JPH05328324A - Refresh system for movement-compensated inter-frame predictive coding - Google Patents

Abstract

PURPOSE:To prevent a fault from remaining in an image at the time of switching a receiving channel in the case of using both of movement- compensated inter-frame predictive coding and periodical refresh by forming an overlap part in an area using an intra-frame coding mode at the time of periodical refresh. CONSTITUTION:A picture element(PE) block (n) has the size of (W1 PE XW1 PE) and a search area SA for the block (n) is an area corresponding to the width of an LW PE around the block (n). An overlap part L of a refresh area RA is obtained by adding only width covering the area SA of the block (n) in the moving direction of the area RA. The width is LW. A frame just before the current frame is refreshed by an area RA0 and an overlap part L0 added to the area RA0. Since an area SA1 is set up so as to be covered with an overlap part L0 required at the time of using the PE block p1 of the preceding frame in order to form a block n1 included in the area RA0, the block p1 is included in the area obtained by combining the area RA0 and the part L0. Thereby data for all refreshed areas are used for the data of the block p1 constituting the block n1.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The field of motion-compensated inter-frame predictive coding, which is used in an image transmission system using motion-compensated inter-frame predictive coding, has improved the method of performing intra-frame coding, namely, refresh. Regarding the scheme.

[0002]

2. Description of the Related Art A conventional example of an image transmission system using motion-compensated interframe predictive coding will be described below.

Since a huge amount of information is generated when image information is digitized, the amount of information to be transmitted is reduced by using the correlation of images. For that purpose, motion-compensated interframe predictive coding and intraframe coding are often used. Then, these encodings are processed in units of a pixel block in which several tens of pixels are collected. Figure 4
(A) shows a transmitter that performs motion-compensated interframe predictive coding.

Video data is input to the input terminal 1,
It is input to the subtractor 2 and the motion detection / compensation circuit 7. In the subtractor 2, when the switch 8 is on, the subtraction process between the inter-frame prediction signal and the input signal is performed via the switch 8. The output of the subtractor 2 is input to the encoder 3 and encoded. The output of the encoder 3 is guided to the output terminal 9 as a prediction error signal and input to the decoder 4 for decoding. The decoded signal is the same as the output of the subtractor 2 and is input to the adder 5. The output of the motion detection and compensation circuit 7 is also input to the adder 5, and the added output is input to the frame memory 6. The motion detection / compensation circuit 7 compares the output of the frame memory 6 (the signal of the previous frame) with the signal of the current frame input to the input terminal 1 to convert the image of the previous frame into the image of the current frame. It detects whether or not there is motion, obtains a motion vector, and corrects and outputs the image of the previous frame by the amount of motion. Therefore, the output of the adder 5 is stored in the memory 6 as a signal obtained by predicting the signal of the current frame. If there is movement, the switch 8 is turned on and a signal obtained by predicting the signal of the current frame from the previous frame,
The difference (that is, prediction error) signal from the signal of the current frame is obtained from the subtractor 2. Further, the operation of the above transmitter will be described.

A frame of data currently being transmitted is frame P, and a frame to be transmitted next is frame N.
The switch 7 is assumed to be closed. The data of the frame P is decoded into image data by the decoder 4 and the adder 5, and is stored in the frame memory 6. Then, the motion detection / compensation circuit 7 detects the motion of the image occurring between the frames, and mutates the data of the previous frame in accordance with the motion to generate the predicted image data of the next frame, thereby improving the efficiency of encoding. I am trying.

FIG. 4B shows the outline of the operation of the motion detection / compensation circuit 7. In the motion detection and compensation circuit 7,
First, the pixel block n is defined in the current frame N. next,
In the previous frame P stored in the frame memory 6,
An area SA is defined so as to include the position occupied by the pixel block n in the frame N. SA is called a search area. An arbitrary pixel block p1 having the same size as the pixel block n is cut out from the area SA, and the sum of absolute values of difference values d between pixels occupying the same positions between p1 and n is calculated. p
The position of 1 is moved one pixel at a time, and p1 at which the sum is the minimum is p. Then, the shift between the positions of p and n is set as a motion vector V. The difference value d between p and n becomes the prediction error signal d. The motion vector V and the prediction error signal d are transmitted as the data of the pixel block n in the frame N.

FIG. 5 shows the structure of the receiver. Input terminal 10
The prediction error signal is input to the decoder 11 via the. Since the decoded prediction error signal is incomplete as an image, the signal from the motion compensation circuit 15 is added by the adder 12 via the switch 13. The output of the adder 12 is delivered to the output terminal 17 and stored in the frame memory 14. The output of the frame memory 14 is input to the motion compensation circuit 15, and the position is corrected according to the motion vector from the terminal 16. That is, the position of the image of the previous frame is corrected by the motion vector, and the prediction error signal is added to complete the image of the current frame.

The operation will be further described. It is assumed that the data of the currently shown frame P is stored in the frame memory 14 and the switch 13 is closed. The pixel block p in the previous frame P in FIG. 4B is displaced by the motion compensation circuit 15 by the transmitted motion vector V, and the adder 12 adds the pixel prediction error signal d to add the pixel prediction error signal d to the frame N. It is decoded into a pixel block n in.

