WO1996022656A1

WO1996022656A1 - Predictive encoding method for picture information

Info

Publication number: WO1996022656A1
Application number: PCT/JP1995/001952
Authority: WO
Inventors: Yasuo Suto; Masahiko Achiha; Katsumi Usami
Original assignee: Hitachi, Ltd.
Priority date: 1995-01-18
Filing date: 1995-09-27
Publication date: 1996-07-25
Also published as: TW312887B

Abstract

A moving picture is divided into a first block and a second block in such a way that they are shifted from each other by a half block. Movement vectors (V1 and V2) of the divided first and second blocks are acquired and two pieces of movement compensation picture information (BX1 and BX2) corresponding to the respect vectors are generated. Then movement compensation information (BXo) for inter-frame encoding is obtained by combining the two pieces of information (BX1 and BX2). Therefore, prediction encoding processing can be performed with nearly the same accuracy as that when a moving picture is divided into smaller blocks, and the amount of movement vector information can be reduced to about 1/2 of the amount which is obtained when a moving picture is divided into smaller blocks.

Description

Description Predictive coding method for image information

The present invention relates to a motion compensation technique used for video coding, for example, a technique effective when applied to a video processing apparatus included in a TV (television) telephone system in ISDN, a video conference system, and the like. . Background art

In the case of moving images, in a scene with little motion, there is a large correlation between adjacent frames, and large data compression is performed using this. Such data compression is called interframe coding. In inter-frame coding, in order to perform more efficient information compression, a motion-compensated predictive coding scheme that uses a correlation between a previous frame and a current frame and encodes and transmits only a moving part is relatively efficient. Many have been adopted. For example, in a TV telephone system or a TV conference system, a TV camera is fixed and an object to be photographed by the TV camera is exclusively a human. In this case, a still portion such as a background in a whole photographed image occupies a large proportion. Also, when several frames are drawn on the TV screen per second, the difference in the image between adjacent frames is small. In such a case, the stationary portion, which is the redundant portion of the image information between frames, is not transmitted, and the difference portion, that is, the portion where the image moves, is detected, and only that portion is encoded. This is the motion compensation prediction coding method, which enables large information compression.

Specifically, the motion vector of the target area is detected between the current frame and the previous frame, and the position shifted by the motion vector in the previous frame is detected. As the predicted value, the difference (prediction error) between the predicted value and the pixel of interest in the current frame is transmitted. The receiving side holds the information of the previous frame, and obtains the image of the current frame by using the motion vector transmitted together with the prediction error in a procedure generally opposite to the above.

As a method of detecting a motion vector, there is a technique called block matching or pattern matching. As a reference describing the motion compensation prediction, there is “Interface (issued on December, 1991, p. 152)” published by CQ Publisher.

In the motion compensation prediction method, it is effective to reduce the prediction error as much as possible to increase the compression rate and reduce the amount of encoded data. If the size of the block is reduced and fine motion compensation is performed, the prediction accuracy can be improved, but then the number of blocks per frame increases, and the motion vector to be transmitted is increased. The data volume increases. On the other hand, when the size of a block is increased, the amount of data in the motion vector decreases, but the amount of encoded data increases due to a large motion compensation prediction error, and the block-like distortion increases.

An object of the present invention is to eliminate motion vector information.

Another object of the present invention is to realize accurate motion compensation prediction with reduced block distortion. Disclosure of the invention

According to the present invention, input image information used for motion compensation for inter-frame encoding of a moving image is divided into blocks of a plurality of pixels, and image information similar to a target block in a current frame is obtained from information of a previous frame. A method for detecting a motion vector for predictive encoding of input image information based on the information of the previous frame by determining the position of an area having This is divided into a plurality of first blocks and a plurality of second blocks arranged in such a manner that the first block is shifted half a block vertically and horizontally, and the first block and the second block are divided. The motion vectors (V1, V2) are detected for each.

The motion vectors obtained by the two types of block segmentation are used to generate motion-compensated predicted image information (ΒΧ Ι, BX 2) using the image information of the previous frame for each motion vector of the first and second blocks. To be served.

For example, assuming that the size of the first and second blocks is NX N pixels, respectively, the positions of the first and second blocks are shifted vertically and horizontally by a half block (NZ2), and the motion vector for the first block and the second block are shifted. When a motion vector corresponding to is detected, two types of motion vectors are obtained for each area (reference block) of one size smaller than the size (NZ2) (N / 2). The final predicted value (BXo) is obtained by combining the motion-predicted predicted image information (also simply referred to as predicted value) created based on the motion vector obtained by the two types of block segmentation. Therefore, it is possible to obtain the final predicted value with the same high accuracy as that in the case of detecting the motion vector in the one-step small-sized block unit. In this way, the two systems are divided into blocks so that they are shifted from each other by half a block, the motion vector is obtained for each of the divided blocks, motion-assisted image information is generated, and these are synthesized. This makes it possible to perform predictive coding with almost the same accuracy as when the block size is reduced by one level.

In terms of the amount of motion vector information required (the number of motion vectors), the motion vector is smaller than in the case where the motion vector is detected in units of one-step smaller blocks. Is almost halved. For example, one frame is divided into K blocks vertically and L blocks horizontally. When the motion vector is detected by simply setting the block size to 12, the number of motion vectors is 2 KX 2 L = 4 K. On the other hand, as in the present invention, when detecting a motion vector by performing two types of block separation with a shift of a half block without changing the block size, the number of motion vectors is determined. Is

K L + (K + 1) X (L + 1) = 2 K L + K + L + 1

Thus, the number of motion vectors can be reduced to about 1 Z 2. Therefore, in consideration of the fact that the information to be transmitted or the information compressed by encoding must include a motion vector in the motion-compensated predictive coding, the amount of transmitted information is reduced, and the moving image compression ratio is reduced. Improvement can be achieved. As described above, the motion vector information amount can be reduced to about 12 in the case where the block division is performed with the one-step smaller size.

Further, in order to reduce the block distortion, the motion compensated prediction image information (BX 1) related to the first block and the motion compensated prediction image information (BX 2) related to the second block are Weighting processing for reducing the weight of the end pixels is performed, and they are combined to obtain final motion-compensated predicted image information. That is, in the motion compensated prediction image created based on the motion vector, block distortion is generally caused due to poor matching of the motion compensated prediction image with an adjacent block at a block end. At this time, the first block and the second block are separated from each other in the vertical and horizontal half-block size phases, so that the above-described block distortion also occurs when the half-block size phase is shifted. Focusing on this feature, the combining means (11) performs a weighting process to eliminate the weight of the predicted value corresponding to each block edge of the first block and the second block. Thereby, the above-mentioned block distortion can be reduced by g. The generation of the motion-compensated predicted image information can be performed for each reference block having the size of the half block vertically and horizontally. In this case, the weighting process is performed on the motion compensated prediction image information of the reference block relating to one of the first block and the second block with respect to the reference block in the vertical and horizontal directions. 2 is performed by multiplying by a function of sin ² and the motion-compensated prediction image information of the reference block relating to one of the first block and the second block is vertically and horizontally aligned with the reference block. On the other hand, they can be performed by multiplying the functions of c 0 s ² each having a period of π / 2. The motion-compensated predicted image information (Β ο ο) obtained by the synthesis is subtracted from the corresponding image information of the current frame, whereby prediction error information (ε) is obtained.

