodiment, and the coordinate position for obtaining the unknown motion vector information in the input image 313 is defined. The processes of the fluid analysis unit 301 and the mapping unit 302 are the same as those in the ninth and tenth embodiments.

(Embodiment 12) FIG. 18 is a block diagram showing an embodiment of an image transforming apparatus according to the eleventh invention.
In the description of this embodiment, the parts corresponding to those in FIG. 17 are designated by the same reference numerals, and in the present embodiment, in addition to the configuration of FIG. 17, a motion vector updating unit 305, a motion vector / error database 306, and a comparison unit 307 It is provided.

When nothing is stored in the motion vector / error database 306, the motion vector updating unit 305 gives the representative motion vector information 312 input from the motion analyzing unit 304 to the fluid analyzing unit 301 as it is, and The value of the vector information 312 is written in the motion vector / error database 306. The comparison unit 307 outputs the output image 315 and the reference image 31 generated by the mapping unit 302 based on the motion vector information 314 of the entire region calculated by the fluid analysis unit 301 based on the motion vector information 312 at that time.
7 is compared to obtain the error of the output image 315, and the error information 3
18 is given to the motion vector updating unit 305. The motion vector updating unit 305 writes this error information 318 in the motion vector / error database 306 in association with the motion vector information 312.

When the error information 318 is smaller than a predetermined value, the motion vector updating unit 305 further gives the output request signal 320 to the switch 308 arranged on the output side of the mapping unit 302, thereby outputting the output image 315.
Control to take out. On the other hand, when the error information 318 is equal to or larger than the predetermined value, the motion vector updating unit 305 slightly changes the value of the motion vector information in the vicinity thereof, and gives new motion vector information 319 to the fluid analysis unit 301. However, when the number of updates of the motion vector information exceeds a preset value, the motion vector / error database 306
Further, the value of the motion vector information when the value of the error information 318 is the smallest is read and given to the fluid analysis unit 301,
The contents of the motion vector / error database 306 are erased and the output request signal 320 is given to the switch 308.

The fluid analysis unit 301, the mapping unit 302,
The processes of the feature amount detection unit 303 and the motion analysis unit 304 are
Same as Example 11. Examples 9 to 11 described above
According to the output image 315 obtained by transforming the input image 313.
To obtain the motion vector information 312 of a representative point / line or area in the input image 313 or the discrete motion vector information 312 between the input image 313 and the reference image 317, the unknown motion vector in the input image 313 is obtained. Information is obtained by fluid analysis to generate motion vector information 314 of the entire region of the input image 313, and pixels of the input image 313 are mapped based on the motion vector information 314 of the entire region to reduce the amount of calculation of the input image. An output image 315 obtained by transforming 313 is obtained.

Furthermore, by using a fluid analysis method based on fluid dynamics, complex physical phenomena such as compressible flow, viscous flow and vortex flow and discontinuous phenomena such as shock waves can be obtained as image effects.

The above-described image transformation apparatus according to the present invention can be applied to the motion compensation circuit and the image transformation section in the moving picture coding apparatus and the moving picture decoding apparatus described above, and also applied to the video effect as described above. can do.

[0127]

As described above, according to the present invention, the following effects can be obtained. In the moving picture coding apparatus according to the first aspect of the present invention, a feature point in the image of the reconstructed image signal that is the image signal of the previous frame is determined as a representative point, and the reconstructed image is calculated using the motion vector of this representative point. In order to generate a predicted image signal by performing motion compensation on the signal,
The amount of information of the prediction error signal is reduced and the coding efficiency is improved by performing the motion compensation adapted to the characteristics of the image signal of the previous frame. In this case, in the moving picture decoding apparatus, a representative point is determined in the same manner as the moving picture coding apparatus, and motion compensation is performed using the motion vector of the representative point to generate a predicted image signal, It is not necessary to transmit extra additional information such as patch shape information from the encoding device side as in the conventional case, and the encoding efficiency is not reduced.

In the moving picture decoding apparatus according to the second invention, the coding prediction error signal and the motion vector information transmitted from the moving picture coding apparatus according to the first invention are received and decoded. Reproduce the original video signal.

