JPS58107785A - Encoder between movement compensation frames - Google Patents

Abstract

PURPOSE:To reduce the deterioration of picture quality generated from the movement compensation remarkably, by comparing an evaluation value to a dynamic vector detected once with an evaluation value to a shift vector, and estimating inter-frame when the difference is a prescribed value or below. CONSTITUTION:A TV signal inputted to a terminal 1 is applied to a delay circuit 3, a dyanmic vector detector 20 and an inter-frame evaluation operator 24. An output of the delay circuit 3 is subtracted from an estimation signal from a variable delay circuit 14 at a subtractor 5 and an estimation error signal is given to a quantization device 7. An estimation error signal quantitized is encoded at an encoder 30 and applied to a terminal of a multiplexer 34. An evaluation value of a dynamic vector detected at a dynamic vector detector 20 is applied to a discrimination circuit 36 together with an evaluation value to a shift vector detected at the device 24, and the difference of the both is discriminated and a corrected dynamic vector is obtained. This signal is encoded 31 and applied to the multiplexer 34, and multiplexed with the estimation error signal for output.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a motion compensated interframe coding apparatus for television signals.

In the digital transmission of television signal.

An interframe encoding method is used in which difference signals between adjacent frames (frame difference signals) are encoded and transmitted. According to this method, two conventional pulse code modulations (
PCM).

The number of transmission bits can be significantly reduced, and a high compression rate can be obtained especially for still images and images with small movement. However, for images containing large movements,
Since the frame difference signal becomes larger, the compression ratio decreases.

Therefore, in order to maintain a high compression rate even for images containing large motions, a method called ``motion compensated interframe coding'' was devised. This method detects the motion of a television signal, generates a prediction signal that compensates for the motion of the television signal, and performs predictive coding. This motion compensated interframe coding method will be explained with reference to FIG. Figure 1 (
1(a) shows the previous frame, and FIG. 1(b) shows the current frame. In Figure 1, the object was at point B' in the previous frame.

The current frame shows movement to point A. In motion compensated interframe coding.

The amount of displacement V (this t is called a motion vector) between points A' and B' in the previous frame, which are at the same position on the television screen, with respect to point A in the current frame is calculated, and the displacement amount V of point A in the current frame is calculated. As a predicted signal of signal value Y(?), instead of signal value Y'(?) of point A', which is the predicted signal of simple interframe coding, signal value Y'(r+v
)k is used. r is a position vector indicating the position on the television screen. The prediction error signal (Y(?)-Y'(r+v)) in motion-compensated interframe coding is the prediction error signal (Y(?)-Y'(r+v)) in simple interframe "--1--coding".
Y(?) -Yf? ) ), so motion-compensated interframe coding can realize efficient coding even for images with large motion.

As a method for detecting this motion vector, for example, Yoshihiro Ninomiya, "Motion Correction in Interframe Coding", Institute of Electronics and Communication Engineers Image Engineering Study Group, Material No. IE78-6, 1978
May 26, 2013, hereinafter referred to as document 1) can be used. In this method, a television image 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 (with reference to the same position on the television screen).
An evaluation value indicating the similarity between the signal in the previous frame no-flock at a position shifted by □ vector) and the signal in the block of the current frame is calculated, and the evaluation value is calculated for the block of the previous frame with the highest degree of similarity. Detect the shift vector as a motion vector. Note that the evaluation value for this similarity determination is the sum of absolute values of 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.

Alternatively, the number of difference signals whose absolute value exceeds a certain threshold is considered. (When using such evaluation values, the one with the smallest evaluation value has the highest degree of similarity.

Although the principles and advantages of motion-compensated interframe coding have been explained above, motion-compensated interframe coding has the following drawbacks.

That is, in simple interframe coding.

When the quantization characteristics for the prediction error signal are made coarser,
The quantization noise remains on the screen, creating a dirty window (di
There is a deterioration in image quality peculiar to interframe coding called interframe coding. On the contrary.

