JPH07264599A - Encoding device and decoding device for dynamic image - Google Patents

Abstract

PURPOSE:To execute accurate information transmission for a restricted code amount and to efficiently encode/decode a motion vector by outputting the encoded signal of each area from an encoding means together with importance and a motion vector. CONSTITUTION:A moving vector of each area divided by an area dividing circuit 100 is stored in a motion vector memory 102 for a fixed period. An importance judging circuit 104 judges the importance of motion detected in each area or that of the area while referring to past motion information stored in the memory 102. Then an encoding circuit 105 executes encoding in accordance with the judged importance. In the case of low importance, quantizing step size is increased to execute rough quantization and a code amount is reduced though an quantizing error is large. When the importance is high, the quantizing step size is reduced to execute accurate quantization and a minute image having a large code amount and reduced at its quantizing error can be regenerated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to communication using moving images,
The present invention relates to a moving picture coding device and a decoding device in an information terminal for broadcasting, storage and the like.

[0002]

2. Description of the Related Art In order to efficiently compress and code moving image data having a large amount of information, a method of efficiently reducing redundancy by utilizing correlation in the time axis direction or the space axis direction is used. There is. Among them, the methods used as the redundancy reduction method in the time axis direction include “motion compensation” and “difference between frames”. Motion compensation uses the correlation of the temporal change of the moving object and The inter-difference uses the time invariance of the stationary part. In particular, the motion compensation measures the amount of movement of the moving object (the amount of movement of the object from the previous frame to the position of the current frame) between the current frame and the previous frame, and measures the direction and size. After generating the motion vector indicating the height, the element of the motion vector is encoded and transmitted to the receiving side.

If the transmission code amount is limited, it is necessary to control the generated information amount by using some method. These control methods include: Increase the quantization step size in the compressed part.

2. Reduce the number of frames to be transmitted. 3. The encoding target area is limited to a part of the frame. 4. The size of the motion vector is determined, and only the area having a large motion is encoded. However, each of them has the following problems.

[0005] 1 is a blurred image with the entire resolution lowered. 2 becomes awkward. Regardless of whether or not 3 is truly necessary information, the uncoded part remains unnatural in the reproduced image.

[0006] In the case of 4, the small movement part is not transmitted, so that if the necessary information is included in it, accurate information cannot be transmitted to the receiving side. In particular, the method of determining important information from the motion vector was performed by using the size of the detected motion vector as described in 4 above, but the size of the motion does not always match the information that the user really needs. The current situation is that there are none.

For example, a portion that always passes through the screen in a certain direction is often the background even if the movement is large. In other words, when performing video communication in a car or train, the scenery outside the window passes by at the speed of the car or train, but the importance of information is often different from the information that you want to tell the other party. . On the other hand, even if the movement is not great,
The part that repeats up and down, left and right within a certain range is the information that should be sent to the other party or the part that the user wants to pay attention to, and is therefore highly important as information.

Therefore, if the importance of information is judged only by the magnitude of the movement as in the conventional case, the information which the sender originally wants to convey to the receiver is not properly judged, and the degree of compression of each information is also increased. Since it becomes inaccurate, there is a drawback that the demand for transmitting information as accurately as possible within a limited amount of transmission code is not sufficiently satisfied.

Further, conventionally, a motion prediction device has been used in a moving image coding device or the like. FIG. 10 shows an example of a moving picture coding device. The current coded frame 401 is input to the motion prediction device 402 and the subtractor 403, and motion prediction processing is performed using the past frame 404 as a reference frame.

The contents of the motion estimation device 402 are shown in FIG. The current frame 401 is input to the subtractor 405, and the past frame 404 is already held in the frame memory 406. The motion vector 408 is sent from the controller 407 to the frame memory 406, and the data 409 of the past frame which is deviated from the current frame 401 by the amount of the motion vector 408 is read from the frame memory 406 and input to the subtractor 405. It The error 410 between the current frame 401 and the past frame 409 is sent to the controller 407. The controller 407 switches the motion vector 408 to several types and sends the motion vector 411 in which the error 410 is reduced to the motion vector encoder 412. Motion vector 4 at the same time
The past frame 409 corresponding to 11 is output. Note that the motion prediction may be performed in frame units or in block units obtained by dividing a frame.