Furthermore, when a motion vector is detected by two types of block segmentation shifted from each other by half the vertical and horizontal half blocks, two types of motion vectors are obtained for each one-step smaller area in each of the vertical and horizontal directions in each block. By selecting a motion vector with higher accuracy for each block and using it to create a motion-compensated prediction image, the prediction error can be reduced, so that a high-precision image can be decoded. Will be possible.

A moving picture processing apparatus to which the motion vector detection method or the predictive coding method is applied, converts input image information (X) used for motion compensation for inter-frame coding of a moving image into a plurality of pixels. The prediction code of the input image information based on the information of the previous frame is determined by determining the position of a region having image information similar to the first block of interest in the current frame from the information of the previous frame by dividing into one block. A first motion vector detecting means (37) for detecting a first motion vector (VI) for image transformation, and an arrangement in which the input image information is arranged by shifting the first block by half a block vertically and horizontally. Image information similar to the second block of interest in the current frame from the information of the previous frame A second motion vector detecting means for detecting a second motion vector (V 2) for predictive encoding of input image information based on the information of the previous frame by determining the position of the region having (38) and first motion compensation for creating motion-compensated predicted image information (BX 1) for the first block using the first motion vector of the first block and image information of the previous frame. Means (9), and a second motion for generating motion compensated prediction image information (BX 2) for the second block using the second motion vector of the second block and the image information of the previous frame. Compensation means (10); motion compensation obtained by combining the motion compensated prediction image information created by the first motion compensation means with the motion compensation prediction image information created by the second motion compensation means. First combining means (11) for creating predicted image information; Motion compensated prediction picture information and the current frame of the corresponding image information and subtracting means for obtaining a prediction error information by subtracting the

(4) and can be configured. The synthesizing unit reduces the weight of the block end pixel, as described in the method, for the motion compensation prediction image information related to the first block and the motion compensation prediction image information related to the second block. Weighting process for them and synthesize them.

The moving image processing apparatus further includes a first storage unit (3) for holding the image information of the current frame, and a motion compensated prediction image information that is combined with the prediction error information generated by the subtraction unit. It is possible to have a first adding means (8) for adding, and a second storage means (50) for holding the image information obtained by the first adding means as image information of a previous frame. is there. In consideration of a configuration for decoding the information encoded by the above configuration, the video processing device includes the prediction error information (BX o) and the first and second motion vectors (VI, V 2) And means for transmitting (13, 14). And coupled to the transmission means via a transmission line (15) or the like, Means for receiving the prediction error information and the first and second motion vectors (16, 17); and motion compensation using the received first motion vector and image information of the previous frame on the receiving side. Third motion compensating means (39) for generating predicted image information, and fourth motion generating means for generating a motion-compensated predicted image information using the received second motion vector and the image information of the previous frame on the receiving side. A motion compensation stage (40), and combining the motion compensated prediction image information created by the third motion compensation means with the motion compensated prediction image information created by the fourth motion compensation stage. Second combining means (52) for creating the motion-compensated predicted image information, and decoding by adding the synthesized motion-compensated predicted image information (BX o) and the received prediction error information (ε '). Second addition means (19) for acquiring image information (X '); and Image information and a third storage means (5 1) for holding an image data of the receiving front frame. A moving image processing apparatus that applies a method of creating a motion-compensated predicted image using the accuracy method among the two types of blocks detected by two types of block separation shifted vertically and horizontally by half blocks. The input image information (X) used for motion compensation for inter-frame coding of a moving image is divided into first blocks of a plurality of pixels, and the information of the previous frame is used to extract the first block of interest in the current frame. Motion vector detection for detecting the first motion vector for predictive encoding of input image information based on the information of the previous frame by determining the position of an area having image information similar to that of the previous frame. Means (37), the input image information is divided into a second block in which the first block is displaced by half a block in the vertical and horizontal directions, and the information of the previous frame is similar to the second block of interest in the current frame. Image of A second motion vector detecting means for detecting a second motion vector for predictive encoding of input image information based on the information of the previous frame by determining a position of an area having information; And the more accurate one of the detected first motion vector and second motion vector. First compensation means (31, 32) for selecting a motion vector, and motion compensation prediction for the selected motion vector using the selected motion vector and the image information of the previous frame. A first motion compensation means (33) for generating image information (BXo); and a prediction by subtracting the motion-compensated prediction image information generated by the first motion compensation means and the corresponding image information of the current frame. And subtraction means (14) for obtaining error information (ε). Furthermore, the first storage means (3) for holding the image information of the current frame, and the prediction error information (f) generated by the subtraction means, First adding means (8) for adding motion-compensated predicted image information (BXo), and second convenient means (5) for holding the image information obtained by the adding means as image information of the previous frame. 0). Considering a configuration for decoding the encoded information for the moving image processing apparatus as well, the prediction error information, the first motion vector, the second motion vector, and the first selecting means Means (13, 14) for transmitting the control information (g) for designating the motion vector selected in step (1), and means for transmitting the information via a transmission line (15) or the like. Receiving means (16, 17) and a second selecting means (4) for selecting a motion vector specified by the control information from the received first motion vector and second motion vector. 3) and a second motion compensation means (44) for creating motion-compensated predicted image information using the motion vector selected by the second selection means and the image information of the previous frame on the receiving side; The motion-compensated predicted image information formed in the second motion compensation step and the received prediction error information are added. Second adding means (19) for obtaining decoded image information by calculating the third information, and a third storage means for holding the image information obtained by the second adding means as image information of the previous frame on the receiving side. The moving image processing apparatus can further comprise means (51) and (25). BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a block diagram of a coding system prediction processing unit in a TV telephone system according to one embodiment of the present invention.

FIG. 2 is a block diagram of a decoding system prediction processing unit in the TV telephone system.

FIG. 3 is a block diagram of an image communication system in the TV telephone system.

FIG. 4 is an explanatory diagram of first and second blocks used in the processing of the encoding system prediction processing unit.

FIG. 5 is a diagram for explaining the basic principle of motion compensation in video coding. FIG. 6 is an explanatory diagram of the weighting process for the motion-compensated predicted image C 1 (X, Y).

FIG. 7 is an explanatory diagram of the weighting process for the motion compensated prediction image C 2 (X, Y).

FIG. 8 is an explanatory diagram of an example of a block size and a reference block size, which are units of motion vector detection in the present embodiment.

FIG. 9 is an explanatory diagram of separation of the first block in the device of the present embodiment.

FIG. 10 is an explanatory diagram of the separation of the second block in the apparatus of this embodiment.

FIG. 11 is a flowchart showing the first half of the encoding process using the apparatus of this embodiment.

FIG. 12 is a flowchart showing the latter half of the encoding process following the process of FIG.

FIG. 13 is a block diagram of an encoding prediction processing section in a system according to another embodiment of the present invention. FIG. 14 is a block diagram of a decoding system prediction processing unit in a system according to another embodiment of the present invention.

FIG. 15 is a flowchart showing the first half of an encoding process using a system according to another embodiment of the present invention.