In motion-compensated interframe coding, a new problem occurs in image quality in which the dirty window appears to move. This new image quality degradation is particularly noticeable in static portions of the television image when the signal-to-noise ratio (S/N ratio) of the input television signal is low. In motion-compensated interframe coding, the cause of such image quality deterioration is a first-order problem. As mentioned above, in motion-compensated interframe coding, the signal of the previous frame that has the highest similarity to the block signal of the current frame is used as the predictive signal and predictive coding is performed. However, if the input television signal contains noise, If the block is stationary, the shift vector (0
, 0) is not necessarily the maximum, and a stationary part may be determined to be moving.

Therefore, the dirty window, which is stationary in interframe coding, appears to be moving in motion compensated interframe coding, resulting in serious image quality deterioration.

The object of the invention arises through motion compensation.

Image quality deterioration caused by dirty windows appearing distorted,
An object of the present invention is to provide a motion compensated interframe coding device that is significantly reduced in size.

According to the present invention, one frame of an input television signal is divided into a plurality of blocks, a motion vector representing the motion of the television image is detected for each block, and a prediction signal is generated that compensates for the motion of the image. and motion compensation for predictively encoding the input television signal based on the prediction signal.

In the interframe encoding device, the blocks of the input television signal are spatially shifted by various shift vectors representing various displacement amounts within a predetermined range with respect to the same position on the television screen as a reference. a block consisting of a previous frame signal at a position;
means for obtaining an evaluation value indicating a degree of similarity with the block of the input television signal; and detecting a shift vector indicating a high degree of similarity among the evaluation values obtained for the various shift vectors as a motion vector. and compares the evaluation value for the detected motion vector with the evaluation value for the shift vector indicating that the television signal of the block is static +)-, t, and determines the difference between the two evaluation values. means for correcting the motion vector by outputting a shift vector indicating that the motion vector is above a certain value and that the television signal of the TV signal is stationary if it is less than the certain value; A motion-compensated interframe coding device is obtained, which is characterized in that it includes means for generating a predictive signal based on a motion vector and predictively encoding the input television signal.

Next, the principle and embodiments of the present invention will be described.

The drawback of motion-compensated interframe coding is image deterioration in which the dirty window appears to be moving.As mentioned above, the above-mentioned evaluation value changes due to noise, etc., and a static part of an image is incorrectly determined to be moving. By being done. Therefore, in a motion vector detection method that takes into account changes in the evaluation value due to noise, that is, for a block of the current frame signal in a stationary part of one image, a shift vector (hereinafter referred to as If a motion vector detection method is used in which a shift vector (0,0) is detected as a motion vector in the description, the above-mentioned image quality deterioration can be alleviated.

In other words, in the present invention, the evaluation value /Dv and the shift vector (o,
Compare the evaluation value for O) and /DJ.

If the difference between the two is less than a certain value, the block is assumed to be stationary and inter-frame prediction is performed; on the other hand, if the difference is greater than a certain value, it is assumed to be moving and the motion vector is calculated. A prediction signal is generated using

In the following explanation, the evaluation value indicating the degree of similarity between the signal of the block of the yt frame and the signal of the previous frame at a position shifted by the shift vector will be described. The smaller the value, the higher the degree of similarity. thing.

For example, the above-mentioned "intra-block signal of the current frame and the previous frame (D) that is shifted by the shift vector."

This will be explained as the number of signals whose absolute value from the in-track signal exceeds a certain threshold.

Next, the principle of the present invention will be explained in more detail based on this premise. That is, in one aspect of the present invention.

As mentioned above. and Dv (Do-DV) (Inter-frame prediction is performed assuming that this value is always stationary.

That is, a predicted signal is generated using the shift vector (0, 0). Conversely, if (Do-Dv) is greater than or equal to a certain value, a motion vector is used to generate a predicted signal. In other words, (DODV) is compared with a constant value, and based on the comparison result, the motion vector )/is is modified to a modified motion vector, and a predicted signal is generated from the modified motion vector.

