JPH0484586A - Moving compensation prediction coding device and decoding device - Google Patents

Abstract

PURPOSE:To attain proper moving compensation by applying moving compensation to a data by inter-field prediction by means of adaptive prediction and applying no moving compensation to a data by inter-frame prediction or moving compensation in a range of nearly one line in the vertical direction. CONSTITUTION:A data input signal from a data input terminal 26 is fed to an in-field decoder, from which a reproduction prediction residue is obtained and it is fed to an adder 8, from which a reproduction picture signal is obtained and fed to a field memory 9 and an output signal resulting from delaying the input signal by one field is obtained and fed to a field memory 10 and a movement compensation device 29. The field memory 10 delays the input signal by one field, the one-frame delay signal is outputted and fed to an adaptive prediction device 27, the movement compensation device 29 applies movement compensation to a picture input signal for each block and the result is fed to an adaptive prediction device 27. The adaptive prediction device 27 uses a prediction mode signal at a prediction mode input terminal 31 to select a proper signal of two kinds of inter-frame and inter-field for movement compensation prediction signals in the unit of blocks and gives the selected signal to the adder 8.

Description

Detailed Description of the Invention (Industrial Application Field) High-efficiency encoding that efficiently encodes signals with a smaller amount of code in recording, transmission, and display devices that process digital signals, especially for moving images. The present invention relates to a signal encoding device and decoding device.

(Prior art) In high-efficiency encoding of moving images, a frame that uses the correlation between frames of image signals to predict the frame to be encoded using already encoded frames, and encodes only the prediction error. There is inter-predictive coding.

In recent years, motion-compensated interframe prediction, in which prediction is performed by moving an image in accordance with its movement, has become more common.

As an example of developing this and adapting the prediction method, the Journal of the Television Society vol. 1.39, No. IO, pp. 861
868, r 15/30Mb/+ between motion compensated frames.
“Inter-field/intra-field adaptive predictive coding method” (1)
985-10j etc.

This encodes television relay signals at a relatively high rate, and uses interfield (intraframe) prediction in addition to motion compensated interframe prediction and intrafield prediction for interlaced signals.

Motion compensated interframe prediction is also used in video conferencing and the like. In this case, the image to be encoded is generally a non-interlace signal obtained by thinning out one field of an interlace signal.

(Problems to be Solved by the Invention) The encoding method shown as a conventional example has the following problems in encoding an interlaced signal.

With 3 types of prediction, inter-field prediction is 17% in motion or vertically.
This is effective only when the length is about 2 lines; motion compensated interframe prediction is effective when the motion is stationary or moving or more and is within the range that motion compensation can follow; when the motion is even larger, intra-field prediction is effective. be.

As a result, inter-field prediction is rarely used, and the prediction efficiency is not much different from using only motion-compensated inter-frame prediction and intra-field adaptive prediction.

Non-interlaced signals, which are obtained by thinning out one field of interlaced signals used in video conferences, etc., contain many aliased components in the vertical direction due to the thinning, and have a problem in that appropriate interframe prediction cannot be performed.

The present invention was made with attention to the above points, and the inter-field prediction is motion compensated using adaptive prediction, and the inter-frame prediction is performed without motion compensation, or even with motion compensation, within a range of about one line in the vertical direction. By doing so, the main motion compensation prediction is performed between fields, and there are few changes in the image during that time.
It becomes easier to cover large movements, enables appropriate motion compensation, and only requires a small increase in processing amount.For interlaced signals, by using multiple past frames for prediction,
The object of the present invention is to provide a motion compensation predictive encoding device and a decoding device that can reduce the influence of aliasing components when there is little motion, increase prediction efficiency, and further reduce the amount of data in predictive encoding. be.

(Means for Solving the Problems) In order to achieve the above objects, (1) Means for obtaining signals in which signals of a plurality of past frames or fields are each motion-compensated by predictive coding of a moving image signal. (2) providing a motion compensated predictive coding device characterized by having an adaptive prediction means for inputting the plurality of motion compensated signals and outputting a signal closest to the prediction target signal as a prediction signal; (2) In predictive coding of video signals,
A means for obtaining a motion-compensated signal of one field before, a means for obtaining a signal of one frame before without motion compensation, inputting the two types of signals, and outputting the signal closer to the prediction target signal as a prediction signal. (3) Predictive decoding of a moving image signal to obtain a motion-compensated signal for each of a plurality of past frames or fields. and a prediction means for inputting the plurality of signals, selecting an input signal based on prediction information sent from an encoding side, and outputting the selected input signal as a prediction signal. It is something to do.