FIG. 16 is a flowchart showing the latter half of the encoding process following the process of FIG. BEST MODE FOR CARRYING OUT THE INVENTION

In order to explain the present invention in more detail, this will be described with reference to the accompanying drawings.

Here, first, in order to facilitate understanding of the present invention, the basic principle of the motion vector detection method will be described.

As shown in FIG. 5, a pixel at the pixel position (N k, N 1) at the upper left corner of each block 23 composed of NX N pixels is now assumed that the image 21 of the m-th frame is the encoding target frame. Let the value be Cm (Nk, Nl). Here, the sum of absolute values S, i of the difference between the block 24 in the previous frame (the (m-1) th frame) 22 whose position is shifted by (i, j) and the current frame (the mth frame) is expressed by the following equation. Calculate as follows.

N-l N-1

S i j = ∑ ∑ I Cm (N k + p, N l + q)

^{P =} ° ^{q =} °-Cm- l (N k + p + i, N l + q +) I And variously change (i, j), for example, change from 0 to Sat I, 0 to ± J, Calculate S u and find the minimum value S. The (i ', j') power corresponding to the found minimum value S-factory is taken as the motion vector at that time.

Motion-compensated predictive coding calculates the pixel value Cm (N k + p, NI + p), using the pixel value Cm-1 (Nk + p-, N 1 + p-j ') of the block shifted by (, j') in the previous frame as the prediction value Perform encoding. An image composed of the predicted values is a motion-compensated predicted image. The encoded output data is obtained by encoding the motion vector values ', j') and NXN prediction error values (difference value between the value relating to the current frame and the prediction value) for one block. Things.

In contrast to the basic principle described above, the motion vector detection according to the present invention, as shown in FIGS. 4 (A) and 4 (B), shows the size of a frame or a specific image area. The first block and the second block, which are equal to each other and have different division positions, are prepared, and the motion vector is individually obtained for each block. The division of the second block shown in FIG. 4 (B) is shifted from the block division shown in FIG. 4 (A) by half a block in each of the vertical and horizontal directions.

Assuming that the sizes of the first and second blocks are NXN pixels, respectively, the positions of the first and second blocks are shifted vertically and horizontally by NZ2, and the kill vector for the first block and the motion vector for the second block are shifted. When a vector is detected, two types of motion vectors are obtained for each one-step smaller area of size (NZ2) (N / 2). The final predicted value is obtained by combining motion-compensated predicted image information (also simply referred to as predicted value) created based on the motion vector obtained by the two types of block segmentation. Therefore, a final predicted value can be obtained with the same high precision or the same small prediction error as in the case where the motion vector is detected in units of one-step smaller size blocks.

At this time, in terms of the amount of information of the required motion vector (the number of motion vectors), the motion vector is set in units of blocks each having the smaller size by one step. The number of motion vectors is almost halved compared to the case of detecting. For example, based on the case where one frame is divided into K blocks vertically and L blocks horizontally, the motion vector when the motion vector is detected by simply setting the block size to 12 is used. Is 2 KX 2 L = 4 KL. On the other hand, as in the present invention, when the motion vector is detected by performing two types of block separation with a shift of half a block without changing the block size, the motion vector is not changed. The number is

K X L + (K + 1) X (L + 1) = 2 K L + K + L + 1

Thus, the number of motion vectors can be reduced to about 12. Therefore, in consideration of the fact that in the motion compensation predictive coding, the information to be transmitted or the information compressed by the encoding must include a motion vector, the amount of transmitted information can be reduced, and the moving image compression can be reduced. Rate improvement can be achieved.

In addition, a motion-compensated predicted image created based on a motion vector generally causes block distortion due to poor matching of the motion-compensated predicted image with an adjacent block at a block end. At this time, since the first block and the second block are separated from each other in the vertical and horizontal half block size phases, the above-described block distortion also occurs when the half block size phase is shifted. Will be. Focusing on this feature, weighting processing is performed to reduce the weight of the predicted value corresponding to each block end of the first block and the second block. Thereby, the block distortion can be reduced.

Also, as described above, the positions of the first and second blocks are shifted vertically and horizontally by two to detect the motion vector for the first block and the motion vector for the second block. 2) Two types of motion vectors can be obtained for each smaller area of X (N / 2). In each block, of the two types of motion vectors obtained by the minimum value S j · of the sum of absolute values S ii, the motion vector having the smaller minimum value S j ′ (relative accuracy By selecting a motion vector with a high degree of motion and using it to create a motion-compensated prediction image, the prediction error can be reduced, so that it is possible to decode an accurate 髙ぃ image.

Hereinafter, an embodiment corresponding to the above-described principle will be described. FIG. 3 shows a moving image communication device in a TV telephone system which is an embodiment of the moving image processing device according to the present invention.

The moving image communication device includes a transmitting device SEN for transmitting moving image data, a receiving device for receiving the moving image data transmitted from the transmitting device SEN, which is coupled to the transmitting device SEN by a communication line 15. The equipment REC is established. The communication line 15 is ISDN, though not particularly limited. The transmission device SEN is configured as follows, for example. An image receiving unit 1 such as a TV camera for inputting a moving image is provided, and an AZD (analog digital) converter 2 for converting the moving image data from the image receiving unit 1 into a digital signal is provided at a subsequent stage. ing. At the subsequent stage of the AZD converter 2, a frame memory 3 for storing digital signals from the AZD converter 2 is provided. The frame memory 3 has a storage capacity of at least one frame of an image, and stores a current frame to be encoded. At the subsequent stage of the frame memory 3, an encoding prediction processing unit 26 (to be described in detail later) for performing inter-frame prediction encoding processing is provided. Further, a subtracter 4 for obtaining a prediction error f by subtracting the image information X output from the frame memory 3 and the predicted value BXo output from the coding prediction processing unit 26 is provided. Further, a switch 5 for selectively transmitting the prediction error ε output from the subtractor 4 and the image information X from the frame memory 3 to the subsequent quantization section 6 is provided. Have been.