In the motion vector encoder 412, the method of assigning the code 413 to the value of the motion vector as it is is the simplest, but as another method, the motion vector of the frame temporally adjacent to the current frame and the current frame. There is also used a method of obtaining the difference between the motion vectors of the above and adding the reference numeral 413 to the difference value. Further, when performing motion prediction in block units, motion vectors of spatially adjacent blocks may be referenced. Here, the motion vector to be referred to is called a prediction vector, and the error between the transmission vector and the prediction vector is called a prediction error. When the temporal and spatial variation of the motion vector is small, that is, when the motion of the subject or the motion of the camera is small, the appearance frequency near the prediction error of zero is shown in FIG.
It becomes higher as shown in FIG. Therefore, if a method called variable length coding, which assigns a code closer to zero prediction error than in other cases, is used, the total code amount can be reduced as compared with the case where a code is added to the value of the motion vector as it is.

Returning to FIG. 10, a motion vector code 413.
Is output as it is to the outside. The past frame 409 is input to the subtractor 403 and the adder 414. Subtractor 40
From No. 3, the error 415 between the current frame 401 and the past frame 409 is sent to the pixel prediction error encoder 416. The pixel prediction error encoder 416 encodes the error 415 by, for example, orthogonal transform encoding, and the pixel prediction error code 4
17 is output to the outside and is also input to the pixel prediction error decoder 418. The pixel prediction error decoder 418 decodes the pixel prediction error 419 and inputs it to the adder 414. The adder 4145 adds the pixel prediction error 419 to the past frame 409, reproduces the image, and sends the image to the motion prediction apparatus in preparation for encoding the next frame.

FIG. 12 shows a moving picture decoder corresponding to this moving picture encoder. Sign 417 of pixel prediction error
Is the pixel prediction error decoder 424.
25 is decoded and sent to the adder 426. On the other hand, the motion vector code 413 is decoded into the motion vector 429 by the motion vector decoder 428, and the frame memory 427
Sent to. The frame memory 427 already holds the past frame, the past frame 530 designated by the motion vector 429 is read out, and the adder 4
Sent to 26. In the adder 426, the pixel prediction error 425 is added to the past frame, which is output to the outside as the reproduced image 431 and written in the frame memory 427.

As described above, the motion prediction measuring device is often used as a part of the moving image coding device. In that case, from the viewpoint of reducing the image communication cost and saving the image storage medium, it is preferable that the code amount of the motion vector is small. Therefore, the device for encoding the difference between adjacent vectors as described above in the description of the motion vector encoder 412 in FIG. 11 has been devised. However, in this method, when the motion of the subject or the camera is strong, the bias of the prediction error of the adjacent vector to zero becomes smaller as compared with FIG. 13, and the code amount of the motion vector increases. .

[0015]

As described above, in the prior art, accurate information transmission is not performed within the limited transmission code amount, and an efficient encoding method of motion vectors is considered. There was a drawback that it was not done.

Therefore, an object of the present invention is to provide a moving picture coding apparatus and a decoding apparatus capable of accurately transmitting information for a limited transmission code amount and efficiently coding / decoding motion vectors. To do.

[0017]

According to a first aspect of the present invention, there is provided a moving picture coding apparatus, wherein the moving picture coding apparatus divides an image signal inputted for each frame into a plurality of areas, and the dividing means. Motion compensation means for detecting the motion vector of the previous frame image signal of the input frame for each area, and a motion vector memory for accumulating and storing the motion vector of each area detected by the motion compensation means for each input frame. Subtraction means for obtaining a difference between each of the divided areas and the area of the previous frame image signal corresponding to the motion vector detected by the motion compensation means,
Importance determination means for determining the importance of each area according to the motion vector from the input frame of each area accumulated and stored in the motion vector memory to the past frame of a certain period, and each area from this importance determination means Coding means for coding the differential signal of each area from the subtracting means according to the importance of, and the area of the previous frame image signal by decoding the coded signal of each area from the coding means according to the importance. Decoding means for adding to the motion compensating means as a previous frame image signal of the frame image signal to be input next, and output the encoded signal of each area from the encoding means together with the importance and the motion vector. And means for doing so.

In the moving picture decoding apparatus of the first invention, the input means for inputting the encoded signal of each area of the image signal for each frame together with the importance and the motion vector, and the area of each area from this input means Decoding means for decoding the coded signal according to the degree of importance, frame memory storing the previous frame image signal of the frame image signal input to the input means, and the motion vector of each area from the input means Extracting means for extracting the corresponding area from the previous frame image signal in the frame memory, and adding means for adding the decoded signal of each area from the decoding means and the corresponding area from the extracting means. , A means for storing the addition result from the adding means in the frame memory as a previous frame image signal for the next input frame image signal. Characterized in that was.

