JPH04105487A - Prediction coder and decoder - Google Patents

Abstract

PURPOSE:To reduce a coding data quantity by providing a means obtaining plural frame and field signals coded precedingly in the prediction coding skipping a frame or a field of a moving picture and a prediction means generating a prediction signal closest to a signal of a frame or a field being a prediction object from the plural signals to the coder and the decoder. CONSTITUTION:Squaring devices 31, 32 and block accumulators 33, 34 are acted similarly and a mean square value of three kinds of prediction residues is fed to a comparator 30. A changeover switch 35 is selected so that a prediction signal whose square error (mean square value) is minimum among the three prediction signals from the result of the comparator 30 and the signal is outputted via a prediction signal output terminal 36. When a picture is not moving based on the adaptive prediction in this way, since a same field as a predicted field is in existence at a preceding (2N) or a succeeding (N) field, the same field is selected in the still state, and complete prediction is implemented. When any movement exists, it is highly possible that an N- preceding field with a close time distance is used at a 1st stage. Proper prediction direction is changed at a preceding and a succeeding field in a 2nd stage.

Description

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

(Prior art) In high-efficiency coding of moving images, interframe prediction utilizes the correlation between frames of image signals, predicts the frame to be predicted using already coded frames, and codes only the prediction error. There is encoding.

In recent years, motion-compensated interframe predictive coding, which predicts images by moving them according to their movements, has become more common.

JTC (Joint Technical Com) of IsO (International Organization for Standardization) and IEC (International Electrotechnical Commission)
+n1ffee) I's 5C (S++bCommi Be
e) In WG (Working Group) 8 of 2, as a standard method for video encoding, cyclic predictive encoding is performed as the first stage processing by skipping several frames, and the frames in between are used as the second stage processing. , a coding method that predicts from the frames encoded in the first stage before and after the frame is being considered, and progress is underway with plans to compile a standardization proposal in September 1990.

This method causes less deterioration in image quality even if the amount of coded data of the second-stage prediction residual is reduced, and is more efficient than simple cyclic prediction from the previous frame.

In this method, the first stage prediction operation is basically the same as normal interframe predictive coding. Jump frame 3
Converting into frames corresponds to encoding a moving image of 30 frames per second by dropping the frames into 10 frames per second.

The second-stage predictive encoding is processing of frames that were not encoded in the first stage, and is predicted from the frames before and after the frames that have been encoded in the first stage.

That is, frames in the first stage are predicted between frames in the first stage, while frames in the second stage are predicted from those in the first stage.

The second stage is adaptive prediction, and the prediction method is based on patent application No. 1-10841 filed by the same applicant and inventor as the present application.
Since it is the same as that explained in No. 9 "Adaptive Interframe Prediction Method", detailed explanation will be omitted.

In adaptive prediction, the optimal one is selected from among the following four types of prediction on a block-by-block basis.

■Linear prediction from the front and rear frames ■Prediction from the previous frame ■Prediction from the next frame ■Prediction from a fixed value (no prediction) This kind of adaptive prediction improves prediction efficiency, and the second stage is acyclic prediction. Even if the amount of residual encoded data is reduced, there is little effect on image quality.

(Problems to be Solved by the Invention) A problem arises when trying to apply a conventional encoding method to an interlaced signal.

Since images of interlaced signals move between fields even within the same frame, it is appropriate to encode the signals in units of fields. In this case, there are two types depending on whether the field interval N of the first stage is an even number or an odd number.

FIG. 6 is a diagram for explaining the operation of conventional inter-field prediction. FIG. 6(A) shows the case where the field interval N is an even number, and FIG. 6(B) shows the case where the field interval N is an odd number.

In FIG. 6, squares indicate each field and arrows indicate the direction of prediction. The numbers are frame numbers, and 0 and e are odd fields.

This is an even field.

Further, in the figure, the hatched field is the first-stage predictive coding field, and the lower part shows the field in which proper prediction cannot be made in a stationary state.

When N shown in FIG. 6A is an even number, the first-stage predictive coding field is only an even number field or an odd number field (in the figure, 0 odd number field).