In the moving picture coding apparatus according to the third invention, the pattern around the representative point is analyzed for the reconstructed image signal, and the motion vector resolution information is obtained based on the analysis.
Since the motion vector of the representative point is obtained at the resolution specified by this motion vector resolution information, in the case of a complicated pattern, the motion vector is accurately detected with high resolution so that the prediction error signal becomes small, and the flat part of the image is detected. The resolution of the motion vector is reduced in the simple part of the pattern such that the accuracy is not higher than necessary, thus improving the coding efficiency by efficiently reducing the information amount of the motion vector and the motion vector search. Speed can be improved.

In the moving picture coding apparatus according to the fourth aspect of the present invention, the first motion-compensated picture signal obtained by performing motion compensation on the reconstructed picture signal and the motion compensation on the characteristic picture signal are obtained. Of the second motion-compensated image signals, the one having the smaller evaluation value of the error for each small area from the input image signal is selectively used as the predicted image signal, so that the reconstructed image signal, that is, the image signal of the previous frame is always generated. The prediction error becomes smaller and the coding efficiency is improved as compared with the conventional method in which only the motion-compensated signal is used as the prediction image signal.

In the moving picture decoding apparatus according to the fifth invention, the coded prediction error signal, the motion vector information and the first motion compensated picture signal transmitted from the moving picture coding apparatus according to the fourth invention. The original moving image signal is reproduced by receiving the selection information indicating which of the second and the second motion-compensated image signal has been selected as the predicted image signal and performing decoding.

In the moving picture coding apparatus according to the sixth and seventh inventions, motion parameters (motion compensation parameters) are quantized in order to reduce the amount of transmission information, and dequantized motion parameters are motion compensated. A motion compensation is performed on the reconstructed image signal as a motion vector for the motion vector to obtain a predicted image signal. In this case, due to the quantization error when quantizing the motion parameter, the motion vector for motion compensation deviates slightly from the optimum value in the sense that the prediction error is the smallest, and as a result, the information amount of the prediction error signal is reduced. It acts to increase.

Therefore, the input image signal is deformed by the image deformation parameter by the motion parameter corresponding to the quantization error, and the prediction error of the predicted image signal with respect to the obtained modified image signal is obtained. Since the prediction error obtained in this way has a small value equivalent to the prediction error before quantization of the motion parameter, the increase in the information amount of the prediction error signal can be suppressed low. Further, the reconstructed image signal in this case is an image signal obtained by modifying the input image signal by the quantization error, but the quantization error is less than 1/2 of the quantization step size, and Does not occur, the transmission information amount is reduced with almost no deterioration in image quality, and the coding efficiency is improved.

In the moving picture decoding apparatus according to the eighth invention, the moving picture decoding apparatus according to the sixth or seventh invention receives and decodes the coded prediction error signal and the motion vector information transmitted from the moving picture coding apparatus. To reproduce the original moving image signal.

In the image transforming apparatus according to the ninth to eleventh inventions, when an image obtained by transforming the input image is obtained,
The unknown motion vector information in the input image is obtained by fluid analysis from the motion vector information of a representative point / line or area in the input image or the discrete motion vector information between the input image and the reference image. By generating the motion vector information of the entire area and mapping the pixels of the input image based on this motion vector information, an output image obtained by transforming the input image with a small amount of calculation can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a moving picture coding apparatus according to a first embodiment.

FIG. 2 is a diagram showing representative points and triangular patches for explaining the operation of the embodiment.

FIG. 3 is a block diagram showing a configuration of a moving picture decoding apparatus according to a second embodiment.

FIG. 4 is a block diagram showing a configuration of a moving picture coding apparatus according to a third embodiment.

FIG. 5 is a diagram showing motion vector search points for explaining the operation of the embodiment.

FIG. 6 is a diagram showing polygon patch motion compensation.

FIG. 7 is a block diagram showing a configuration of a moving picture coding apparatus according to a fourth embodiment.

8 is a block diagram showing a configuration example of a characteristic image memory in FIG.

9 is a block diagram showing another configuration example of the characteristic image memory in FIG.

10 is an explanatory diagram of the size of the characteristic image memory in FIG.

FIG. 11 is a block diagram showing a configuration of a moving picture decoding apparatus according to a fifth embodiment.

FIG. 12 is a block diagram showing a configuration of a moving picture coding apparatus according to a sixth embodiment.