It is clear from the following that even if such comparison/determination means is added, the amount of information to be transmitted hardly increases. That is, as described above, the evaluation value is a measure for determining the degree of similarity between the block of the current frame and the previous frame signal at a position shifted by the shift vector, but the difference between the D0 and Dv is (Do -Dv) is small (that is, less than a predetermined value), which means that the similarity between the current frame block and the previous frame block (prediction signal in the case of simple interframe prediction) corresponding to the shift vector (0, 0) is also small. Instead of using a prediction signal generated using motion vectors.

Even if predictive coding is performed using an interframe prediction signal, if the constant value is made small, the increase in the amount of information generated in that block is extremely small.

Note that this constant value is based on the signal-to-noise ratio (S/N ratio) of the target television signal, the size of the block that becomes the signal (er) for motion vector detection, and the evaluation for performing the above-mentioned similarity comparison. It is determined by considering the maximum possible value.

That is, the above constant value is.

The higher the S/N ratio is, the smaller the value can be set, and the larger the maximum possible value of the evaluation value is, the larger the value can be set.

As described above, according to the present invention, although it is possible to reduce the image quality deterioration in which the dirty window appears to move, there is almost no increase in the amount of information to be transmitted compared to conventional motion compensated interframe coding. 1. The effects of applying the present invention are extremely significant.An embodiment of the motion compensated interframe coding apparatus of the present invention shown in FIG. 2 will be described. In Figure 2, analog/digital conversion (A/D conversion) is performed on a television signal (hereinafter referred to as "TV signal").
) is input to terminal 1. The TV signal input to terminal 1 is sent to delay circuit 3 through signal line 2. Motion vector detector 20. and inter-frame evaluation value calculator 24
is input. The delay circuit 3 is.

In order to match the timing of the input TV signal outputted from the delay circuit 3 to the signal line 4 and the predicted signal outputted from the variable delay circuit 14 to the signal line 15 in the subtracter 5 (that is, the above-mentioned modified motion vector detection and used to correct the time required to generate the predicted signal). The subtractor 5 is
A difference signal between the TV signal input from the delay circuit 3 via the signal line 4ft and the predicted signal input from the variable delay circuit 14 via the signal line 15 is output.
! This difference signal (prediction error signal) is input to a quantizer 7 through a signal line 6 and is quantized. This quantized prediction error signal is input to a first encoder 30 and an adder 9 via a signal line 8. Here, the first encoder 30 encodes (for example, variable length encoding) the quantized prediction error signal and outputs it to the signal line 32. Further, the adder 9 adds the quantized prediction error signal from the signal line 8 and the prediction signal from the signal line 15. That is,
The quantized prediction error signal is converted to 1 by an adder 9.
Locally decoded and sent to frame memory 1 via signal line 16
Written to 0. The signal written in the frame memory 1o is used for detecting a motion vector in the secondary frame and generating a predicted signal.

The motion vector detector 20 receives multiple inputs from the signal line 2.
The above-mentioned motion vector is detected from the V signal and the TV signal of about one frame before which is input to the signal line 11, and the motion vector is output to the signal line 21, and the evaluation value corresponding to this motion vector is transmitted to the signal line. Output to 22. In addition, for the signal of approximately one frame ago that is output from the frame memory 10.1m, the vertical detection range of the motion vector is up to 10M horizontal scanning lines/frame up and down, and the vertical block size is B horizontal scanning lines. Then, signals of (2M+B) horizontal scanning lines are output in parallel. Such a signal on the signal line 1] is input to the delay circuit 12, and the delayed output is output to the signal line 13, but the signal line 13 is also connected to the /IP of the (2M-1-B) line.
It becomes a parallel output. Note that the configuration of the motion vector detector 20 is detailed in the above-mentioned document 1, so a description thereof will be omitted.

The inter-frame evaluation value calculator 24 calculates the shift vector (0, 0
) is obtained and output to the signal line 25. here.

The function of the inter-frame evaluation value calculator 24 corresponds to a part of the function of the motion vector detector 20! be.

Therefore, the inter-frame evaluation value calculator 24 can be exactly the same as the motion vector detector 20. In this case, it is sufficient to operate the motion vector detector only for the shift vector (0,0).