The first moving picture coding apparatus of the second invention is
A motion prediction unit that performs motion prediction on the image signal of the input frame with the image signal of the previous frame and outputs the motion vector information, and the image signal of the previous frame motion-predicted by the motion prediction unit and the input frame. Subtracting means for obtaining the difference from the image signal, coding means for coding and outputting the difference signal from this subtracting means, and the motion-predicted previous frame image for decoding the coded signal from this coding means A moving picture decoding apparatus comprising: a decoding means for adding the signal to the motion predicting means as an image signal of a previous frame for the next input frame, wherein the motion predicting means is an image signal of the input frame. Control means for detecting the motion vector of the image signal of the previous frame, and the motion vector detected by this control means for each input frame A motion vector memory for product storage, a cycle detecting means for detecting a time change cycle of the motion vector stored and stored in the motion vector memory, and a cycle detected by the cycle detecting means for the motion vector of the input frame. Only the past motion vector is read from the motion vector memory, and the motion vector read by this means is used as a predicted value of the motion vector of the input frame.

The first moving picture decoding apparatus of the second invention is
Input means for inputting a coded signal and motion vector information for the frame image signal, decoding means for decoding the coded signal from the input means, and motion by decoding the motion vector information from the input means A motion vector decoding means for outputting a vector, a frame memory storing a previous frame image signal for the frame image signal input from the input means, and a front memory in the frame memory according to a motion vector from the motion vector decoding means A unit for predicting the motion of a frame image signal, the previous frame image signal whose motion is predicted by this unit, and the decoded signal from the decoding unit are added, and the frame memory is used as the previous frame image signal for the next frame image signal. And a motion vector from the motion vector decoding means. A motion vector memory for accumulating and storing for each frame, a cycle detecting means for detecting a cycle of time change of the motion vector accumulated and stored in the vector memory, and a cycle detecting means for detecting the inputted motion vector information. A means for reading a motion vector past by a certain period from the motion vector memory, and a means for supplying the motion vector read by the means to the motion vector decoding means as a predicted value of the input motion vector information. Is provided.

A second moving picture coding apparatus according to the second invention is
A motion prediction unit that performs motion prediction on the image signal of the input frame with the image signal of the previous frame and outputs the motion vector information, and the image signal of the previous frame motion-predicted by the motion prediction unit and the input frame. Subtracting means for obtaining the difference from the image signal, coding means for coding and outputting the difference signal from this subtracting means, and the motion-predicted previous frame image for decoding the coded signal from this coding means A moving picture coding apparatus comprising: a decoding means for adding the signal to the motion prediction means as an image signal of the previous frame for the next input frame, wherein the motion prediction means is an image signal of the input frame. Control means for detecting the motion vector of the image signal of the previous frame, and the motion vector detected by this control means for each input frame A motion vector memory for product storage and a change pattern for the past N sections from the input frame and another past section that is a change pattern within a predetermined error range for the motion vector stored and stored in the motion vector memory are stored. Pattern detecting means for detecting, means for reading the motion vector after another section detected by the pattern detecting means from the motion vector memory, and prediction of the motion vector of the input frame for the motion vector read by this means And a means used as a value.

A second moving picture decoding apparatus according to the second invention is
Input means for inputting a coded signal and motion vector information for the frame image signal, decoding means for decoding the coded signal from the input means, and motion by decoding the motion vector information from the input means A motion vector decoding means for outputting a vector, a frame memory for storing a previous frame image signal for the frame image signal input from the input means, and a previous frame in the frame memory according to a motion vector from the motion vector decoding means A means for predicting the motion of an image signal, and a previous frame image signal whose motion is predicted by this means and the decoded signal from the decoding means are added to the frame memory as a previous frame image signal for the next frame image signal. The means for storing and the motion vector from the motion vector decoding means are stored. A motion vector memory that stores and stores each of the frames, and a change pattern in the past N sections from the input frame and a change pattern that is a change pattern within a predetermined error range with respect to the motion vector stored and stored in the vector memory. Pattern detecting means for detecting another section, means for reading a motion vector after the different section detected by the pattern detecting means from the motion vector memory, and a motion vector read by this means for the input motion Means for supplying to the motion vector decoding means as a predictive value of vector information.

[0023]

According to the first aspect of the present invention, a memory for storing motion information from a certain past to the present is stored, and by utilizing the information on the change in motion from the past stored in the memory, the current frame In addition to the motion information (for example, the size of the motion vector), the importance of the information is determined in consideration of the "way of motion" of the part from the past. Basically, the part where the same motion vector is always observed on the screen (moving region part that always passes in a certain direction) is often the background even if the motion is large, so the importance is set low. To do. On the other hand, the part where different motion vectors are observed depending on time (moving region part where a certain range is repeated up, down, left, and right) is the information to be sent to the other party even if the motion is not large ( Since the sender is pointing the camera in that direction), set the importance to high. In other words, the introduction of the method that considers this way of movement enables appropriate determination of the important information for the user. Further, the compression rate of each information (that is, how finely to encode) is adaptively set according to the importance, and information is transmitted as accurately as possible within a limited transmission code amount.

