JPH047152B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an interframe coding device that compensates for the motion of a television signal.

Digital transmission of television signals uses an interframe coding method that encodes and transmits the difference signal between adjacent frames (frame difference signal), which allows transmission to be made faster than when using normal pulse code modulation (PCM). The number of bits can be significantly reduced, and a high compression ratio (ratio in which the number of transmitted bits is reduced compared to PCM) can be obtained, especially for still images and images with little movement. However, for images that include large movements, the frame difference signal becomes large, so the compression ratio decreases. "Motion compensated interframe coding" has been proposed to maintain a high compression rate even for images containing large motion. In this method, the motion of the television signal is detected, a predictive signal is generated that compensates for the motion of the television signal, and predictive coding is performed.

<Principle of Motion Compensated Interframe Coding> FIG. 1 shows an object that was at point B' in the previous frame moved to point A in the current frame. In motion compensated interframe coding, the amount of displacement υ→ (this υ→
is called a motion vector), and the predicted signal for the signal value Y(r→) at point A in the current frame is the signal value Y'(r→) at point A', which is the predicted signal in the case of simple interframe coding. ) instead of the signal value Y′(r→+υ
→)
Use. Note that r→ is a position vector indicating the position on the television screen.

The prediction error signal Y(r→)−Y′(r→+υ→) in motion compensated interframe coding is smaller than the prediction error signal Y(r→)−Y′(r→) in simple interframe coding. Since this value is small, even images with large motion can be efficiently encoded by motion compensated interframe encoding.

As a method for detecting this motion vector, for example, Ninomiya describes ``Motion correction in interframe coding (IEICE Image Engineering Study Group, document number IE78-
6, May 26, 1978, document 1) can be used. In this method, the television signal is divided into a plurality of blocks, and the television signal in each block of the current frame is divided into various amounts of displacement (referred to as shift vectors) with respect to the same position on the television screen. An evaluation value indicating the degree of similarity between a signal in a block in the previous frame and a signal in a block in the current frame that are shifted by the same amount is determined, and a shift vector for the block in the previous frame with the highest degree of similarity is detected as a motion vector. Note that the evaluation value for this similarity judgment is the sum of the absolute values of the difference signals between the intra-block signal of the current frame and the intra-block signal of the previous frame shifted by the shift vector, or a threshold value at which the absolute value of the difference signal is constant. The number of items exceeding the value is considered.

<Conventional Disadvantages> Although the principles and advantages of motion-compensated interframe coding have been described above, motion-compensated interframe coding has the following disadvantages. In other words, in the above-mentioned interframe coding, when the quantization characteristics for the prediction error signal are coarsened, there is a peculiar phenomenon called the dirty window effect, where quantization noise appears stuck on the television screen in the still part of the image. However, in motion compensated interframe coding, a new image quality deterioration occurs in which the dirty window appears to move, especially when the input television signal-to-noise ratio (S/N ratio) is low. This is noticeable in the still part of the image.

The reason why such image quality deterioration newly occurs in motion compensated interframe coding is as follows.
As mentioned above, in motion-compensated interframe coding, predictive coding is performed using the previous frame signal with the highest similarity to the block signal of the current frame as a prediction signal, but due to noise contained in the input television signal, Even if it is a part, the similarity to a vector indicating that the block is stationary is not necessarily the highest, but rather the similarity to a vector that is not stationary is higher, and the stationary part is determined to be moving. may be done. Therefore, a dirty window that appears stationary in interframe coding will appear to be moving in motion compensated interframe coding. Note that this movement of the dirty window occurs in units of blocks, so if you observe this on the screen, the dirty window will appear to move around randomly in units of blocks, giving an extremely unnatural impression to the viewer. .

<Summary of the Invention> An object of the present invention is to provide a motion-compensated inter-frame encoding device that significantly reduces the factors of image quality deterioration in the above-mentioned motion-compensated inter-frame encoding.

The present invention provides a means for detecting a motion vector for a block of an input television signal, and a vector (zero) indicating that the image is stationary among the motion vectors detected for a block temporally and spatially adjacent to the block. A motion vector correction means for correcting the motion vector for the block to a zero vector regardless of its vector value when the frequency of occurrence of a vector (referred to as a vector) is high; and means for generating a predictive signal and predictively encoding the input television signal.

<Principle of the invention> Next, the principle of the invention will be explained. In this invention, a motion vector detected for a certain block in the current frame (referred to as a block of interest) is compared with a motion vector detected in blocks surrounding the block of interest (referred to as a reference block), and If the motion signal is stationary, the motion vector detection result for the block of interest is corrected to a vector indicating that the block is stationary (that is, a zero vector). By generating a prediction signal based on a motion vector that has undergone such correction (corrected motion vector), serious image quality deterioration in which the dirty window described above appears to move can be alleviated.