In addition, in the process of motion vector detection, the motion vector detector 20 can also perform evaluation value calculation for the shift vector (0,0).
, 0) can also be output from the motion vector detector 20. In this case, the inter-frame evaluation value calculator 24 is not required.

The determination circuit 26 compares the evaluation values (DV) for multiple motion vectors input through the signal line 22 with the inter-frame evaluation values (D.) input through the signal line 25. If the difference between the two (Do-Dv) is greater than a certain value, the signal line 21
The motion vector signal inputted from the above is directly output as a modified motion vector signal to the signal line 23, and if (Do-Dv) is a constant value, the signal indicating the shift vector (, 0, 0) is output as the modified motion vector signal. Output as a signal.

On the other hand, the signal of the previous frame, which is multiple outputted from the frame memory 10, is input to the variable delay circuit 14 via the delay circuit 12. Here, the delay circuit 12 includes the motion vector detection circuit 20. Inter-frame evaluation value calculator 24 and determination circuit 26
is used to compensate for the time required to determine the modified motion vector described above. The variable delay circuit 14 is
The TV signal inputted to the signal line 13 is shifted according to the modified motion vector signal inputted by the signal a23, and outputted to the signal line 15 as a prediction signal. Further, the modified motion vector signal is input to the second encoder 31 via the signal line 23, encoded (for example, variable length encoded), and outputted to the signal line 33.

The multiplexer 34 multiplexes the encoded prediction error signal from the signal line 32 and the encoded modified motion vector signal from the signal line 33.

It comes out to the signal line 35. The output signal of the multiplexer 34 is written into a buffer memory 36 for speed matching with the transmission line 37. The signal written in the buffer memory 36 is read out at the transmission speed of the transmission line 37 and sent to the transmission line 37.

Next, the configuration and operation of the determination circuit 26 will be explained using FIG. Adder 50Fi and signal line 22 provide evaluation values Dv for multiple input motion vectors.

The above-described constant values are added and output to the signal line 51. The comparator 52 receives the value from the signal line 51 and the interframe evaluation value from the signal line 25. Compare with.

If the inter-frame evaluation value is smaller, a signal value of "0" is output to the signal line 53, and if not, a signal value of "1" is output. When the signal input to the signal line 53 is "1", the dart circuit 54 outputs the motion vector signal input to the signal line 21 as a corrected motion vector signal to the signal line 23, and If the input signal is “0”, shift vector (0,0)
A signal indicating this is output to the signal line 23 as a modified motion vector signal. In this way, if the difference between the inter-frame evaluation value (Do) and the evaluation value (DV) for the motion vector is greater than or equal to a certain value, the determination circuit 26 outputs the motion vector signal as a modified motion vector signal. However, if this is not the case, a signal indicating the shift vector (0, 0) is output as a modified motion vector signal.

Here, the criterion for whether or not to perform motion vector correction is "
This section describes in detail how to determine the "constant value." The value of the above-mentioned "constant value" will change if the S/N ratio of the input TV signal or the maximum value of the evaluation value changes (that is, if the definition of the evaluation value is changed), but in practical terms The value may be approximately 10% or less of the possible values. For example, if the size of a block, which is the unit of motion vector detection, is 4 horizontal scanning lines in the vertical direction and 8 samples in the horizontal direction, and the evaluation value is ``the previous frame signal that is shifted by the shift vector from the current frame block signal'', In the case of "the number of signals whose difference signal exceeds a certain threshold", the maximum value that the evaluation value can take is 32, so the "fixed value" may be a value between 1 and 3.

Next, the variable delay circuit 14 will be explained with reference to FIG. In the above explanation, the "signal line 13" in FIG.
was represented by one line, but here, when 12 horizontal scanning lines of data are input in parallel (for example, the vertical block size is 4 horizontal scanning lines, and the vertical detection range of the corrected motion vector is is up to 4 horizontal scan lines above and below)
I will explain about it. Therefore, signal line 13 in FIG.

In FIG. 4, the subscripts 131 to 1312 are displayed separately, and for convenience of explanation, it is assumed that the smaller the subscript N of 13N is, the higher it is on the television screen. Also.