Even in the part where the same motion vector is always observed on the screen, if the motion is small, human beings can sufficiently follow the motion, so that the sender may possibly want to transmit the information. There is (for example, when moving the camera slowly to inform the surroundings of the current location). Therefore, when the movement is small, it is possible to avoid the erroneous determination of the importance by not performing the importance determination in consideration of the above-described movement.

As a result, according to the first aspect of the present invention, information that is truly important to the user is appropriately determined, and the compression rate of each information can be adaptively set according to the degree of importance. It becomes possible to transmit the information originally intended to be transmitted as accurately as possible within the limited transmission code amount.

In the second aspect of the present invention, the storage means for holding a plurality of past motion vectors, the means for detecting the cycle of time change of these motion vectors, and the motion vector traced back in the past by the detected cycle are described above. In the configuration having a unit for reading from the storage unit and a unit for outputting the read motion vector as a current motion prediction vector, or a storage unit for holding N or more past motion vectors, and the storage unit. The latest change pattern of N motion vectors is used as target data, a section having the same change pattern as the target data is found as a reference section from the storage means, and a motion vector following the reference section is stored in the storage means. Means for reading from
When the motion has periodicity, the read motion vector is output as a motion prediction vector of the current processing frame, and when the motion has periodicity, the information amount of the motion vector is reduced by utilizing the property.

Therefore, according to the second aspect of the present invention, it is possible to detect the period of time change while reading the past motion vector from the storage means. For example, when a videophone is used while riding in a vehicle such as a train or an automobile, the motion vector corresponding to the shake of the camera changes almost periodically, so that a motion vector as shown in FIG. 7 is detected, for example. In this figure, the horizontal axis represents time, that is, the frame number, and the current time is tn. Further, the vertical axis represents the x component of the motion vector, and almost periodic fluctuations are seen as shown in the figure. The same applies to the y component of the motion vector. If the cycle T is detected, the motion vector traced back by T in the past is read from the storage unit and output as a prediction vector. A motion vector traced back by this cycle T has a higher probability of being more accurate as a prediction vector than an adjacent vector. Therefore, when variable length coding of a motion vector is performed using the prediction vector according to the present invention, the prediction error is concentrated near zero and the code amount can be reduced. In addition, the camera shake can be corrected by shifting the input image in the direction that cancels the prediction vector, and a visually good image without shaking on the screen can be input to the moving image encoding apparatus.

In determining the prediction vector, first,
A section having the same pattern as the latest N motion vector change patterns is found and used as a reference section (see FIG. 7), and a vector following this reference section is used as a prediction vector for variable-length coding of motion vectors and shake correction. If so, the effect of reducing the code amount can be obtained.

[0029]

【Example】

(First Invention) FIG. 1 is a block diagram of a transmitting side (encoding device) according to an embodiment of the first invention. Input image information is
The area dividing circuit 100 divides each area into a predetermined area and then transfers the area to the motion compensation circuit 101. The motion compensation circuit 101 compares the previous frame information called from the frame memory 108 in which the previous frame for which the encoding / decoding processing has already been completed is stored with the current frame information divided for each area, From where in the previous frame the region of the current frame has moved, that is, how much motion there was in the region up to the current frame is calculated and expressed as a motion vector (this series of operations is called motion compensation). Called). In the difference circuit 103, the area information of the motion-compensated previous frame is subtracted from the area information of the current frame, and the difference information (that is, the information of the current frame that cannot be expressed only by considering the motion information in the previous frame). ) Is the encoding circuit 1
Coded at 05. For each area, the coded information, the motion vector information, and the importance degree information are multiplexed by the multiplexing circuit 109, sent to the transmission path, and sent to the receiving side.
On the other hand, the encoded information for each area is the encoding circuit 1
It is decoded as a motion compensation error at 06, and the previous frame information after motion compensation and the motion compensation error are added at the addition circuit 107, the current frame information is reproduced, and the frame memory 10 is reproduced.
8 is stored.

In contrast to such a conventional motion compensation error coding system, the present invention newly adds a motion vector memory 102 for storing the motion vector obtained for each area for a certain period. For each area in the screen where there is motion, the motion information from the past point in time, which traces back a certain period from the present, to the present is stored, and the motion information from the past stored in this motion vector memory is stored. By referring to the motion information, the motion detected by the importance level determination circuit 104 for each area or the importance level of the area is determined.
This importance information is sent to the encoding circuit 105 at the next stage, where it is encoded according to the importance.