In this case, there is no problem in the first stage, but in the second stage, the lines of the field with a different even-odd field (e-even field in the figure) do not match, so it is not possible to make appropriate predictions in a stationary state. , the prediction efficiency deteriorates.

On the other hand, when N shown in FIG. 6(B) is an odd number, the first-stage predictive coding field becomes an even number and an odd number alternately. In this case, there is no problem in the second-stage prediction because either the previous or the previous field is compatible, but the first-stage prediction results in prediction between fields with different evenness, making it impossible to perform appropriate prediction and reducing prediction efficiency.

Non-interlaced signals do not have the above problems, but the encoding efficiency decreases in images containing aliasing distortion, such as frame images created by thinning out one field of an interlaced signal used in video conferences and the like.

This is because even if motion compensated prediction is performed, the prediction will not necessarily be appropriate due to insufficient accuracy of motion compensation or aliasing distortion.

The present invention has been made by focusing on the above points, and performs the first stage prediction by skipping fields (frames).
The fields (frames) in between are encoded to predict from the first field (frame) as second-stage prediction, and are adaptively predicted from multiple past fields (frames). is used, and the influence of aliasing is reduced in non-interlaced signals. It is an object of the present invention to provide a predictive encoding device and a decoding device that reduce the amount of data required.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the present invention provides (1) encoding that predicts the full norm and field of a moving image using interlaced I-, Prediction n characterized in that it has a means for obtaining signals of a plurality of frames and fields, and a prediction stage that generates a prediction signal closest to the signal of the frame or field to be predicted from the plurality of previous signals. (2) I, which skips frames and fields of moving images.

A method for obtaining signals of multiple frames and fields encoded in the past by decoding that predicts by The present invention provides a predictive decoding device characterized in that it has two stages. 1 row of 1 soy, /l do interval pe odd number 5
': (1) Adaptively predict the first stage prediction from the previous coded field (previous N fields) and the second field of the previous one field (2N field 1iii). 3. The second stage processing is the same as the conventional one. 6. Even for non-interlaced signals, the first stage prediction is adaptively performed from the past two frames.1. The second stage processing is the same as in the conventional example.

(Example) This method is i:/・tarase('F+%,)゛2/
Lee: Ku 1.・Is it possible to apply C to both base signals?The embodiment will be mainly explained in the case of the base signal, which behaves like a ship's signal.

FIG. 1 is a block diagram showing the configuration of an embodiment of a predictive coding apparatus of the present invention.

An image signal input from an image input terminal 1 is supplied to a changeover switch 2.

This image signal is in units of blocks such as 8×8 pixels in accordance with adaptive processing in units of blocks to be described later.

- On the other hand, the vertical synchronization signal input from the synchronization input terminal 3 is supplied to the field counter 4.

The field counter 4 converts the input vertical synchronization signal into N
It generates control signals for the first and second stages and a field number signal in the second stage, and supplies them to the changeover switches 2, 5.6 and the adaptive predictor 7.

First, the first stage predictive encoding operation will be explained.

Changeover switches 2, 5.6 are field counter 4
Terminal C is connected to terminal a for the first-stage predictive encoding process by a control signal supplied from N fields (N is an odd number of 3 or more).

Here, since N is an odd number, the fields in the first stage are alternately even and odd fields.

Therefore, the first stage field signal is supplied to the adaptive predictor 7 and the subtractor 8-1 through the switch 2 and the switch 5.

The subtracter 8 obtains a subtracted 11 prediction residual from the prediction signal output from the adaptive predictor 7 from the first-stage field signal, and supplies it to the intrafield encoder 9 .

The intra-field encoder 9 performs intra-field encoding [7, D CT (Ditc+etp Co51ne
Spatial redundancy is removed by EJ variable length code of Than+[o+m) and Huffman code tη, etc. 1, digital data is obtained and supplied to the changeover switch 6,
Output from the data output terminal 10 (3) On the other hand, the prediction signal under the encoder necessary for prediction processing is made from the encoded signal in order to be the same signal as the decoder side. The necessary intra-field encoded compressed data is not supplied to the intra-field decoder]-1 via the switch 6 connected to the terminal a side only in the first stage.