Regarding the signal line 23 to which the modified motion vector signal is input, the line to which the horizontal component is input is also connected to the signal line 231.
A line into which the vertical component is input is shown as a signal line 232.

In FIG. 4, for example, the signal line 232.

When a vertical shift control signal indicating that the corrected motion vector vertical component (S) is 0 is input, the scanning line selection circuit 270 transfers the signal on the signal line 135 to the signal line 282.
1 to signal line 136 to signal line 282□, signal line 1
The 3° signal is output to the signal line 2823, and the signal on the signal line 13B is output to the signal line 2824. Similarly, when a vertical shift control signal of another value (S) is input, the A-5-S
.

If B=A+1 and C=A+2IC-A+3, the signal on signal line 13A is sent to signal line 282I, the signal on signal line 13B is sent to signal line 2822, and the signal on signal line 13c is sent to signal line 2.
At 823.

The signal on signal line 13D is output to signal line 2824. Each output signal 282. - 2824 are respectively shift control star 2
74-277. 'Reference numbers 301-304
Since the constituent operations of the parts surrounded by the reference characters 301 are exactly the same, only the parts surrounded by the reference characters 301 will be described below.

The signal output to the signal line 2821 is input to the shift register 274. Here 7 foot register 274
The number of output taps is determined by the detection range of the horizontal correction motion vector. for example.

When detecting movement up to the left and right V□ samples, the Tanoso number is 2vH+1. The signals output from each tap of the shift reno star 274 are input to the 2 selector 278 in parallel. The selector 278 selects one of the corrected motion vector horizontal component signals input to the signal line 231, and inputs the selected signal to the selector 290 via the signal line 286.

The selector 290 outputs the signal on the signal line 286 to the signal line 15 at the time when a prediction signal for the signal on the uppermost horizontal scanning line of the current signal block is to be generated. In addition, when generating predicted signals for the signals of the second, eighth, third, and fourth horizontal scanning lines from the top of the current signal block, the signals of signal lines 287°, 288, and 289 are connected to signal line 15, respectively. Output to. In this way, the signal line 15 has
A predicted signal is obtained by compensating the signal of the current frame with the modified motion vector of signal lines 231 + 232.

According to the present invention as described above, it is possible to obtain a motion-compensated inter-frame encoding device that reduces the serious deterioration in image quality that is a disadvantage of motion-compensated inter-frame encoding, such as dirty windows appearing to move.

[Brief explanation of the drawing]

FIGS. 1(a) and (b) are diagrams for explaining the principle of the motion compensated interframe encoding method, and FIG. 2 is a block diagram showing a motion compensated interframe encoding apparatus according to an embodiment of the present invention. 3 is a block diagram showing the configuration of the determination circuit 26 in FIG. 2, and FIG. 4 is a block diagram showing the configuration of the variable delay circuit 14. l... 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, 24... Inter-frame evaluation value calculator, 26... Judgment circuit, 30.31... Encoder, 34... Multiplexer, 36... Buffer memory.

Claims (1)

[Claims] 1. A prediction signal in which one frame of an input television signal is divided into a plurality of blocks, a motion vector representing the motion of a television image is detected for each block, and the motion of the image is compensated. in a motion compensated interframe coding device that generates a prediction signal and predictively encodes the input television signal based on the prediction signal, the motion compensation interframe encoding device generates a motion compensation interframe encoding device that predicts and encodes the input television signal based on the prediction signal. a block composed of previous frame signals spatially shifted by various shift vectors indicating various displacement amounts within the specified range;
and means for obtaining an evaluation value indicating a degree of similarity with the block of the input television signal. Among the evaluation values obtained for the seed shift vectors, a shift vector indicating a high degree of similarity is
and means for detecting it as a motion vector. The evaluation value for the detected motion vector is compared with the evaluation value for the shift vector indicating that the television signal of the block is stationary, and if the degree of approximation between the two evaluation values is small, the motion vector is , means for correcting the motion vector by outputting a shift vector indicating that the professional television signal is stationary if the degree of approximation is large; 1. A motion compensated interframe coding apparatus, comprising: means for generating a prediction signal using the input television signal and predictively coding the input television signal.