In the quantization unit 6, when the image information X is selected by the switch 5, the quantization of the image information X is performed, and when the prediction error ε is selected by the switch 5, the prediction error ε is calculated. Quantization is performed. The quantization value output from the quantization unit 6 is denoted by q. At the subsequent stage of the quantization unit 6, an inverse quantization unit 7 for performing inverse quantization of the quantization value q is provided. The inverse quantization unit 7 outputs a decoded image signal X ′ for the image signal X, and outputs a prediction error f ′ for the prediction error £. A switch 25 for selecting a predicted value BXo output from the encoding system prediction processing unit 26 is provided, and the predicted value BXo (at the time of inter-frame predictive encoding) or 0 value (frame) is provided. An adder 8 is provided for adding the output of the inverse quantization unit 7 and the output of the inverse quantization unit 7. An encoding control unit 12 for controlling whether to use inter-frame predictive encoding or intra-frame encoding is provided, and the encoding control signal p output from the encoding control unit 12 provides The operation of switches 5 and 25 is controlled. That is, when inter-frame prediction is selected by the coding control unit 12, the coding control signal p is set to the high level, and the prediction error ε is selectively transmitted to the quantization unit 6 by the switch 5. You. Further, at this time, the predicted value Β ο ο is transmitted to the adder 8 via the switch 25, and is added to the prediction error f ′ which is the output value from the inverse quantization unit 7. The result of the addition is used as decoded image information X ′, which is input to the encoding prediction processing unit 26. The motion vectors VI and V 2 are obtained from the coding prediction processing unit 26, and these are used as the coding control signal p from the coding control unit 12 and the motion vector VI and V 2 from the self quantization unit 6. After being multiplexed together with the quantized value q by the signal multiplexing unit 13 at the subsequent stage, it is transmitted to the communication line 15 via the transmission unit 14 at the subsequent stage. In addition, when the intra-frame encoding is specified by the control signal q (q = low level ), The image information X is given to the quantization unit 6 by the switch 5, the 0 value is supplied from the switch 25 to the adder 8, and the output of the adder 8 is the decoded image which is the output of the inverse quantization unit Ί. Information X ′ is supplied to the encoding system prediction processing unit 26. The decoded image information X ′ supplied to the encoding system prediction processing unit 26 is handled as information of a previous frame on the transmission side in a later process. The receiving device REC is configured as follows, for example. A receiving unit 16 for capturing moving image data input via the communication line 15 is provided, and the moving image data received by the receiving unit 16 is transmitted to a subsequent signal unit I 17. To be transmitted. The signal separating section 17 is provided corresponding to the signal multiplexing section 13 in the transmitting section SEN. The moving image data transmitted from the receiving section 16 has a quantization value q and a motion vector VI. , V 2, and the coding control signal p. At the subsequent stage of the signal separation unit 17, an inverse quantization unit 18 is arranged. By inverse quantization of the quantization value q transmitted from the signal separation unit 17, a prediction error (at the time of frame 閒 prediction encoding), Alternatively, decoded image information X '(at the time of intra-frame encoding) is obtained.

Also, based on the motion vectors V 1 and V 2 from the signal separation unit 17, a decoding system prediction processing unit 27 for obtaining a prediction value 動き Χ ο for motion compensation, the signal separation unit 1 A switch 22 whose operation is controlled by an encoding control signal ρ from the decoding control signal ρ from the decoding system prediction processing unit 27. ) Or 0 value (in intra-frame encoding) is selectively transmitted to the adder 19. During the inter-frame prediction encoding, the adder 19 calculates the prediction error f ′ transmitted from the inverse quantization unit 18 and the prediction value BX ο transmitted from the decoding prediction processing unit 27. Addition processing is performed, and at the time of intra-frame encoding, addition processing of the decoded image signal X, and the 0 value is performed. The decoded image information X ′ output from the adder 19 is supplied to a decoding prediction processing circuit 27. The decryption The decoded image information X ′ supplied to the system prediction processing unit 27 is handled as information of the previous frame on the receiving side in a later process.

Further, an AZD converter 23 for converting the output signal of the adder 19 into an analog signal is provided at a stage subsequent to the adder 19, and the converted output is provided to a TV monitor 24 at the subsequent stage. Is displayed.

FIG. 1 shows a detailed configuration example of the coding system prediction processing unit 26. The content shown in the figure focuses on inter-frame predictive coding.

As shown in FIG. 1, although the coding system prediction processing unit 26 is not particularly limited, the frame memory 50, the motion vector detection units 37, 38, the motion compensation units 9, 10, And a combining unit 11.

The frame memory 50 stores the decoded image information X, output from the adder 8. This decoded image information X ′ is handled as image information of the previous frame in a later process. The data stored in the frame memory 50 can be transmitted to the motion vector detection units 37 and 38 and the motion compensation units 9 and 10. The motion vector detector 37 compares the input image information X with the image information of the previous frame in the frame memory 50 using the first block shown in FIG. The motion vector V1 is detected for each block of the image. The motion vector detector 38 compares the input image information X with the image information of the previous frame in the frame memory 50 using the second block shown in FIG. The motion vector V2 is detected for each block of the image. The motion compensating unit 9 generates a predicted value BX 1 by moving the position of the previous frame supplied from the frame memory 50 by using the detected motion vector V 1. The motion compensating unit 10 generates the predicted value BX2 by moving the position of the previous frame supplied from the frame memory 50 using the detected motion vector V2. In the synthesizing unit 11, although not particularly limited, After performing weighting processing using a trigonometric function on each of the predicted values BX1 and BX2 output from the motion compensation units 9 and 10 (details will be described later), the final values are synthesized by combining them. The predicted value BX o. The predicted value BXo is transmitted to the above-mentioned subtractor 4 switch 25 shown in FIG.

FIG. 2 shows a detailed configuration example of the decoding system prediction processing unit 27. The content shown in the figure focuses on inter-frame predictive coding.

As shown in FIG. 2, the decoding system prediction processing unit 27 includes, but is not limited to, a frame memory 51, motion compensation units 39 and 40, and a synthesis unit 52.

The decoded image information X ′ output from the adder 19 is stored in the frame memory 51. The decoded image information X ′ stored in the frame memory 51 is handled as the image information of the previous frame on the receiving side. The data stored in the frame memory 51 can be transmitted to the free compensators 39 and 40. The motion compensator 39 moves the position of the previous frame supplied from the frame memory 51 by the motion vector VI transmitted from the signal separator 17 to generate a predicted value BX1. The motion compensator 40 moves the position of the previous frame supplied from the frame memory 51 by using the motion vector V2 transmitted from the signal separator 17, and generates a predicted value BX2. The combining unit 52 performs a predetermined weighting process on the predicted value BX1 and the predicted value BX2 from the motion compensation units 39 and 40, and then combines them to obtain the final predicted value. Get the value BX o. The predicted value BX0 is transmitted to the switch 22 shown in FIG. That is, the decoding system prediction processing unit 27 performs the same processing as the series of processing by the motion compensation units 9 and 10 and the synthesis unit 11 in the coding system prediction processing unit 26 described above. Accordingly, processing for decoding the received image is performed.

Next, the encoding process in the present embodiment will be described in detail. FIGS. 8 to 10 show the above-described two types of block division. In the processing by the motion vector detection sections 37 and 38, as shown in FIG. 8, the detection processing of the motion vector is performed in blocks of 32 × 32 pixels separated by a solid line. Otherwise, the processing is performed in units of 16 × 16 pixel blocks (hereinafter referred to as reference blocks).

The first block (corresponding to FIG. 4 (A)), which is cut off by the solid line shown in FIG. 9, is used by the motion vector detection unit 37. The second block (corresponding to FIG. 4 (B)) separated by a solid line shown in FIG. 10 is used by the motion vector detection unit 38. 9 and 10, the reference block is denoted by X (p, q), the first block is denoted by LI (x, y), and the second block is denoted by L 2 (X, y). Notation.

11 and 12 show flow charts of the decoding process using the apparatus of the present embodiment. The symbols # 1 and # 2 shown in Fig. 11 and the symbols # 1 and # 2 shown in Fig. 12 indicate that the processing is repeated between the corresponding symbols. .