FIG. 13 is a block diagram showing a configuration of a moving picture coding apparatus according to a seventh embodiment.

FIG. 14 is a block diagram showing the configuration of a moving picture decoding apparatus according to an eighth embodiment.

FIG. 15 is a block diagram showing a configuration of an image transformation apparatus according to a ninth embodiment.

FIG. 16 is a block diagram showing the arrangement of an image transformation apparatus according to Example 10.

FIG. 17 is a block diagram showing the arrangement of an image transformation apparatus according to Example 11.

FIG. 18 is a block diagram showing a configuration of an image transformation apparatus according to a twelfth embodiment.

[Explanation of symbols]

101 ... Input image signal 102 ... Subtractor 103 ... Prediction error signal 104 ... Prediction error coding unit 105 ... Encoded prediction error signal 106 ... Prediction error decoding unit 107 ... Decoded prediction image signal 108 ... Adder 109 ... Reconstruction Image signal 110 ... Frame memory 111 ... Image signal of previous frame 112 ... Representative point determination unit 113 ... Representative point information 114 ... Motion compensation circuit 115 ... Predicted image signal 116 ... Motion vector information 117 ... Variable length coding unit 118 ... Variable length Encoding unit 119 ... Multiplexing unit 120 ... Encoding output 121 ... Picture analysis unit 122 ... Motion vector resolution information 131 ... Feature image memory 132 ... Feature image signal 133 ... Motion compensation circuit 134 ... Motion compensation image signal 135 ... Motion vector information 136 ... Error evaluation unit 137 ... Control unit 138 ... Selection control signal 139 ... Predicted image No. selection unit 140 ... Predicted image signal 161 ... Image transformation unit 162 ... Image transformation parameter 163 ... Deformed image signal 164 ... Motion analysis unit 165 ... Motion parameter 166 ... Quantization unit 167 ... Quantization value 168 ... Quantization error information 169 Dequantization unit 170 Motion vector information 172 Motion analysis unit 201 Input signal 202 Separation unit 203 Encoded prediction error signal 204 Side information 205 Variable length decoding unit 206 Variable length decoding unit 207 Coded prediction error signal 208 ... Motion vector information 209 ... Prediction error decoding unit 210 ... Prediction error signal 211 ... Adder 212 ... Reconstructed image signal 213 ... Frame memory 214 ... Previous frame image signal 215 ... Representative point determination unit 216 ... Representative point information 217 ... Motion compensation circuit 218 ... Predicted image signal 221 ... Selection control signal 222 ... motion vector information 231 ... feature image memory 232 ... feature image signal 233 ... motion compensation circuit 234 ... motion compensation image signal 235 ... predicted image signal selection unit 236 ... predicted image signal 241 ... motion parameter 242 ... inverse quantization unit 243 ... Motion vector information 301 ... Fluid analysis unit 302 ... Mapping unit 303 ... Feature amount detection unit 304 ... Motion analysis unit 305 ... Motion vector update unit 306 ... Motion vector / error database 307 ... Comparison unit 308 ... Switch 311 ... Coordinate information 312 ... Motion Vector information 313 ... Input image 314 ... Whole area motion vector information 315 ... Output image 316 ... Unknown motion vector coordinate information 317 ... Reference image 318 ... Error information 319 ... Motion vector information 320 ... Output request signal

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takeshi Datake 1 Komukai Toshiba-cho, Saiwai-ku, Kawasaki-shi, Kanagawa Incorporated Toshiba Research and Development Center

Claims (16)