That is, when the degree of importance is low, the quantization step size is increased to coarsely quantize, and the code amount is reduced instead of the large quantization error. It is also conceivable that the encoding process of that portion is stopped and no information is transmitted to the receiving side, or alternative information (a plain background or the like) is transmitted. In addition, a low-pass filter is applied to that part, and a method that removes high-frequency components before encoding, or when performing orthogonal transform such as discrete cosine transform here, only low-frequency components are subject to encoding, A method of discarding high frequency components may be used. On the other hand, when the degree of importance is high, the quantization step size is reduced and fine quantization is performed to reproduce a fine image with a small quantization error in spite of a large code amount. Further, it is possible to switch to accurate transmission of information by stopping the low-pass filter processing and the high-frequency component removal of the orthogonal transformation which were performed when the degree of importance is low.

Next, a method of determining the degree of importance will be described. The part that constantly passes through the screen in a certain direction is often the background even if the movement is large. In other words, when moving image communication is performed in a car or train, the scenery outside the window passes by at the speed of the car or train, but the importance of the information is often different from the information originally intended to be transmitted to the other party.
Therefore, in this case, it can be considered that the importance is low. On the other hand, even if the movement is not large, the sender repeats the camera up and down, left and right within a certain range, so the sender points the camera in that direction. This is the part that I want you to do and it is highly important as information. In this way, the importance level is set low for parts that generate the same motion vector (parts that always pass in a certain direction), and parts that repeat within a certain range vertically and horizontally. The importance is set high for.

An example is shown in FIG. Now at time t-n,
It is assumed that the motion vector of each area in the frame 201 is obtained as in (a) (arrow in the figure). It is assumed that it has changed as shown in (b) by time t, and the motion vector during that time has been stored in the motion vector memory. Area 202 is, for example, a human, and area 203 is, for example, the background outside the window. It is assumed that a person is moving on a certain range while a person is on a running train, while the background outside the window is passing by at the speed of the train.

A large motion vector is generated in the area 203, but this is due to the movement of the train, and always the same vector is generated. The importance of this portion is set to be low, and the quantization step size in the encoding unit is increased to perform coarse quantization. Alternatively, the high frequency component after the orthogonal transformation is discarded and displayed as an image of only the low frequency component. On the other hand, in the region 202, different motion vectors are detected depending on the time, and the size thereof is not so large as that of the region 203, but the motion always stays within a certain range, and it is considered that the camera is aimed at that part. , The quantization step size in the encoding unit is reduced to perform fine quantization. Alternatively, even the high-frequency component after the orthogonal transformation is displayed as an image including the high-frequency component as an encoding target.

On the other hand, consider an example as shown in FIG. Now time t
It is assumed that the motion vector of each area in the frame 301 is obtained in -n as shown in (a) (arrow in the figure). At time t, a similar vector is obtained as shown in (b),
It is assumed that the motion vector during that time is stored in the motion vector memory. A region 302 is, for example, a landscape that the sender wants to send to the receiving side, and is a case where the surrounding landscape is sequentially transmitted by slowly moving the camera. At this time, the same motion vector is always generated, and if applied in the case of FIG. 2, the importance is set low. However, in this case, the amount of movement is small, and since the human eye can also follow this movement (because the sender moves the camera within the range of movement that the opponent's eyes can follow), the same vector is always generated. Also, when the amount of movement is smaller than the preset threshold value, the importance of that portion is set to be high.

A method for determining whether or not the motion vector for each area changes with time, that is, a criterion for determining that the same motion vector has occurred will be described. Motion vectors detected at the same position in frames that are temporally adjacent to each other (for example, motion vectors detected at the same position in the current frame and the previous frame, respectively. Time is called unit time), and the inner product and the size are calculated. First, the inner product value is a preset value (threshold value S)
If larger, the two vectors are determined to have the same direction and direction (in the case of exactly the same, the inner product value becomes 1). Further, when the difference in magnitude between the two vectors is smaller than a preset value (threshold value T), it is determined that the two vectors have the same magnitude (if they are exactly the same, the difference is 0). When the above two conditions (inner product value> S and size difference <T) are satisfied, it is determined that the vector of the area has not changed within the unit time. Furthermore, when all the motion vectors per unit time from the past stored in the motion vector memory have not changed, or when the change is less than the preset ratio, the motion vector of the area is Judge that it is always constant.