For convenience of explanation, the size of the image is assumed to be a size (16 N pixels X 16 M pixels) defined by N reference blocks in the horizontal direction and M reference blocks in the vertical direction. In the flowcharts shown in FIG. 11 and FIG. 12, the processing is performed while scanning for each reference block from the upper left to the lower right of the image in the horizontal direction. In the following processing, division by 2 (2) is performed.

First, assuming that p == q = 0 (step S1), the processing is started from the upper left of the image. The processing of steps S2A to S5A depends on the first block. This is processing related to motion vector detection, and this processing is performed by the motion vector detection unit 37 and the motion compensation unit 9. The processing of steps S2B to S5B is processing related to motion vector detection by the second block, and this processing is performed by the motion vector detection unit 38 and the motion compensation unit 10. Done.

Next, it is determined whether or not the reference block to be encoded is the reference block located at the upper left of the first block or the second block (steps S2A and S2B). That is, the motion vector detection unit 37 determines whether or not (p = even number and q = even number) holds for the reference block X (p, q) to be coded. (Step S2A). 0 is treated as an even number. In addition, the motion vector detection unit 38 calculates (p = 0, q = 0), (p = 0, and 0) for the reference block X (p, q) to be encoded from now on. It is determined whether any of the following holds: (q = odd number), (p = odd number and q = 0), or ( _p = odd number and q = odd number) (step S2B).

If it is determined in the above step S 2 A that the answer is yes, that is, if it is determined that (P = even number and q = even number) holds, it is coded from now on Reference block X (p, q) force Means that it is located at the upper left of the first block. This is clear from FIG. In this case, the motion vector detection unit 37 searches the frame memory 50, and the motion vector V1 of the first block LI (p / 2, q / 2) is obtained. The signal is latched by an internal register and transmitted to the motion compensator 9, and is also transmitted to the signal multiplexer 13 for signal multiplexing (steps S3A and S5A). The motion compensator 9 creates motion-compensated predicted image information BX1 from the previous frame in the frame memory 50 based on the motion vector V1 sent by fe (step S5A). This motion compensated prediction image information BX1 has the size of one reference block. If it is determined as no in the determination in step S 2 A, that is, if it is determined that (p = even number and q = even number) is not satisfied, the first block at that time is determined. , A motion-compensated predicted image information BX 1 of the reference block size is created based on the motion vector V 1 currently latched in the register in the motion vector detector 37. That is, the detection of the motion vector for one first block is performed based on the image information of the reference block located at the upper left of the first block, and the four reference blocks included in the first block are detected. The motion compensation prediction image information for each is generated using the motion vector detected based on the reference block located at the upper left corner.

If it is determined to be yes in the determination in step S 2 B, that is, (p = 0 and q = 0), (p = 0 and q = odd), (p = odd and If it is determined that either of (q = 0) or (p = odd and q = odd) is satisfied, it means that the reference block X (p, q) to be coded is It means that it is located at the upper left of the second block. This is clear from FIG. In that case, the motion vector detection unit 38 searches the frame memory 50 to create a motion vector V2 of the second block. At this time, if (p = q = 0) holds, the motion vector V 2 of L 2 (0, 0) is detected, and if (p = 0 and q = odd number) holds, Is that if a motion vector V 2 of L 2 (0, q / 2 + 1) is detected and (p = odd and q = 0), then L 2 (p 2 + 1 , 0) is detected, and if (p = odd number and q = odd number) is satisfied, the motion vector of L 2 (p / 2 + 1, q / 2 + l) is obtained. V 2 is detected. Then, the detection result is latched in an internal register and transmitted to the motion compensation unit 10. At the same time, the signal is sent to the signal multiplexing unit 13 for signal multiplexing (step S4B). The motion compensation unit 10 creates a motion-compensated predicted image BX2 from the previous frame in the frame memory 50 based on the transferred motion vector V2 (step S5B). This motion compensated prediction image information BX2 has the size of one reference block. Further, when it is determined to be no in the determination in step S 2 B, that is, (p = 0 and q = 0),

If it is determined that none of (p = 0 and q = odd), (p = odd and q = 0), or (p = odd and q = odd) is not satisfied, For the second block at that time, motion-compensated predicted image information BX2 of the reference block size is created based on the motion vector V2 currently latched in the register in the motion vector detection unit 38. That is, the detection of the motion vector for one second block is performed based on the image information of the reference block located at the upper left of the second block, and the four reference blocks included in the second block are detected. The respective motion-compensated prediction image information for the block is generated using the motion vector detected based on the reference block located at the upper left.

The motion-compensated predicted image information BX 1 and BX 2 created in the above steps S 5 A and S 5 B are transmitted to the synthesizing unit 11, where the motion-compensated predicted image information BX 1 and BX 2 Then, a weighting process using a trigonometric function as shown in FIGS. 6 and 7 is performed and synthesized to create motion-compensated predicted image information BXo (step S6).

The weighting is performed so as to lose the weight of the block end pixel. That is, the motion-compensated predicted image information created based on the motion vector generally causes block distortion due to poor matching of the motion-compensated predicted image with an adjacent block at a block end. At this time, the first block and the second block are separated from each other by a vertical and horizontal half block size phase shift. Therefore, the above-described block distortion also occurs when the half-block size phase is shifted. Focusing on this feature, weighting processing is performed to reduce the weight of the motion-compensated predicted image information (predicted value) corresponding to each block end of the first block and the second block. The weight here is sin ² (πΧN) sin ² (π Y / N) for the first block, taking into account the vertical and horizontal components as shown in Fig. 6. The second block is cos ² (jr X / N) Xcos ² (π Y / N) in consideration of the vertical and horizontal components as shown in FIG. The weighting function is a function of sin ² and cos ² in which the vertical and horizontal directions of the reference block have a period of 2 respectively. The reason for using the trigonometric function is as follows. As is clear from comparison of Figs. 6 and 7, the weight of the first block and the second block decreases at the end of the block, and the weights of the first block and the second block are reduced. Paying attention to the state in which the predicted values for (the summation and synthesis) are combined, the first block and the second block have a phase shift of π / 2 in the above weighting function. In this size, the block distortion is reduced. For example, paying attention to the hatched first block L 1 (X, y) in FIG. 6, the weight of the block end value is lost by the weighting function. In FIG. 7, the position of the first block L 1 (X, y) is a position indicated by a broken line. For each of the second blocks as well, the weight of the value at the block end is eliminated by the weighting function. Therefore, focusing on one reference block X (p, q), the block distortion at the block end of the reference block X (p, q) is reduced. Then, the motion-compensated predicted image information BXo formed through such weighting processing is transmitted to the subtractor 4. The subtracter 4 calculates the image information X output from the frame memory 3 and the motion-compensated predicted image information (predicted value). The prediction error ε is extracted by subtraction processing with BX o (step S7). When the prediction error ε is extracted by the subtractor 4, the prediction error ε is quantized by the quantization unit 6 and the quantized value q is sent to the signal multiplexing unit 13 and the inverse quantization unit 7 (step S 8) Then, the quantized value q is inversely quantized by the inverse quantizer 7 and added to the motion compensated prediction image BXo by the adder 8, so that the decoded image information ( X and are created (Step S9). This reference frame X, is stored in the frame memory 50 (step S10).