[Claims] 1. A prediction error generation means for generating a prediction error signal indicating an error of a prediction image signal with respect to an input image signal, and a prediction error coding means for coding the prediction error signal to generate a coded prediction error signal. An image reconstruction unit that generates a reconstructed image signal from the encoded prediction error signal and the predicted image signal; a representative point determination unit that determines a feature point in the image for the reconstructed image signal as a representative point; A motion compensation means for obtaining a motion vector of the representative point, performing motion compensation on the reconstructed image signal using the motion vector, and generating a predicted image signal for the next input image signal. Video encoding device. 2. A moving picture decoding apparatus for decoding a moving picture signal by receiving at least a coding prediction error signal and motion vector information transmitted from the moving picture coding apparatus, wherein the coding prediction error signal A prediction error decoding unit that decodes the prediction error signal to generate a prediction error signal, an image reconstruction unit that generates a reconstructed image signal from the prediction error signal and the prediction image signal, and a feature in the image regarding the reconstructed image signal Representative point determining means for determining a point as a representative point, motion compensation is performed on the reconstructed image signal using the representative point and the motion vector, and a predicted image signal for generating the next reconstructed image signal. A moving picture decoding device, comprising: a motion compensating means for generating. 3. A prediction error generation means for generating a prediction error signal indicating an error of the prediction image signal with respect to the input image signal, and a prediction error coding means for coding the prediction error signal to generate a coded prediction error signal. An image reconstruction unit that generates a reconstructed image signal from the encoded prediction error signal and the predicted image signal; a representative point determination unit that determines a feature point in the image for the reconstructed image signal as a representative point; With respect to the reconstructed image signal, a pattern around the representative point is analyzed, and a pattern analysis unit that obtains motion vector resolution information indicating the resolution necessary for the motion vector based on the pattern is analyzed with the resolution specified by the motion vector resolution information. The motion vector of the representative point is obtained, the motion vector is used to perform motion compensation on the reconstructed image signal, and the next input image signal is obtained. Video encoding apparatus characterized by comprising a motion compensation unit that generates a predicted image signal. 4. A prediction error generation means for generating a prediction error signal indicating an error of the prediction image signal with respect to an input image signal, and a prediction error coding means for coding the prediction error signal to generate a coded prediction error signal. An image reconstructing means for generating a reconstructed image signal from the encoded prediction error signal and the predicted image signal; and a first motion compensation image generation unit for performing motion compensation on the reconstructed image signal. Motion compensation means, a characteristic image storage means for storing a characteristic image signal which is a signal of a characteristic region of the reconstructed image signal, and a second motion compensated image signal for performing motion compensation on the characteristic image signal. Of the first and second motion-compensated image signals having a smaller evaluation value of the error for each small area from the input image signal, for the next input image signal. Video encoding apparatus characterized by comprising a selecting means for selecting as the prediction image signal. 5. The prediction means uses the selection means as the predicted image signal when an evaluation value of an error between each of the first and second motion-compensated image signals and the input image signal for each small area is a predetermined value or more. No-signal is selected, and the characteristic image storage means stores, as the characteristic image signal, an image signal of a small area of the reconstructed image signal when the selecting means selects the no signal. The moving picture coding apparatus according to claim 4, wherein 6. The characteristic image storage means is in a predetermined state in which an evaluation value of a difference between the reconstructed image signal and the characteristic image signal read from the characteristic image signal storage means for each small area is larger than a set value. 5. The moving picture coding apparatus according to claim 4, wherein the reconstructed image signal when it continues for a period of time or more is stored as a characteristic image signal. 7. The characteristic image storage means stores a characteristic image signal having an image size larger than the image sizes of the input image signal and the reconstructed image signal. 2. The moving picture coding device according to item 1. 8. A moving picture decoding apparatus for decoding a moving picture signal by receiving at least a coding prediction error signal and information of a motion vector transmitted from the moving picture coding apparatus, wherein the coding prediction error signal A prediction error decoding unit that decodes the prediction error signal to generate a prediction error signal, an image reconstruction unit that generates a reconstructed image signal from the prediction error signal and the prediction image signal, and motion compensation for the reconstructed image signal. First motion compensation means for generating a first motion compensated image signal, a characteristic image storage means for storing a characteristic image signal which is a signal of a characteristic region of the reconstructed image signal, and the characteristic image signal Second motion compensating means for performing motion compensation on the above to generate a second motion compensated image signal, and transmitted from the moving image encoding device among the first and second motion compensated image signals. Video decoding apparatus characterized by comprising a selection means for selecting a person designated by the-option information as the predicted image signal to generate the next reconstructed image signal. 