(Operation) As a means to solve the problem shown in the previous section, adaptive prediction is used to motion compensate the inter-field prediction, and the inter-frame prediction is either not motion compensated, or even if motion is compensated, the vertical direction is 1.
Make it within the range of a line.

Furthermore, in a non-interlaced signal, in addition to motion compensation prediction between one frame, prediction between two frames one frame before is used.

In the prediction method of the present invention, the main motion compensation prediction is performed between fields. In this case, if the motion compensation is the same, the fields are closer in time, so there are fewer changes in the image between them, making it easier to cover large movements. On the other hand, when the image is nearly stationary, prediction can be performed favorably using interframe prediction.

For non-interlaced signals, frames with matching aliasing components are used, so prediction efficiency is improved compared to using only one frame.

(Embodiment) FIG. 1 is a block diagram showing a first embodiment of a motion compensation predictive coding apparatus of the present invention.

Although its operation is almost the same as that of a conventional encoding device, the method of forming a prediction signal is different.

In FIG. 1, an image signal input from an image signal input terminal is input to a predictive subtractor 2. It is fed to a motion compensator 3 and an adaptive predictor 4.

This image signal is assumed to be continuously input in blocks of 8×8 pixels in accordance with later adaptive processing and the like.

The prediction subtracter 2 subtracts the prediction signal output from the adaptive predictor 4 from the image input signal, obtains a prediction residual, and supplies the prediction residual to the intrafield encoder 5 .

The intra-field encoder 5 performs intra-field encoding on the input signal, and outputs it from a data output terminal 6 as digital data from which redundancy has been removed, and also outputs it to the intra-field decoder 7.

The internal processing of the intrafield encoder 5 may be simple as in the conventional example, but
Efficiency can be improved by using variable length codes such as Co51ne Transform) and Huffman codes.

On the other hand, since the predicted signal in the encoding device is the same signal as that in the decoding device, it is necessary to create it from an intra-field encoded signal.

Therefore, the intra-field decoder 7 decodes the intra-field coded signal, obtains a reproduced prediction residual, and adds it to the adder 8.
is supplied to.

The adder 8 adds the reproduction prediction residual and the corresponding prediction signal to obtain a reproduction image signal and supplies it to the field memory 9.

The field memory 9 obtains an output signal delayed by one field from the input signal, and supplies the output signal to the field memory 10 and the motion compensator 3.

Like the field memory 9, the field memory 10 delays the input signal by one field, and as a result, a one frame (two field) delayed signal is outputted and supplied to the adaptive predictor 4.

On the other hand, the motion compensator 3 detects a motion vector for the image input signal for each block using the reproduced image signal delayed by one field, and uses the detected motion vector as a one-field delayed signal (motion Due to compensation, it is no longer exactly one field) and is supplied to the adaptive predictor 4.

Furthermore, since motion vector information is also required on the decoding side,
It is output from the motion vector output terminal 11 and transmitted to the decoding side.

The adaptive predictor 3 selects an appropriate one for each block from two types of prediction signals, one between frames and one between motion compensation fields, and outputs it from a prediction mode output terminal 12.

FIG. 2 is a block diagram showing an example of the internal configuration of the adaptive predictor in FIG. 1.

In FIG. 2, the current field signal comes in from input terminal 13 and is supplied to subtractor 14.15.

Further, a one field delayed signal and a frame delayed signal are input from input terminals 16 and 17, respectively, and are supplied to subtracters 15 and 14, respectively.

The subtracter 14 calculates the difference between the current field signal and the one-frame delayed signal and supplies it to the absolute value converter 18.

The absolute value converter 18 takes the absolute value of the input signal and supplies it to the block accumulator 19.

Similarly, the subtracter 15 calculates the difference between the current field signal and the one field delayed signal and supplies it to the absolute value converter 20.

The absolute value generator 20 takes the absolute value of the input signal and supplies it to the block accumulator 21 .