FIG. 4 shows the receiving side (decoding device) according to the present invention.
It is a block diagram of. The coding information, the motion vector, and the importance information transmitted from the coding apparatus shown in FIG. 1 are separated by the separation circuit 110, the coding information and the importance information are sent to the decoding circuit 11, and the motion vector is stored in the frame memory. 112
Sent to. The decoding circuit 111 decodes the coded information with a compression degree corresponding to the importance degree information. Frame memory 1
12 stores the decoded image of the previous frame, from which the area corresponding to the motion vector is read out and the adder 11
3 is output. The adder 113 adds the decoded difference information from the decoding circuit 111 and the area read from the frame memory 112. The added decoded image is stored in the frame memory 112 and is used as the image information of the previous frame when the coding information of the next frame is decoded.

The decoded image in the frame memory 112 is displayed and output from the monitor 114. (Second Invention) The first embodiment of the second invention will be described.

FIG. 5 is a block diagram of a motion predicting section for detecting a motion vector cycle in the encoding apparatus of the first embodiment. The current encoded frame 401 is input to the subtractor 405, and the previous frame 404 is already held in the frame memory 406. Motion vector 40 from controller 407
8 is sent to the frame memory 406, and the current frame 401
The data 409 of the previous frame deviated by the amount of the motion vector 408 is read from the frame memory 406 and input to the subtractor 405. The error 410 between the current frame 401 and the past frame 409 is sent to the controller 407.
The controller 407 switches the motion vector 408 to several types and sends the motion vector 411 with which the error 410 is reduced to the motion vector encoder 412 and the motion vector memory 420. At the same time, the previous frame 409 corresponding to the motion vector 411 is output.

The motion vector memory 420 stores and holds past motion vectors, and the cycle detector 421 detects the cycle of the change over time. The cycle detector 421 separately processes the x and y components of the motion vector. First,
The x-component data is transformed into the frequency space using Fourier transform or the like. Then, the reciprocal of the peaking frequency is taken as the cycle of the x component. For the y component, the cycle of the y component is similarly obtained. From the motion vector memory 420, the x component of the motion vector, which is traced back in time by the period, is read out as the predicted value of the x component. The same applies to the y component. These x and y components are combined and sent to the motion vector encoder 412 as a prediction vector 422.

The motion vector encoder 412 performs variable length coding on the difference value between the motion vector 411 and the prediction vector 422, and outputs the code 413 to the outside. Further, the prediction vector 422 is also sent to the controller 407, and the motion vector 408 close to the prediction vector 422 is preferentially generated from the controller 407, and when the error 410 becomes smaller than a preset allowable error. By ending the motion vector search, it is possible to perform the motion vector search at high speed while maintaining the efficiency of the variable length coding of the motion vector at a good level.

If the motion vector 408 is set so that the vicinity of the prediction vector 422 becomes dense,
Even if the number of candidates for the motion vector 411 is the same, the prediction error can be reduced, and the efficiency of variable length coding can be improved.

Next, a second embodiment of the second invention will be described. FIG. 6 is a block diagram of a motion prediction unit that detects a change pattern of a motion vector in the encoding device according to the second embodiment. The difference from FIG. 5 is that the period detector 42
The point is that the pattern detector 423 is used instead of 1.
The pattern detector 423 first detects the motion vector memory 420.
The x components of the latest N motion vectors are read out and held as target data for pattern detection.
Next, the x component of the motion vector is read from the section consisting of N motion vectors whose time goes back in the past, and the error from the target data is calculated. In this way, the error is calculated while switching the sections variously, and the section in which the error becomes small is determined as the reference section having the same change pattern as the target data. Then, the x component of the next motion vector in the reference section is used as the predicted value. Similarly for the y component,
The x and y components are combined and sent to the motion vector encoder 412 as a prediction vector 422. Since the other parts are the same as those in FIG. 5, description thereof will be omitted.

Next, the coding devices in the first and second embodiments will be described together. FIG. 8 is a block diagram of this encoding device. The periphery of the motion vector decoder 428 is different from that of the decoding device described in the related art. The motion vector 429 is sent to the frame memory 427 and written in the motion vector memory 432. The cycle detector 433 detects the cycle of the motion vector held in the motion vector memory 432, and the motion vector traced back by the detected cycle is the predicted vector 43.
4 from the motion vector memory 432 to the motion vector decoder 428. Motion vector decoder 428
Then, the code 413 of the motion vector is decoded to obtain the prediction error, and the prediction vector 434 is added to the prediction error to obtain the motion vector 429. Similar to the first and second embodiments of the moving picture encoder, the period detector 433 may be replaced with a pattern detector.