The intra-field decoder 11 decodes the input signal 1 to obtain a reproduced residual l signal, which is supplied to the adder 1-2.

The adder 12 adds the predicted signal output from the adaptive predictor 7 to the reproduced residual signal to obtain a reproduced image signal.
It is supplied to the field memory 13.

The field memory 13 stores input signals for one field, and when a new first stage field is input, the signal in the field memory 13 is transferred to the field memory 14.

The output signals of the field memories 13 and 14 are supplied to the adaptive predictor 7 and become predicted signals.

The adaptive predictor 7 performs adaptive prediction using the current field signal supplied from the changeover switch 5, the previous field signal supplied from the field memory 13, and the subsequent field signal supplied from the field memory 14, and generates a predicted signal. and supplies it to the subtracter 8 and adder 12, and also obtains the corresponding prediction mode. Since this information is also required on the decoding side (decoding device side), it is transmitted to the decoding side via the prediction mode output terminal 16. are doing.

The contents of field memories 13 and 14 are used repeatedly for N fields, and a predicted signal is generated for all fields.

Furthermore, in the first stage, the memory contents are updated sequentially starting from the portion for which prediction processing has been completed.

Next, the second stage predictive encoding operation will be explained.

Changeover switches 2, 5.6 are field counter 4
Terminal C is connected to terminal C for second-stage predictive encoding processing by a control signal supplied from the terminal C.

Therefore, the second stage field signal is supplied to the N-1 field memory 15 through the changeover switch 2.

The N-1 field memory 15 stores the first
In order to encode the field of the second stage first, the signal of the field of the second stage is delayed by (N-1) fields and is supplied to the adaptive predictor 7 and the subtracter 8 via the changeover switch 5. There is.

The subsequent processing is the same as that for the first field, so it will be omitted, but since the second field is not used for prediction, there is no need to decode it, and the switch 4-6 is on the terminal side. , no data signal is supplied to the intra-field decoder 11.

Furthermore, the contents of the field memories 13 and 14 are not rewritten during the second stage prediction process.

The adaptive predictor 7 is similar to the one described in the above-mentioned Japanese Patent Application No. 1-108419 "Adaptive Inter-frame Prediction Method".
The first stage and the second stage differ in the part of the prediction signal that uses a mixture of two fields.

Also, there may be an independent mode without prediction as in the previous application.

The adaptive processing selects the optimal one from the following three types of prediction in units of blocks of approximately 8 x 8 pixels. The numbers in parentheses are for the first step.

0 Prediction from both fields ■ Prediction from previous (2N previous) fields ■ Prediction from subsequent (previous N) fields FIG. 2 is a block diagram showing an example of the configuration of the adaptive predictor in FIG. 1.

In FIG. 2, the output signal of the field memory 13 in FIG. 1 is input from the front field input terminal 17, and the multiplier 1
8. The signal is supplied to a subtracter 19 and a block delay device 20.

Similarly, the output signal of the field memory 14 shown in FIG. 1 is input from the rear field input terminal 21 to the multiplier 22. The signal is supplied to a subtracter 23 and a block delay device 20.

The multiplier 18 multiplies the input signal by six and supplies it to the adder 25.

Similarly, the multiplier 22 multiplies the input signal by (1-α),
It is supplied to the adder 25.

The adder 25 adds the two signals of the front field and the rear field, obtains the prediction signal from both fields, and sends the predicted signal to the subtracter 2.
6 and block delay device 20.

The coefficient α of the multiplier is determined by the field number input from the field number input terminal 24.

In the first stage, the value is -1 if linear prediction is performed, but in reality, it is l/2 a,', which is an equal addition of both, based on the S/N of the predicted signal. 3 The second stage is the position of the field subjected to linear prediction and l-1 prediction, and in case of N or 3, it is set to 27'3 for the earlier field and 173 for the later field.