9. The characteristic image storage means stores an image signal of a small area of the reconstructed image signal when the selecting means selects no signal as the characteristic image signal. The moving picture decoding apparatus according to claim 8. 10. The moving image decoding according to claim 8, wherein the characteristic image storage means stores a characteristic image signal having an image size larger than the image size of the reconstructed image signal. Device. 11. An image transforming unit for transforming an input moving image signal according to an image transforming parameter to generate a transformed image signal, and a prediction for generating a prediction error signal indicating an error of the predicted image signal with respect to the transformed image signal. Error generation means, prediction error coding means for coding the prediction error signal to generate a coded prediction error signal, and image reconstruction for generating a reconstructed image signal from the coded prediction error signal and the predicted image signal. Configuring means, motion analyzing means for analyzing a motion between both image signals from the input image signal and the reconstructed image signal to extract a motion parameter, and quantizing the motion parameter to quantize the motion parameter and the image. Quantization means for generating transformation parameters, and inverse quantization for dequantizing motion parameters quantized by the quantization means to obtain motion vectors Means, performs motion compensation using the motion vector for said reconstructed image signal, the video encoding apparatus characterized by comprising a motion compensation unit that generates a predicted image signal for the next input image signal. 12. An image transforming unit that transforms an input moving image signal according to an image transforming parameter to generate a transformed image signal, and a prediction that generates a prediction error signal indicating an error of the predicted image signal with respect to the transformed image signal. Error generation means, prediction error coding means for coding the prediction error signal to generate a coded prediction error signal, and image reconstruction for generating a reconstructed image signal from the coded prediction error signal and the predicted image signal. Configuring means, motion analysis means for analyzing the motion between the image signals from the input image signal and the reconstructed image signal to extract the image deformation parameter and motion compensation parameter, and the motion compensation parameter Quantizing means for quantizing, dequantizing means for dequantizing the motion parameter quantized by the quantizing means to obtain a motion vector, A moving picture coding apparatus, comprising: a motion compensating unit that performs motion compensation on a reconstructed image signal using a motion vector and generates a predicted image signal for a next input image signal. 13. A moving picture decoding apparatus for decoding a moving picture signal by receiving at least a coded prediction error signal and information of a quantized motion parameter transmitted from the moving picture coding apparatus, wherein: Error decoding means for decoding a generalized prediction error signal to generate a prediction error signal, image reconstruction means for generating a reconstructed image signal from the prediction error signal and the predicted image signal, and the quantized motion Dequantizing means for dequantizing parameters to obtain a motion vector, and motion compensation using the motion vector for the reconstructed image signal to generate a predicted image signal for generating the next reconstructed image signal. A moving picture decoding apparatus comprising: a motion compensating means. 14. The coordinate information in the input image is designated based on coordinate information of a representative point / line or area in the input image and motion vector information indicating movement of the point / line or area. Fluid analysis means for obtaining unknown motion vector information of coordinate positions other than the coordinate position by fluid analysis and outputting motion vector information of the entire area of the input image, and based on the motion vector information output from this fluid analysis means, An image transformation apparatus, comprising: a mapping unit that obtains an output image by transforming the input image by mapping pixels of the input image. 15. A motion vector detecting means for detecting discrete motion vector information indicating a motion between an input image and a reference image, and a motion vector detecting means for detecting a feature amount in the input image to obtain a motion vector. Based on the motion vector information detected by the motion vector detecting means and the feature amount detecting means for determining the coordinate position that is not detected, the unknown motion vector information of the coordinate position determined by the feature amount detecting means is obtained by fluid analysis. And a fluid analysis unit that outputs motion vector information indicating the movement of the entire area between the input image and the reference image, and by mapping the pixels of the input image based on the motion vector information output from the fluid analysis unit. Mapping means for obtaining an output image obtained by transforming the input image. 16. A motion vector detecting means for detecting discrete motion vector information indicating a motion between an input image and a reference image, and a motion vector detecting means for detecting a feature quantity in the input image to obtain a motion vector. Based on the motion vector information detected by the motion vector detecting means and the feature amount detecting means for determining the coordinate position that is not detected, the unknown motion vector information of the coordinate position determined by the feature amount detecting means is obtained by fluid analysis. And a fluid analysis unit that outputs motion vector information indicating the movement of the entire area between the input image and the reference image, and by mapping the pixels of the input image based on the motion vector information output from the fluid analysis unit. Mapping means for obtaining an output image obtained by transforming the input image, and an output image obtained by this mapping means An error detecting means for detecting an error from the reference image and a motion vector updating means for updating the motion vector information detected by the motion vector detecting means when the error detected by the error detecting means is equal to or more than a predetermined value. An image transforming device comprising.