Since the above processing is performed in units of the reference block except for the detection of the motion vector, the address of the reference block is updated after the processing in step S10, and the step S10 is performed on the reference block of the updated address. The processing of 2A, S2B to S10 is performed (step S11). That is, it is determined whether or not p <N-1 is satisfied (step S111). If it is determined in this determination that p <N-1 is satisfied (yes), then After p is incremented (step S114), the above steps S2A and S2B are discriminated again. If it is determined in step S 1 1 1 that p−N−1 does not hold (no), the value of p is set to 0, and then q <M−1 holds. It is determined whether or not this is the case (steps S112, S113). In this determination, if it is determined that q <M—1 holds (yes), q is incremented (step S115), and the determination in steps S2A and S2B is performed again. Is performed. If it is determined that q <M—1 does not hold (no) in the determination in step S.113, the last reference block of the plurality of reference blocks to be encoded is processed. Since the processing of has been completed, the encoding processing for the current frame is terminated. Next, the results of simulation by a computer will be described. The image data used was 160 x 96 pixels, the block size was 16 x 16 pixels, and the search area for motion vector detection was 115 to 15 pixels.

The first motion vector and the second motion vector are detected for the first block and the second block shown in FIGS. 4 (A) and (B), respectively. Based on these motion vectors, motion-compensated predicted images C 1 (X, Y) and C 2 (X, Y) are created. Next, the motion-compensated predicted images CI (X, Y) and C 2 (X, Y) are subjected to weighting processing using trigonometric functions as shown in Figs. 6 and 7, and then synthesized. To obtain the final motion-compensated predicted image C (X, Y). The final motion-compensated predicted image C (X, Y) is represented by the following equation.

C (X, Y) = C 1 (X, Y) xsin ² (XZN) Xsin ² (Υ Ν)

+ C 2 (X, Y) cos 2 (π X / N) x cos 2 (π Y / N)

Prediction is performed using this final motion prediction image C (X, Y), and the obtained prediction error is encoded. It has been confirmed that such a simulation can simultaneously reduce the amount of motion vector information and block distortion.

According to the above embodiment, the following effects can be obtained.

[1] The motion vector detection unit 38, apart from the processing in the motion vector detection unit 37, performs a multi-pixel process so that the first block is divided by half a block in the vertical and horizontal directions of the frame. The block is cut into the second block, and the motion vector is detected for each of the cut blocks. In this way, two systems of block segmentation are performed so as to be shifted from each other by half a block, a motion vector is obtained for each of the segmented blocks, motion-compensated image information is generated, and the motion-compensated image information is synthesized. Predictive coding can be performed with almost the same accuracy as when the block size is reduced by one level. Wear. Furthermore, the amount of motion vector information can be reduced to about 12 in the case where the block division is performed in the one-step smaller size.

[2] The motion compensation prediction image generated by the motion compensation unit 9 based on the detection result of the motion vector detection unit 37 and the motion compensation unit 10 based on the detection result of the motion vector detection unit 38. The weighting process to reduce the weight of the block edge pixels is performed on each of the motion-compensated prediction images generated by the above, and the combining unit 11 that combines them is adopted. Therefore, the objective evaluation of the motion compensated prediction image can be enhanced. As a result, the threshold value of the prediction error can be set high, so that the error signal can be largely eliminated in the moving image communication device.

[3] Due to the above-mentioned effects, a reduction in the amount of vector information and a reduction in block distortion are achieved, so that a good video transmission can be achieved in a TV telephone system.

Next, a second embodiment of the present invention will be described.

FIG. 13 shows another configuration example of the encoding system prediction processing unit 26 shown in FIG. In the figure, the coding prediction processing unit 26 includes three first motion vector detection units, a second motion vector detection unit 38, a comparator 31, a switch 32, a motion compensation unit 33, And a frame memory 50. The first motion vector detection unit 37 compares the input image information X with the previous frame in the frame memory 50 using the first block shown in FIG. It has a function to obtain the motion vector V1 of the block. The second motion vector detection unit 38 compares the input image information X with the previous frame in the frame memory 50 by using the second block shown in FIG. It has the function of obtaining the motion vector V2 of the block. The comparator 31 has a function of comparing the minimum value of the absolute value sum S ij of the difference. That is, the sum of absolute values S of the differences obtained by the motion vector detection unit 37 , _{3 and} the sum of the absolute values of the differences S: j obtained by the bridle vector detector 38 (compared as S 2 ′). U. The information g of this comparison result is multiplexed together with the motion vectors V 1 and V 2 by the signal multiplexing unit 13, and transmitted to the communication line 15 via the transmission unit 14. Further, the operation of the switch 32 is controlled by the comparison result information g of the comparator 31. If the sum of the absolute values of the differences obtained by the motion vector detector 37 is S! If the minimum value S 1 j ′ of j is smaller than the absolute value sum S, j of the differences obtained by the motion vector detection unit 38, the minimum value S 2 of j, the switch 32 The output of the torque detector 37 is selected and transmitted to the subsequent motion compensator 33. On the other hand, the absolute value sum S ij of the difference obtained by the motion vector detector 38 is the minimum value S ij of the difference S ij, and the sum S of the difference obtained by the motion vector detector 37. If it is smaller than the minimum value S 1 J of J, the output of the motion vector detection unit 38 is selected and transmitted to the subsequent stage motion compensation unit 33. That is, the motion vector with the higher accuracy is selected and supplied to the motion compensator 33. The motion compensator 33 has a function of generating a predicted value BXo by moving the position of the previous frame stored in the frame memory 50 based on the motion vector transmitted via the switch 32. Have. The predicted value BXo is transmitted to the switch 25 and the subtractor 4 described with reference to FIG. FIG. 14 shows another configuration example of the decoding-based prediction processing unit 27 shown in FIG. As shown in FIG. 14, the motion vector transmitted from the signal separator 17 and the motion vector transmitted from the signal separator 17 for latching the motion vector V1 are shown. A vector latch 42 for latching the vector V 2 is provided, and the outputs of the vector latches 41 and 42 are selectively transmitted to the motion compensator 44 via a switch 43 at the subsequent stage. It is as follows. The above switch 43 is the ratio transmitted from the signal separation unit 17. The operation is controlled by the comparison result information g. That is, the motion vectors V 1 and V 2 used for the processing in the motion compensator 44 are selected in accordance with the comparison result in the comparator 31 shown in FIG. I have. The motion compensator 44 moves the previous frame from the frame memory 51 using the transmitted motion vector to generate a predicted value BXo. The predicted value BXo can be transmitted to the adder 19 via the switch 22.

FIGS. 15 and 16 show flowcharts of the encoding process in the apparatus of this embodiment. The symbols # 3, # 4, and # 5 shown in Figs. 15 and 16 and the symbols # 3, # 4, and # 5 shown in Fig. 16 are processed by the corresponding symbols. It indicates that they are continuous. The processing shown in FIGS. 15 and 16 is similar to that of the embodiment described above, except that the image size is determined by the horizontal N reference block X the vertical M reference block (16 N pixels X 16 M pixels) Then, it is assumed that scanning is performed for each reference block in the horizontal direction from the upper left of the image to the lower right, and the processing is advanced.