In the above motion compensation interframe predictive coding system, there are intraframe coding processing and interframe coding processing as processing modes. The intraframe coding process is called refresh. Intra-frame coding is a transmission method in which pixel data only in a frame is directly coded and transmitted without prediction. Therefore, when the intraframe coding is used, the receiver can reproduce the screen of the frame N without using the data of the previous frame P. In the inter-frame coding process, the data of the next frame is created by adding the data of the prediction error to the data of the previous frame to reduce the amount of information. By utilizing the intraframe coding process and the interframe coding process, the amount of transmission information is reduced as a whole to improve efficiency.

However, if the data received by the receiver is erroneous at a certain point in time when the inter-frame coding processing is continuously performed, the error generated there remains uncorrected and remains thereafter. become. In order to eliminate the propagation of this error, it is necessary to periodically use the intra-frame coding and perform refresh to prevent the accumulation of errors in the data in the frame. For refreshing, one screen is refreshed at a time, or one screen is divided into several areas and one area is refreshed at a time, and one screen is refreshed over several frames or fields. There are periodic refreshes.

Here, a description will be given of the periodic refresh, which can relatively easily control the transmission buffer during transmission. Note that intraframe coding is performed by comparing the data amount when motion compensation interframe predictive coding is applied to a pixel block with the data amount when intraframe coding is performed. Even when there are few, it is forced to be executed. When intraframe coding is performed, the switches 8 and 13 in the system are open.

FIG. 6 is an explanatory diagram of the periodic refresh. In the figure, the frame is divided into strips, and a region of one strip is forcibly intra-frame-encoded and transmitted per frame. Strip a for intra-frame coding
The position of 1 is sent in sequence for each frame. For example, if one frame is divided into 11 and a1 is defined, when the screen is sent for 11 frames, intra-frame coding is performed once for all strips and the entire screen is refreshed once. become. The basic operation of the system for performing motion compensation interframe predictive coding is as described above.

When a television receiver is used, the user switches to a desired channel from a plurality of channels at a desired time and uses it. Therefore, consider a case where the receiving channel is switched from A to B on the receiver side. At this time, since it is not known that the receiver has switched the channel on the transmitter side of the B channel, the screen data is continuously transmitted using the motion compensation interframe predictive coding. However, the frame P displayed by the receiver is for the A channel, and does not have the image data for the B channel necessary for constructing the screen of the new channel. Therefore, the B-channel screen cannot be configured in the area using the motion-compensated interframe predictive coding, and a new channel cannot be received in a perfect form for a while from the user's point of view.

The periodic refresh is effective for constructing a new B channel image when the channel switching is performed. When the channel is switched from A to B, the pixel block for which motion compensation interframe predictive coding is used in the area a2 in FIG. 6 does not become a new B channel image. However, the portion a1 becomes a B channel image because the image can be reproduced without using the image data of the frame P as described above. Then, after 11 frames, the data for one frame on the screen becomes the B channel, and the new channel should be received.

However, in this system, since motion compensation is used at the time of encoding, the data of the entire screen after 11 frames may not be completely switched to that of a new channel.

FIG. 7 (7A) shows an example in which the data of the entire screen after 11 frames is not completely switched to that of a new channel. A is a signal of A channel (actually, it is a disturbed signal after 11 frames), B is a signal of B channel, which is a refreshed area, and is hereinafter referred to as area A and area B. The pixel block of interest in the area A is surrounded by a square. Figure 7
It is assumed that the channel B side has been refreshed at the time shown in A (a). At this time, if the pixel block in the area A is moving as shown in FIG. 7A (b), since motion compensation is performed, the data of the old channel is mixed in the subsequent processing. Area A
Signal of the area A is captured in a part of the area B, and a part X of the area A is converted into a new area B (1
It is possible that the area consisting of only the data of one frame before has already been refreshed) will be eroded. This part X is shown in FIG. 7A by refreshing the next frame.
Even when the surroundings are rewritten to the data of the new channel as at the time (c), the data of the old channel is left as it is.

FIG. 7B shows an example in which the image data b in the area A cannot be erased depending on the relationship between the cycle of the periodic refresh and the movement of the portion X. For example, it is assumed that there is a portion X having disordered image data in the refresh area α (FIG. 7B (a)). Area α
Suppose that the part X moves to the region β before being refreshed (time point FIG. 7B (b)), and the same part X moves to the region α again immediately before refreshing the region β (time point diagram). 7B (c)), although the areas α and β are both refreshed, the portion X having disordered image data remains unerased. If this is repeated, no matter how many times the areas α and β are refreshed, the portion X having the disturbed area A data remains.

[0018]

As described above, when the motion-compensated interframe predictive coding and the periodic refresh are used together, the disturbed image erodes the refresh area and the refresh becomes invalid, and immediately when the decoding operation is started. However, there is an inconvenience in use that a good image is not reproduced. It is an object of the present invention to provide a refresh method of motion-compensated interframe predictive coding that can reproduce a good image in a short time even when switching channels.