FIG. 9 shows another embodiment in which the present invention is used for screen shake correction. The motion vector 435 is held in the motion vector memory 436, and the prediction vector 437 is obtained in the same manner as in the above-described embodiment. The prediction vector 437 is input to the shake compensator 438. The current frame 401 is also input to the shake compensator 438, and the current frame 401 is shifted in the direction in which the motion is suppressed by the amount of the predicted vector. After that, the motion prediction is performed using the current frame 439 in which the motion is suppressed. Further, when incorporated in a moving picture coding apparatus, the current frame 439 of which motion is suppressed is also output to the outside of the motion predicting apparatus. Then, instead of the current frame 401, the current frame 439 in which the motion is suppressed is input to the differencer 403 of FIG.

[0046]

According to the present invention, the information that is truly important to the user can be properly determined, and the desired information can be accurately transmitted within the limited transmission code amount. Furthermore, by utilizing the periodicity of the motion vector, it is possible to code the motion vector with a small code amount even when the motion is intense.

[Brief description of drawings]

FIG. 1 is a block diagram of an encoding device according to a first invention.

FIG. 2 is a diagram showing a first example of a temporal change of a motion vector.

FIG. 3 is a diagram showing a second example of a temporal change of a motion vector.

FIG. 4 is a block diagram of an encoding device according to the first invention.

FIG. 5 is a block diagram of an encoding device of a first embodiment according to the second invention.

FIG. 6 is a block diagram of an encoding device of a second embodiment according to the second invention.

FIG. 7 is a diagram showing the periodicity of time change of a motion vector.

FIG. 8 is a block diagram of a decoding device according to first and second embodiments of the second invention.

FIG. 9 is a block diagram of another embodiment according to the second invention.

FIG. 10 is a block diagram of a conventional moving image encoding device.

FIG. 11 is a block diagram of a conventional motion prediction unit.

FIG. 12 is a block diagram of a conventional moving picture decoding apparatus.

FIG. 13 is a first diagram showing a distribution of prediction errors.

FIG. 14 is a second diagram showing the distribution of prediction errors.

[Explanation of symbols]

 100 ... Region division unit 101 ... Motion compensation unit 102, 420, 432 ... Motion vector memory 103, 403, 405 ... Subtraction circuit 104 ... Importance determination circuit 105, 416 ... Encoding circuit 106, 111, 418, 424 ... Decoding Circuits 107, 113, 414, 426 ... Addition circuit 108, 112, 406, 427 ... Frame memory 109 ... Multiplexing circuit 110 ... Separation circuit 114 ... Monitor 401 ... Current frame 402 ... Motion prediction device 404 ... Previous frame 407 ... Controller 408, 411, 429 ... Motion vector 409, 430 ... Motion predicted previous frame 401, 415 ... Difference signal 412 ... Motion vector encoder 413 ... Motion vector code 417 ... Encoding information 419, 425 ... Decoding information 421, 433 ... Cycle detector 422, 434 ... Prediction vector 423 ... pattern detector 428 ... motion vector decoder 431 ... decoded image information

Claims (8)