First, assuming that p = q = 0 (step S11), the processing is started from the upper left of the image. The processing of steps S12A to S14A is processing relating to the motion vector detection by the first block, and this processing is performed by the motion vector detection unit 37. The processing of steps S12B to S14B is processing related to the motion vector detection by the second block, and this processing is performed by the motion vector detection unit 38.

Next, it is determined whether or not the reference block to be encoded is a reference block located at the upper left of the first block or the second block (steps S12A and S12B). That is, the motion vector detection unit 37 determines whether or not (P = even number and q = even number) is satisfied for the reference block X (p, q) to be coded. Be (Step S 1 2 A). In addition, the motion vector detection unit 38 calculates (p = 0 and q = 0), or (p = 0 and q = odd) for a reference block to be coded from now on. It is determined whether or not (p = odd number and q = 0), or (p = odd number and q = odd number) is satisfied (step S12B).

If it is determined in the above step S12A that the answer is yes, that is, if it is determined that (p = even number and q = even number) is satisfied, it is coded from now on. The reference block X (p, q) is located at the upper left of the first block. This is clear from FIG. In this case, the motion vector detection unit 37 searches the frame memory 50 to find the motion vector V1 of the first block LI (pZ2, q / 2), and the detection result is stored in the internal memory. And sent to the signal multiplexing unit 13 for signal multiplexing (steps S13A and S14A).

If it is determined yes in the above-described step S12B, that is, (p = 0 and q = 0), (p = 0 and q = odd), (P = odd and , Q = 0) or (p = odd and q = odd), it is determined that the reference block X (p, q) power to be coded from now on It means that it is located at the upper left of the second block. This is clear from FIG. In this case, the motion vector detection unit 38 searches the frame memory 50 to create the motion vector V2 of the second block. At this time, if (p = q = 0) holds, the motion vector V 2 of L 2 (0, 0) is detected, and (p = 0 and q = odd number) holds. , A motion vector V 2 of L 2 (0, q / 2 + 1) is detected, and if (p = odd number and q = 0) holds, then L 2 (p / 2 + 1) , 0) is detected. When (p = odd number and Q = odd number) holds, the motion vector V 2 of L 2 (p 2 +1, q / 2 + 1) is detected. Then, the detection result is latched in an internal register and sent to the signal multiplexing section 13 for signal multiplexing (step S14B).

Thereafter, the motion vector detectors 37 and 38 compare the calculated minimum value S 1 j ′ with S 2 i ′ (step S 15). Also, in the determination of the above steps S12A and S12B, when it is determined to be no, the determination of the above step S15 is also performed. If it is determined in the comparison of step S 15 that S i ＞> S 2 j is satisfied (yes), the comparison result information g output from the comparator 31 is set to level “1” ( Step S16A), the motion vector V1 output from the motion vector detection unit 37 is transmitted to the motion compensation unit 33 via the switch 32, and the motion compensation unit 33 A predicted value BXo is created (step S17A). In the comparison in step S15, if it is determined that j is not satisfied (no) in S1 "'> S2, the ratio 较 result information g output from the comparator 31 is at level" 0 ". (Step S16B), the motion vector V2 output from the motion vector detection unit 38 is transmitted to the motion compensation unit 33 via the switch 32, and the motion compensation unit 33 In step 3, motion-compensated prediction image information (prediction value) BXo is created (step S17B

Then, as in the case of the above-described embodiment, the predicted value BX 0 created in the above step S 178 or S 177 B is transmitted to the subtractor 4, and in the subtracter 4, the frame memory 3 A prediction error ε is extracted by a subtraction process between the image information X output from and the predicted value BXo (step S18). Further, when the prediction error ε is extracted by the subtracter 4, the prediction error ε is quantized by the quantization unit 6, and the quantized value q is converted into the signal multiplexing unit 13 and the inverse quantization unit 7. (Step S19). Then, the quantized value q is inversely quantized by the inverse quantizer 7 and added to the motion-compensated prediction image BXo by the adder 8 to obtain the reference frame X used for the next encoding process. 'Is created (step S20). This reference frame X 'is stored in the frame memory 50 (step S21).

Since the above processing is performed for each reference block except for the detection of the motion vector, the address of the reference block is updated after the processing of step S21, and the above steps are performed for the reference block of the updated address. The processing of S12A, S12B to S21 is performed (step S22). That is, it is determined whether or not p <N—1 is satisfied (step S221). If it is determined in this determination that p <N—1 is satisfied (yes), After p is incremented (step S224), the above-mentioned steps S12A and S12B are discriminated again. If it is determined in step S 2 21 that p <N—1 does not hold (n 0), the value of p is set to 0, and then q <M—1 holds. It is determined whether or not to perform (steps S22, S223). In this discrimination, if it is determined that q−M−1 is satisfied (yes), q is incremented (step S 2 25), and the above steps S 12 A, S 1 2 B is determined. If it is determined that q <M—1 does not hold (no) in the determination in step S22.3, it is determined that the last reference block of the plurality of reference blocks to be encoded is processed. Since the processing for has been completed, the encoding processing for the current frame is terminated.

Thus, each reference block has two types of motion vectors, and when detecting a motion vector, the minimum value of the difference is smaller by comparing the minimum value S 1 j and S 2 j of the difference. Movement vector, that is, the more accurate killing Since the prediction error can be reduced by selecting the vector, the receiving apparatus REC can decode an accurate image. In this embodiment, in addition to the motion vector, it is necessary to generate comparison result information g for each reference block and transmit it, but this comparison result information g is a 1-bit data amount. Therefore, increasing the amount of information is not a problem. Although the invention made by the inventor has been specifically described based on the embodiments, the present invention is not limited thereto, and it is needless to say that various modifications can be made without departing from the gist of the invention. .

For example, in the above embodiment, weighting processing of a motion-compensated prediction image is performed using a trigonometric function. However, in addition to the trigonometric function, a linear function having a triangular shape and a bilinear function having a semicircular shape are used. Various functions such as the following function and even the rectangle function can be used. Further, the size of the reference block is not limited to the size described in the above embodiment, and may be smaller, but the size of the above embodiment is preferable from the viewpoint of efficient data processing. . Industrial applicability

In the above description, the case where the invention made by the inventor is mainly applied to the videophone system which is the background of use has been described. However, the present invention is not limited to this. It can be widely applied to multimedia systems and the like.

Claims

The scope of the claims

1. The input image information is divided into blocks of a plurality of pixels, and the position of a region having image information similar to the block of interest in the current frame is determined from the information of the previous frame, thereby obtaining the information of the previous frame. In a method for detecting a motion vector for predictive coding of input image information based on

One frame is divided into a plurality of first blocks, and a plurality of second blocks arranged so that the first blocks are shifted half a block in the vertical and horizontal directions. A motion vector detection method characterized by detecting the motion vector described above for each of the two blocks.