Next, the motion vector correction method of the present invention will be explained in more detail with reference to FIG. Figure 2 a-c
is a diagram showing an example of the arrangement of reference blocks, in which X indicates the block of interest, and A to P indicate the positions of the reference block on the television screen with respect to the block of interest X. Fig. 2a shows the case where reference blocks A to D are arranged in the same frame as the block of interest X, on the left side and above the block of interest X, and Fig. 2b shows the case where reference blocks A to D (in the same frame as the block of interest X) are arranged. (same as in Figure 2a), block E one frame before the block of interest X, and F to M spatially surrounding E are placed as reference blocks; Reference blocks A~ surround the block of interest X in the same frame as block X.
This is a case where D, N to R are arranged.

In this invention, a reference block is determined in the temporal and spatial vicinity of the block of interest X, as shown in FIGS. Regardless of the result, the motion vector of the block of interest X is corrected to a zero vector. The reason for performing such a correction is that if a reference block in the temporal and spatial vicinity of the block of interest X is stationary,
This is based on the fact that the block of interest is often stationary due to the high correlation of television signals. Note that other logic may be used as the correction logic when correcting the motion vector of the block of interest X to a zero vector, but this will be described later.

By correcting the motion vector as described above, it is possible to prevent blocks that are supposed to be stationary from being determined to be moving due to noise, etc., which can cause serious image quality deterioration such as the dirty window appearing to move as described above. It has the great advantage of being able to reduce

<Example> Next, an example of the present invention will be described with reference to the drawings. FIG. 3 is a block diagram showing the configuration of a motion compensated interframe coding apparatus according to the present invention. In Figure 3, analog/digital converted television signals (hereinafter referred to as "TV signals") are shown.
) is input to terminal 1. A TV signal input to a terminal 1 is input to a delay circuit 3 and a motion vector detector 20 via a signal line 2.
Delay circuit 3 has an input output via signal line 4.
It is used to match the timing of the TV signal and the prediction signal output to the signal line 15 in the subtracter 5, that is, to correct the time required for the above-mentioned motion vector detection, motion vector correction, and prediction signal generation.

TV input from delay circuit 3 through signal line 4
The signal is subtracted by a subtracter 5 from the predicted signal output from the variable delay circuit 14 to the signal line 15, and the difference signal, that is, the prediction error signal, is input to the quantizer 7 via the signal line 6 and quantized. signal line 8
The signal is input to the first encoder 30 and the adder 9 via the encoder 30 and the adder 9. Here, the first encoder 30 encodes the quantized prediction error signal, for example, performs variable length encoding. The quantized prediction error signal input to the adder 9 via the signal line 8 is added to the prediction signal input via the signal line 15, locally decomposed, and written to the frame memory 10 via the signal line 16. This signal is used for motion vector detection and prediction signal generation in the next frame.

The motion vector detector 20 detects a motion vector from the TV signal inputted through the signal line 2 and the TV signal approximately one frame before inputted from the frame memory 10 via the signal line 11, and detects the motion vector from the TV signal inputted through the signal line 21. and supplies it to the motion vector correction circuit 22. Note that for the signal output from the frame memory 10 approximately one frame before, the maximum detection range in the vertical direction of the motion vector is ±M horizontal scanning lines/frame, and the vertical size of the block is N horizontal scanning lines. , (2M+N) horizontal scanning line TV
Signals are output to signal line 11 in parallel. Furthermore, the TV signal on the signal line 11 is input to the delay circuit 12, and its delayed output is output to the signal line 13, and the signal line 13 also outputs the TV signals of (2M+N) horizontal scanning lines in parallel. Note that the configuration of the motion vector detector 20 is detailed in, for example, the above-mentioned document 1, so a description thereof will be omitted.

The motion vector modification circuit 22 modifies the motion vector based on the logic of motion vector modification described above and outputs the modified motion vector to the signal line 23. The correction operation of the motion vector correction circuit 22 will be described later.
The TV signal of approximately one frame before is output from the frame memory 10 and is input to the variable delay circuit 14 via the delay circuit 12. Here, the delay circuit 12
is used to compensate for the time required for the motion vector detection circuit 20 and the motion vector modification circuit 22 to determine the aforementioned modified motion vector.

The variable delay circuit 14 is input by the signal 13
The TV signal is shifted in accordance with the modified motion vector signal input through the signal line 23 and outputted to the signal line 15 as a predicted signal. This variable delay circuit 14 is configured to be able to store image signals two-dimensionally using randomly accessible memory,
The image signal of the previous frame corresponding to the modified motion vector signal is sequentially read out pixel by pixel. Note that in parallel with this readout, writing of the image signal supplied by the delay circuit 12 is performed by addressing independently from the readout side.

Further, the modified motion vector signal is inputted to the second encoder 31 via the signal line 23 and encoded, for example, variable length encoded, and outputted to the signal line 33. The prediction error signal is multiplexed with the predicted error signal and output to the signal line 35, and written to the buffer memory 36 for speed matching with the transmission line 37.
The data is read out at a transmission speed of 1 and sent to the transmission line 37.