The block accumulators 19 and 21 convert the respective input signals into 8
The data are accumulated in block units such as ×8 pixels and supplied to the comparator 23.

The comparator 23 compares the magnitude of the two block cumulative values, outputs an output signal for selecting the prediction error, that is, the smaller block cumulative value, and switches the changeover switch 24.
At the same time, information on which one has been selected is transmitted from the prediction mode output terminal 2 to the decoding side for each block.

On the other hand, the one-field delayed signal and the one-frame delayed signal are each supplied to a changeover switch 24,
Controlled by the output signal of the comparator 23, the block cumulative value, that is, the one with the smaller prediction error is selected and outputted from the prediction signal output terminal 25.

In such adaptive prediction, when the image is still, the frame interval is selected, and otherwise, the motion compensation field interval is selected.

In the case of FIG. 2, there are two modes of predicted values, but it goes without saying that the number of types of predictions may be increased, such as an intermediate between the two modes or four modes in which a fixed value is added.

FIG. 3 is a block diagram showing an embodiment of a motion compensated predictive decoding device of the present invention. Components that are the same as those in FIG. 1 are designated by the same reference numerals.

In FIG. 3, a data input signal transmitted from the encoding device is input from the data input terminal 26 and is supplied to the intra-field decoder 7.

The intra-field decoder 7 decodes the intra-field encoded signal to obtain a reproduction prediction residual and supplies it to the adder 8 .

The adder 8 adds the reproduced prediction residual and the corresponding prediction signal supplied from the adaptive predictor 27, obtains a reproduced image signal, and outputs the reproduced image signal via the reproduced image signal output terminal 28, and also outputs the reproduced image signal to the field memory. 9.

The field memory 9 obtains an output signal obtained by delaying the input signal by one field, and supplies the output signal to the field memory 10 and the motion compensator 29.

The field memory 10, like the field memory 9, delays the input signal by one field, and as a result, a one frame (two field) delayed signal is outputted and supplied to the adaptive predictor 27.

On the other hand, the motion compensator 29 uses the reproduced image signal delayed by one field, using the transmitted motion vector signal ]-1 input from the motion vector input terminal 30, to calculate the motion relative to the image input signal for each block. It is compensated and supplied to the adaptive predictor 27.

The adaptive predictor 27 selects the appropriate one for each block from two types of prediction signals between frames and between motion compensation fields according to the transmitted prediction mode signal input from the prediction mode input terminal 31, and adds the selected one. It is supplied to the container 8.

Compared to the motion compensator 3 and adaptive predictor 4 of the encoding device shown in FIG. 1, the motion compensator 29 and adaptive predictor 27 do not require motion vector detection or mode determination, so the amount of processing is significantly reduced. Less is better.

In the second process, motion compensation is performed only between fields. Since the correlation between images is higher by the amount of time difference between fields than between frames, it is more advantageous to perform motion compensation between fields or than to perform motion compensation between frames as in the conventional example.

However, when the motion is small or when the motion is only in the horizontal direction, it is effective to perform motion compensation for interframe prediction as well.

FIG. 4 is a block diagram showing a second embodiment of the motion compensation predictive coding apparatus of the present invention. Components that are the same as those in FIG. 1 are designated by the same reference numerals.

The only difference from the first embodiment shown in FIG. 1 is that a motion compensator 32 is added to perform motion compensation between frames, and only the differences will be described below.

In FIG. 4, motion compensation is used for both interframe prediction and interfield prediction. However, the type of motion compensation is different between fields and between frames, and while between fields is basically the same as before, between frames deals with movements only in the horizontal direction that cannot be predicted well between fields, and minute movements in the vertical direction. The vertical range is approximately ±1 line.

FIG. 6 is a diagram for explaining the range of motion vectors in the motion compensator.

The motion vector range of the motion compensator 3 that outputs the inter-field prediction signal is ±7 pixels both vertically and horizontally, as shown in FIG. 6(A).

On the other hand, the range of the motion vector of the motion compensator 32 that outputs the interframe prediction signal is ±1 because the horizontal corresponds to the same amount of motion as the interfield, as shown in FIG. 6(B).
The number of pixels is 5, but the vertical number is ±11 pixels.

Therefore, detection of motion vectors between frames requires a small number of vectors, and detection of motion vectors between frames is simple.

On the other hand, this method can also be applied to non-interlaced signals.