On the other hand, the current field signal, which is the output signal of the changeover switch 5 shown in FIG.

The subtracter 19 calculates the difference between the current field signal and the subsequent field signal, obtains a prediction residual, and supplies the prediction residual to the overlapping unit 28 .

The squarer 28 takes the first power value of the prediction residual and supplies it to the block accumulator 29 .

The block accumulator 29 cumulatively adds the first power value of the prediction residuals for one block (5, obtains the 6th power average value of the prediction residuals, and supplies it to the comparator 30! Vessel 31.32
The block accumulators 33 and 34 operate in the same manner, and each time the −4 power mean value of the three types of prediction residuals is supplied to the comparator 30, the comparator 30 receives the three types of predictions. - of the residual
゛Compare the root mean values 11, judge whether it is the minimum value or not (obtain the corresponding prediction shade, transmit it to the decoding side via the prediction mode output terminal P16, and switch 1) On the other hand, the block delay device 20 delays the prediction signal of (j above 3-) during this judgment process and supplies it to the changeover switch 35.

The changeover switch 35 selects the -power error (-power mean value) according to the result of the comparator 30 among the three predicted bj numbers.
The predicted signal with the minimum is selected, and the predicted signal output terminal 36
3 Through such adaptive prediction, if the image is not moving, the predicted field will be the same as the predicted field.
, Noyield, ml (2NM) Mataha (N'1ij
) l:=ff exists 1'-Luno, the same field or selection island 1 in the stationary state, perfect prediction is possible.

If there is movement, the first stage uses the N previous field which is close in time distance, which has high roundness1, and in the second stage, the appropriate prediction direction changes back and forth depending on the degree of movement9.On the other hand, It is possible to use motion compensation for name prediction3.
In this case, even if the image is moving, if the motion is only in the horizontal direction, appropriate prediction can be made using 2N previous fields, making the effectiveness of the present method remarkable.

FIG. 3 shows the configuration of an embodiment of the predictive decoding device of the present invention.
゛It is a block diagram.

The operation of the decoding device that is
The operation is almost the same as that of the local decoding section of the encoding device shown in the figure.

In FIG. 3, intra-field encoded compressed data that is output from the data output terminal 10J in FIG.
is being supplied to.

--13. The vertical synchronization signal input from the synchronization input terminal 39 is supplied to the field counter 4.

The field counter 4 converts the input vertical synchronization signal into N
It counts the digits and generates control signals for the first stage and second stage, and a field number signal in the second stage.

Then, the 3° intra-field decoder 11, which supplies the changeover switch 40 and the adaptive predictor 7, converts the input signal into V:,
A reproduced residual signal is obtained (IT, supplied to the adder 1-:2).

The adder 12 converts the predicted signal output from the adaptive predictor 7 into the reproduced residual signal no-addition I1. τ, get the reproduced image signal, switch, and...! Terminal a of 40.

via d. , supply 1 to Field Mail 111 +1.

In the first stage prediction code V:, the changeover switch 40 is
As shown in the figure, the terminals, 1 and d
1 terminal C is connected. 3. Field meter L [: 3] accumulates the input signal for one field, and when a new first stage one field is input) 7, it simultaneously outputs the input signal. The image signal written in a cursive manner is input to an adaptive quantizer 7, a field memory 4), and a terminal r1 of a changeover switch 40).

The reproduced image is output from the reproduced image output terminal 41 via C.

This is because the field memory 13 is shared as a delay device in order to temporally correct the fact that the field in the first stage of the encoding device C is encoded before the second stage.

The output signals of the field memories 13 and 14 are supplied to the adaptive predictor 7 and become predicted signals.

The adaptive predictor 7 performs adaptive prediction using the front field signal supplied from the field memory 13 and the rear field signal supplied from the field memory 14 according to the prediction mode input from the prediction mode input terminal 42,
A predicted signal is obtained and supplied to the adder 12.

The contents of field memories 13 and 14 are used repeatedly for N fields, and a predicted signal is generated for all fields.