[Claims] 1. An area dividing means for dividing an image signal input for each frame into a plurality of areas, and a motion vector for each area divided by the area dividing means and a preceding frame image signal of the input frame. , A motion vector memory for accumulating and storing the motion vector of each area detected by the motion compensating means for each input frame, and the motion compensating means for each of the divided areas. Subtraction means for obtaining the difference between the detected motion vector and the area of the previous frame image signal, and the motion vector from the input frame of each area accumulated and stored in the motion vector memory to the past frame of a certain period According to the importance determining means for determining the importance of each area, and the subtraction procedure according to the importance of each area from the importance determining means. From the encoding means for encoding the differential signal of each area, and the encoded signal of each area from this encoding means is decoded according to the degree of importance, added to the area of the previous frame image signal, and input next. A decoding means for supplying to the motion compensating means as a previous frame image signal of the frame image signal, and a means for outputting the encoded signal of each area from the encoding means together with the importance and the motion vector. A featured moving image encoding device. 2. The importance determining means is configured to, if the respective errors of the motion vectors from the input frame accumulated and stored in the motion vector memory to a past frame of a certain period are within a range of a predetermined value. Is determined to be low, and when the respective errors are out of the range of a predetermined value, the degree of importance of the area is determined to be high. 2. The moving picture coding apparatus according to claim 1, wherein the moving picture coding apparatus encodes a high degree of importance and a high degree of compression with respect to a highly important area. 3. The importance determining means does not determine the importance of the area as low when the respective errors are within a predetermined value range and the magnitudes of the respective motion vectors are smaller than a threshold value. Item 2. The moving picture coding device according to item 2. 4. Input means for inputting the coded signal of each area of the image signal together with the importance and the motion vector for each frame, and decoding the coded signal of each area from this input means according to the importance. Decoding means, a frame memory storing the previous frame image signal of the frame image signal input to the input means, and the previous frame image signal in the frame memory according to the motion vector of each area from the input means The extraction means for extracting the corresponding area, the addition means for adding the decoded signal of each area from the decoding means and the corresponding area from the extraction means, and the addition result from this addition means And a means for storing in the frame memory as a previous frame image signal for the frame image signal input to Apparatus. 5. A motion predicting means for predicting a motion of an image signal of an input frame with an image signal of a previous frame and outputting motion vector information thereof, and an image signal of a previous frame motion-predicted by the motion predicting means. And subtraction means for obtaining the difference between the image signal of the input frame, encoding means for encoding and outputting the difference signal from this subtraction means, and decoding the encoded signal from this encoding means to perform the motion prediction. A moving image decoding apparatus comprising: a decoding unit that adds to the image signal of the previous frame and supplies the image signal of the previous frame to the next input frame to the motion prediction unit, wherein the motion prediction unit is The control means for detecting the motion vector of the image signal of the previous frame with respect to the image signal of the input frame, and the motion vector detected by this control means are input to the input frame. A motion vector memory for storing and storing for each frame, a cycle detecting means for detecting a cycle of time change of the motion vector stored and stored in the motion vector memory, and a motion vector for the motion vector of the input frame detected by the cycle detecting means. A moving image including means for reading a motion vector past by a certain period from the motion vector memory, and means for using the motion vector read by this means as a predicted value of the motion vector of the input frame. Image coding device. 6. Input means for inputting a coded signal and motion vector information for a frame image signal, decoding means for decoding a coded signal from this input means, and motion vector information from said input means. And a motion vector decoding means for decoding a motion vector to output a motion vector, a frame memory storing a previous frame image signal for the frame image signal input from the input means, and a motion vector from the motion vector decoding means in accordance with the motion vector. A means for predicting the motion of the previous frame image signal in the frame memory, and a previous frame image signal for the next frame image signal by adding the previous frame image signal whose motion is predicted by this means and the decoded signal from the decoding means. Means for storing in the frame memory as a signal, and a motion from the motion vector decoding means. A motion vector memory for accumulating and storing a vector for each frame, a cycle detecting means for detecting a time change cycle of the motion vector accumulated and stored in the motion vector memory, and a cycle detecting means for the inputted motion vector information. Means for reading the past motion vector by the detected period from the motion vector memory, and supplying the motion vector read by this means to the motion vector decoding means as a predicted value of the input motion vector information A video decoding apparatus comprising: 7. A motion predicting means for predicting a motion of an image signal of an input frame with an image signal of a previous frame and outputting motion vector information thereof, and an image signal of a previous frame motion-predicted by the motion predicting means. And subtraction means for obtaining the difference between the image signal of the input frame, encoding means for encoding and outputting the difference signal from this subtraction means, and decoding the encoded signal from this encoding means to perform the motion prediction. A moving image decoding apparatus comprising: a decoding unit that adds to the image signal of the previous frame and supplies the image signal of the previous frame to the next input frame to the motion prediction unit, wherein the motion prediction unit is The control means for detecting the motion vector of the image signal of the previous frame with respect to the image signal of the input frame, and the motion vector detected by this control means are input to the input frame. A motion vector memory that stores and stores each frame, and a change pattern in the past N sections from the input frame and a past change pattern that is a change pattern within a predetermined error for the motion vector stored and stored in the motion vector memory. Pattern detecting means for detecting another section, means for reading the motion vector after the different section detected by the pattern detecting means from the motion vector memory, and motion vector read by this means for the motion of the input frame And a means for use as a predicted value of a vector. 8. Input means for inputting a coded signal and motion vector information for a frame image signal, decoding means for decoding a coded signal from this input means, and motion vector information from said input means. And a motion vector decoding means for decoding a motion vector to output a motion vector, a frame memory storing a previous frame image signal for the frame image signal input from the input means, and a motion vector from the motion vector decoding means in accordance with the motion vector. A means for predicting the motion of the previous frame image signal in the frame memory, and a previous frame image signal for the next frame image signal by adding the previous frame image signal whose motion is predicted by this means and the decoded signal from the decoding means. Means for storing in the frame memory as a signal, and a motion from the motion vector decoding means. A motion vector memory that stores and stores a vector for each frame, and a change pattern in the past N sections from the input frame and a change pattern within a predetermined error for the motion vector stored and stored in the motion vector memory. Pattern detecting means for detecting another past section, means for reading out a motion vector after another section detected by the pattern detecting means from the motion vector memory, and motion vector read out by this means And a means for supplying the predicted value of the motion vector information to the motion vector decoding means.