<Movement Vector Correction Circuit> Next, referring to FIG. 4, the motion vector correction circuit 22
The configuration and operation of will be explained. The motion vector signal input through the signal line 21 is sent to the binarization circuit 72.
and is input to the delay circuit 71. Binarization circuit 72
inputs the signal value "0" to the shift register 74 via the signal line 73 if the motion vector inputted through the signal line 21 is a zero vector, and otherwise inputs the signal value "1" to the shift register 74. shift register 7
Each output of 4, that is, the signals of signal lines 77 to 79, is input to a determination circuit 80, which determines whether or not to modify the motion vector described above. If not, the signal value is “1”
is output to the signal line 81 and then to the gate circuit 76.

Here, the number of output signals of the shift register 74 matches the number of reference blocks already described. For example, when the reference block is determined as shown in FIG. 2a, the number of output signals from the shift register 74 is four. In this case, the binarized motion vector signal (binary motion vector signal) for reference block A in FIG. The binary motion vector signal for
9 will be output. The one-way vector signal is also input to the delay circuit 71 via the signal line 21. Here, the delay circuit 71 determines the time required for motion vector correction, that is, from the binarization circuit 72 to the determination circuit 80.
A delay amount corresponding to the processing time required is required. The motion vector signal output from the delay circuit 71 to the signal line 75 is input to the gate circuit 76. The gate circuit 76 outputs a signal indicating a zero vector to the signal line 23 when a signal value "0" is input from the signal line 81, and outputs a signal indicating a zero vector to the signal line 23 when a signal value "1" is input from the signal line 81. The motion vector signal input through the signal line 75 is directly transmitted to the signal line 2.
Output to 3. The motion vector is corrected in the manner described above.

<Modification> Although this invention has been described in detail above, in this invention, even if the motion vector correction logic is changed as follows, it can be realized by simply changing the logic of the determination circuit 80 in FIG. can.

First of all, in the above explanation, for example, FIGS.
In the above explanation, the motion vector of the target block X is corrected to a zero vector only when a reference block is determined as shown in FIG. It is also possible to use majority logic in which the motion vector detection result of the target block X is corrected to a zero vector when the frequency of occurrence of zero vectors is high.

Further, as a modification of this majority logic, it is also possible to decide whether or not to correct the motion vector detection result of the block of interest X to a zero vector, taking into account the temporal and spatial occurrence of zero vectors between reference blocks. For example, if you change the placement of the reference block to
In the case defined as in figure a, the reference block A,
In cases where the moving parts in the reference block are connected, such as when the motion vector of B is not a zero vector, there is a high possibility that the block of interest X is also a moving part connected to this, so the motion of the block of interest X is It is also possible to add an additional condition that the vector is not modified.

Furthermore, assuming that the arrangement of the reference blocks is determined as shown in Fig. 2b, zero vectors occur frequently in the motion vectors in reference blocks E to M, but non-zero vectors are connected in reference blocks A to D. If this occurs, it is often the case that an image that was stationary in the previous frame starts moving in the current frame, so the motion vector detection result of the block of interest X may not be corrected.

Further, it is not necessary to limit the arrangement of reference blocks as shown in FIGS. 2a to 2c, and it is also possible to increase the number of reference blocks. In that case, the number of output taps of shift register 74 in FIG. 4 increases.

As explained above, according to the present invention, it is possible to realize a motion-compensated inter-frame encoding device that has the effect of reducing the serious deterioration in image quality, which is a disadvantage of motion-compensated inter-frame encoding, in which dirty windows appear to move.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining the motion compensated interframe coding method, FIG. 2 is a diagram showing an example of the arrangement of reference blocks, and FIG. 3 is a block diagram showing the configuration of the motion compensated interframe encoding device of the present invention. 4 are block diagrams showing the configuration of the motion vector correction circuit 22. 1... Television signal input terminal, 3, 12...
...Delay circuit, 5...Subtractor, 7...Quantizer, 9
... Adder, 10 ... Frame memory, 14 ...
Variable delay circuit, 20... Motion vector detector, 22
... Motion vector correction circuit, 30, 31 ... Encoder, 34 ... Multiplexer, 36 ... Buffer memory, 71 ... Delay circuit, 72 ... Binarization circuit, 74 ... Shift register, 76 ... Gate circuit , 80...determination circuit.

Claims (1)

[Claims] 1 Divide one frame of the input television signal into a plurality of blocks, detect a motion vector representing the motion of the television image for each block, generate a prediction signal that compensates for the motion of the image, and In a motion compensated interframe coding device for predictively coding a television signal, there is provided a means for detecting a motion vector for a block of the input television signal, and a means for detecting a motion vector for a block temporally and spatially adjacent to the block. When a motion vector indicating that the image is stationary among the vectors occurs frequently, a motion vector that corrects the motion vector for the block to a motion vector indicating that the image is stationary regardless of its vector value. A motion compensated interframe coding device comprising: a correction means; and a means for generating a predictive signal based on a corrected motion vector that is an output of the correction means, and predictively coding the input television signal. .