FIG. 5 is a diagram for explaining the state of prediction.

Fig. 5(A) shows the case of interlaced signal, Fig. 5(B)
) indicates the case of a non-interlaced signal.

Odd field (0) for interlaced signals
and an even field (e).

In the figure, squares indicate fields or frames, numbers indicate their numbers, and arrows indicate predictions.

In the case of an interlaced signal, inter-field prediction is between one frame, and inter-frame prediction is between two frames.

In the case of a non-interlaced signal, the configurations of the encoding device and decoding device are the same as in FIGS. 1 and 3, except that the field memory is replaced with a frame memory and the intra-field processing is replaced with intra-frame processing. The operation is also the same.

However, since pixels exist at the same position between frames,
In the adaptive prediction operation, the field is not used directly between frames; if the image is in a static state, the area between one frame is used for prediction.

This method is effective when there are many aliased components, such as non-interlaced signals with one-field thinning used in video conferences. When there is a lot of movement or when the object is stationary, prediction between one frame is advantageous, but when there is little movement, the adaptation of the aliasing component becomes more important than the temporal distance, and a two-frame delay is selected depending on the movement. There are cases.

Therefore, the motion compensation range for a two-frame delay may be narrow as described above.

(Effects of the Invention) The motion-compensated predictive encoding device and decoding device of the present invention motion-compensate inter-field prediction by adaptive prediction, and only a motion compensator can perform inter-frame prediction. By making the range about the same as a line, the main motion compensation prediction is performed between fields, and since the fields are closer in time, there are fewer changes in the image between them, and large movements can be easily distorted, making it suitable for motion compensation becomes possible.

Furthermore, since the range of motion compensation between frames is limited, the amount of processing increases only slightly.

For interlaced signals, adaptive prediction is performed between one frame and between two frames, but by using multiple past frames for prediction, the influence of aliasing components can be reduced when there is little movement, increasing prediction efficiency.

By being able to perform more appropriate predictions, prediction residuals are reduced, and the amount of data can be further reduced in predictive encoding.

As explained above, the motion compensated predictive encoding device and decoding device of the present invention have extremely excellent practical effects.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a first embodiment of the motion compensation predictive coding device of the present invention, FIG. 2 is a block diagram showing an example of the internal configuration of the adaptive predictor in FIGS. A block diagram showing an embodiment of a motion compensated predictive decoding device of the invention, FIG. 4 is a block diagram showing a second embodiment of the motion compensated predictive coding device of the invention, and FIG. 5 explains the state of prediction. FIG. 6 is a diagram for explaining the range of motion vectors in a motion compensator. ■...Image signal input terminal, 2...Prediction subtractor, 3゜29.32...Motion compensator, 4, 27...Adaptive predictor, 5...Intrafield encoder, 6 ...Data output terminal, 7.Intra-field decoder, 8.Adder, 9.10..Field memory 11..Motion vector output terminal, 12..Prediction mode output terminal,
13゜16.17.22...Input terminal, 14.15.
... Subtractor, 18.20 ... Absolute value generator, 19.21
... block accumulator, 23 ... comparator, 24 - switching switch] 8 switch, 25 ... prediction signal output terminal, 26 ... data input terminal, 28 ... reproduced image signal output terminal, 30...
-Motion vector input terminal, 31...Prediction mode input terminal.

Claims (3)

[Claims] (1) Means for obtaining a signal obtained by motion-compensating each of the signals of a plurality of past frames or fields in predictive coding of a video signal, and inputting the plurality of motion-compensated signals and obtaining the signal closest to the prediction target signal. A motion compensation predictive coding device characterized by having an adaptive prediction means for outputting a signal as a prediction signal. (2) In predictive coding of a video signal, means for obtaining a motion-compensated signal of one field before, and means for obtaining a signal of one frame before motion-compensated, and inputting the above two types of signals to predict the target. A motion compensation predictive coding device characterized by having an adaptive prediction means for outputting a signal closer to the signal as a prediction signal. (3) Means for obtaining motion-compensated signals of a plurality of past frames or fields in predictive decoding of a video signal, and inputting the plurality of signals and using prediction information sent from the encoding side. A motion compensation predictive decoding device characterized by having a prediction means for selecting an input signal and outputting the selected input signal as a prediction signal.