[0019]

In an area where the intraframe coding mode is used in periodic refresh, an overlapping part where the areas overlap each other is provided, and the size of this overlapping part is set according to a search range of a motion vector. Is done.

[0020]

[Operation] Due to the presence of the above-mentioned overlapping portion,
Even in the presence of motion compensation, the data is not brought into the refreshed area from the non-refreshed area, and when the channel is switched by the receiver, the screen is refreshed for one cycle. The whole can be completely switched to the image of the new channel.

[0021]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is an explanatory view of an embodiment of the present invention. As shown in the figure, the pixel block n is W1 pixel × W1
The search area SA having a pixel size and corresponding to the pixel block n is an area having a width of LW pixels in the vertical and horizontal directions of n. The overlapping portion L of the refresh area RA is added to the refresh area RA in the moving direction by the width covering the search area SA of the pixel block n. The width of this overlapping portion L is LW in the figure. Under the above conditions, a state will be described in which periodic refresh is performed while performing motion compensation interframe prediction coding.

As shown in FIG. 2, the previous frame from the present time is refreshed by the refresh area RA0 and the overlap portion L0 added thereto. Here, when the pixel block p1 of the previous frame is used to create the pixel block n1 in the refresh area RA0 of the current frame, the refresh overlap portion L0 is set to cover SA1. p1 surely exists inside the region where RA0 and L0 are combined. Therefore, as the data of p1 for forming n1, all the data in the refreshed area is used.

If the pixel block n2 of the previous frame is used to create the pixel block n2 outside the refresh area RA0 of the current frame, the refresh overlap portion L0 and the overlap portion L0 of the pixel block n2 are used. Data from areas other than the area refreshed by the refresh area RA0 and its overlapping portion L0 is brought into the overlapping area.
Data from areas other than the refreshed area will never be brought into the refresh area RA0.

This is expressed from the viewpoint of channel switching. Of the area B consisting of only the data of the new channel B because it was refreshed by the refresh area RA0 of the previous frame of the current frame and its overlapping portion L0, the data of the old channel is mixed in the refresh area RA0 for the above reason. Area A
Data is not brought in. Data of the old channel is brought into the overlap portion L0 of the refresh, but this L0 is rewritten by the refresh area RA1 to the data of the new channel at the time of the next refresh. That is, the block having the data of the A channel does not invade the refreshed area from the moving direction of the refresh.

On the other hand, all the pixel blocks are located on the moving direction side of the periodic refresh once while the periodic refresh is performed for one screen. Therefore, the image displayed on the receiver consists of completely new channel B images while one or two screens are periodically refreshed by the B channel image data.

In the above-mentioned embodiment, the overlap portion is added to the moving direction of the refresh area over time, but as shown in FIG. 3 (A), it is opposite to the moving direction or in FIG. 3 (B). Similarly, when the refresh areas are added to both sides of the refresh area, the image of the entire screen can be changed to an image of a new channel while the refresh is performed for one or two screens.

[0028]

As described above, when the motion-compensated interframe predictive coding and the periodic refresh are used together, the interference at the time of switching the receiving channel can be eliminated by providing the overlap portion in the refresh area. ..

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a search area, a refresh area, and an overlapping portion according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of motion compensation signal processing when refresh is performed by the method of the present invention.

FIG. 3 is an explanatory view of an overlapping portion in another embodiment of the system of the present invention.

FIG. 4 is a block diagram showing the principle of the transmitter according to the present invention and an explanatory diagram of motion compensation.

FIG. 5 is a block diagram showing the principle of a receiver according to the present invention.

FIG. 6 is a conceptual diagram for explaining periodic refresh.

FIG. 7 is an explanatory diagram for explaining a problem when channel switching is performed by a conventional refresh method.

[Explanation of symbols]

W1, W2 ... Pixel block width, SA ... Search area,
RA ... refresh area, L ... overlap portion added to RA, LW ... motion vector search range, RA0
... refresh area of one previous frame of current frame, L0 ... overlap portion added to RA0, n1, n2 ... pixel block of current frame, p
1, p2 ... Pixel block of previous frame, RA1
... refresh area of current frame, SA1 ... n1 search area, SA2 ... n2 search area, L1
... overlap portion added to the refresh area RA on the opposite side to the moving direction of RA, L2 ... overlap portion added to the refresh area RA on the moving direction of RA, N, P ... The frame that was there, n,
p, p '... Pixel block, SA ... Search area, d ...
Difference value between pixels, V ... Motion vector.

Claims (3)

[Claims] 1. In an image transmission method using motion-compensated interframe predictive coding, intraframe coding is forcibly performed in one area in one frame or one field (hereinafter, this is referred to as refresh).
When you refresh the data of the entire screen by shifting this area over several frames or several fields by using, the area that uses the intra-frame coding mode is set according to the search range of the motion vector. Refresh method for motion-compensated inter-frame predictive coding, characterized in that they are overlapped with each other. 2. The refresh method for motion-compensated interframe predictive coding according to claim 1, wherein the overlapped portion is set on the side of shifting the area in the area. 3. The refresh method for motion-compensated interframe predictive coding according to claim 1, wherein the overlapped portion is set on a side opposite to a side on which the area is shifted.