2. Dividing the input image information into a first block of a plurality of pixels, and determining the position of an area having image information similar to the first block of interest in the current frame from the information of the previous frame, Detecting a first motion vector for predictive encoding of input image information based on frame information, and converting the input image information into a second block in which the first block is shifted half a block vertically and horizontally. By determining the position of an area having image information similar to the second block of interest in the current frame from the information of the previous frame, predictive coding of input image information based on the information of the previous frame is performed. Detecting a second motion vector;

Using the first motion vector of the first block and the image information of the previous frame to create motion-compensated predicted image information related to the first block;

Using the second motion vector of the second block and the image information of the previous frame to create motion-compensated predicted image information for the second block;

The motion-compensated prediction image information related to the first block and the second block A step of performing a weighting process on the motion-compensated predicted image information to reduce the weight of the block edge pixels and combining them with each other, and a step of combining the motion-compensated predicted image information with the motion-compensated predicted image information.

3. The predictive encoding method for image information according to claim 2, wherein the motion compensated prediction image information is created for each reference block having the size of the half block in each of the vertical and horizontal directions.

4. The weighting process is performed on the motion-compensated prediction image information of the reference block relating to one of the first block and the second block, with a period _{πΖ2 with} respect to the length and width of the reference block. Multiplied by the function of sin ² to be performed, and for the motion-compensated predicted image information of the reference block relating to one of the first block and the second block, the vertical and horizontal directions of the reference block are used. 4. The predictive encoding method for image information according to claim 3, wherein the predictive encoding is performed by multiplying by a function of cos ² having a period of πΖ2.

5. A step of obtaining the prediction error information by subtracting the motion-compensated predicted image information obtained in the synthesizing step and the corresponding image information of the current frame is further provided. The predictive coding method for image information according to item 3 of the range.

6. The input image information is divided into a first block of a plurality of pixels, and the position of an area having image information similar to the first block of interest in the current frame is determined from the information of the previous frame. Detecting a first motion vector for predictive encoding of input image information based on frame information, and converting the input image information to a second block arranged by shifting the first block by half a block vertically and horizontally. By determining the position of an area having image information similar to the second block of interest in the current frame from information of the previous frame, predictive coding of input image information based on the information of the previous frame is performed. Detecting a second motion vector; Selecting a motion vector with higher accuracy from the detected first motion vector and second motion vector,

Generating a motion-compensated prediction image information using the selected motion vector. A method for predictive encoding of image information, comprising:

7. The image according to claim 6, wherein the selection of the motion vector and the creation of the motion compensated prediction image information are performed for each reference block having the size of the half block in each of the vertical and horizontal directions. A predictive encoding method for information.

8. The input image information is divided into first blocks of a plurality of pixels, and the position of an area having image information similar to the first block of interest in the current frame is determined from the information of the previous frame. First motion vector detecting means for detecting a first motion vector for predictive encoding of input image information based on frame information,

The input image information is divided into a second block in which the first block is displaced by half a block in the vertical and horizontal directions, and from the information of the previous frame, a region having image information similar to the second block of interest in the current frame is determined. Determining a position to detect a second motion vector for predictive encoding of the input image information based on the information of the previous frame; and a first motion vector detecting means for detecting a first motion vector of the first block. First motion compensating means for creating motion-compensated predicted image information relating to the first block using the motion vector and image information of the previous frame;

A second motion compensation unit that creates motion-compensated predicted image information for the second block using the second motion vector of the second block and the image information of the previous frame;

A first method of creating motion-compensated predicted image information by combining the motion-compensated predicted image information created by the first motion compensating means with the motion-compensated predicted image information created by the second motion compensating means. Combining means; Calculating means for subtracting the synthesized motion-compensated predicted image information and the corresponding image information of the current frame to obtain prediction error information, the moving image processing apparatus comprising:

9. The combining means performs a weighting process on the motion-compensated predicted image information related to the first block and the motion-compensated predicted image information related to the second block to reduce the weight of the block end pixel. 9. The moving image processing apparatus according to claim 8, wherein the moving image processing apparatus synthesizes them.

10. The motion compensating means and the synthesizing means perform information creation in units of reference blocks each having the size of the half block vertically and horizontally,

Further, the synthesizing means includes a period π Ζ 2 for the motion compensated prediction image information of the reference block according to one of the first block and the second block with respect to the length and width of the reference block. The weighting process is performed by multiplying the function of sin ² to be performed, and the motion compensation prediction image information of the reference block related to one of the first block and the second block is used as the reference block. 10. The moving image processing apparatus according to claim 9, wherein the weighting process is performed by multiplying the height and width by a function of cos ² having a period of πΖ2.

1. A first memory for holding the image information of the current frame, and a first memory for adding the synthesized dead compensation predicted image information to the prediction error information generated by the subtraction unit. Addition means;

9. The image processing apparatus according to claim 8, further comprising: a second storage unit for holding the image information obtained by said first adding unit as image information of a previous frame. A moving image processing apparatus according to claim 1.

12. The moving picture processing device according to claim 11, further comprising means for transmitting the prediction error information and first and second motion vectors.

13. A means coupled to the transmitting means for receiving the prediction error information and the first and second motion vectors.

Third motion compensation means for creating motion-compensated predicted image information using the received first motion vector and image information of the previous frame on the receiving side;

Fourth motion compensation means for creating motion-compensated predicted image information using the received second motion vector and the image information of the previous frame on the receiving side;

A second step of creating motion-compensated predicted image information by combining the motion-compensated predicted image information created by the third motion compensating unit with the motion-compensated predicted image information created by the fourth motion compensating unit; Combining means;

Second adding means for adding the combined motion-compensated predicted image information and the received prediction error information to obtain decoded image information;

A third storage unit for holding the image information obtained by the second addition unit as image information of the previous frame on the receiving side, and a third storage unit. 13. The moving image processing device according to item 12.

14. The input image information is divided into first blocks of a plurality of pixels, and from the information of the previous frame, the position of a region having image information similar to the first block of interest in the current frame is determined. First motion vector detecting means for detecting a first motion vector for predictive encoding of input image information based on the information of the previous frame,

The input image information is divided into a second block in which the first block is shifted by half a block in the vertical and horizontal directions. A second motion vector detecting means for detecting a position to detect a second motion vector for predictive encoding of input image information based on the information of the previous frame; and the detected first motion. A first selecting means for selecting a motion vector having a higher accuracy from the vector and the second motion vector; First motion compensation means for creating motion-compensated predicted image information related to the selected motion vector using the selected motion vector and image information of the previous frame;

Subtraction means for subtracting the motion-compensated predicted image information created by the first motion compensation means from the corresponding image information of the current frame to obtain prediction error information. Moving image processing apparatus.

15. A first storage unit for holding the image information of the current frame, and the prediction error information generated by the subtraction unit, the motion compensated prediction image information generated by the first motion compensation unit. First adding means for adding

A second storage unit for holding the image information obtained by the adding unit as image information of a previous frame;

Means for transmitting the prediction error information, a first helper vector, a second movement vector, and control information for designating the movement vector selected by the first selection means;

Receiving means coupled to the transmitting means;

Second selecting means for selecting a motion vector specified by the control information from the received first motion vector and second motion vector,

Second motion compensating means for creating motion-compensated predicted image information using the motion vector selected by the second selecting means and image information of the previous frame on the receiving side;

A second adding unit that adds the motion-compensated predicted image information formed by the second motion compensating unit and the received prediction error information to obtain decoded image information; 15. The moving image processing according to claim 14, further comprising: a third storage unit for storing image information as image information of a previous frame on the receiving side. apparatus.