Furthermore, in the first stage, the memory contents are updated sequentially starting from the portion for which prediction processing has been completed.

Next, the second stage predictive encoding operation will be explained.

The changeover switch 40 changes from the state shown in FIG. 3 for the second-stage predictive encoding process in response to a control signal supplied from the field counter 4, and switches between terminals a, c1, and d.
is connected, and the output signal of the adder 12 is output as is from the reproduced image output terminal 41.

This operation returns the frames to their original order.

FIG. 4 is a diagram for explaining the operation of inter-field prediction in this embodiment. The prediction is shown when N is 3 in an interlaced signal.

In the inter-field prediction method of the present invention, in the first stage processing,
The previous coded field (N fields before) is a field with different evenness, or the previous coded field (2N fields before) is the same.

Adaptive prediction is made from both, so if the image is moving, the previous prediction field (3 fields ago) is closer and the image correlation is higher, so there is a higher probability that it will be selected. If the field is stationary, the next previous field (6 fields before) with the same field relationship is selected.

Although this method is particularly effective for interlaced signals, it can also be directly applied to motion compensated interframe prediction for non-interlaced signals.

In this case, it doesn't make much sense without motion compensation.

FIG. 5 is a diagram for explaining the interframe prediction operation of this embodiment.

In non-interlaced signals, as shown in Figure 5, the fourth
The field in the diagram becomes the frame.

In this case, the signal 2N frames earlier is also used, so even if the motion compensation prediction signal N frames earlier cannot be predicted successfully due to the influence of aliasing distortion, it may be successfully predicted 2N frames earlier, and the molecular efficiency is improved. Ru.

In the embodiment of the non-interlaced signal, the fields of the embodiment described above are used as frames as they are.

This is because a frame of a non-interlaced signal and a field of an interlaced signal can be considered to be the same except that the lines are offset.

In this case, the position of the line matches the immediately previous frame, so there is no particular effect in a stationary state, but if the image moves, the closer the line is, the better the prediction may be due to the accuracy of motion compensation and the aliasing distortion included in the image. It will be more effective if it fits well even if it is far away.

This becomes noticeable in images containing aliasing distortion, such as frame images created by thinning out one field of an interlaced signal.

(Effects of the Invention) The predictive encoding device and decoding device of the present invention perform first-stage prediction by skipping fields (frames), and the intervening fields (frames) are used as second-stage prediction to perform first-stage prediction. Coding that predicts from the field (frame) of
By adaptively predicting from multiple past fields (frames), for interlaced signals, fields with matching even and odd numbers are used in the stationary state, and for non-interlaced signals, the influence of aliasing distortion can be reduced, improving prediction efficiency. The improvement reduces the prediction residual, so
This has extremely excellent practical effects, such as being able to reduce the amount of encoded data.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the predictive coding device of the present invention, FIG. 2 is a block diagram showing a configuration example of the adaptive predictor in FIG. 1, and FIG. 3 is a predictive decoding device of the present invention. FIG. 4 is a diagram for explaining the operation of inter-field prediction in this embodiment, and FIG. 5 is a diagram for explaining the operation of inter-frame prediction in this embodiment. FIG. 6 is a diagram for explaining the operation of inter-field prediction in the conventional example. ■...Image input terminal, 2, 5, 6, 35.40...
Changeover switch, 3... Synchronization input terminal, 4... Field counter, 7... Adaptive predictor, 8, 19°2
3.26... Subtractor, 9... Intra-field encoder, 10... Data output terminal, 1j... Intra-field decoder, 12.25... Adder, 13.14...・・・
Field memory, 15...N-1 field memory, 16... Prediction mode output terminal, 1-7... Previous field input terminal, 1.8.22... Multiplier, 20...
・Block delay device, 21... Rear field input terminal, 2
4...Field No. input terminal, 12゛7 Current field input terminal, 28, 31. ．． 32. -1 power, 29
．． 3 (. 34... Block accumulator, 30 Comparator, 36. Predicted signal output terminal - 138 Data input terminal, 39... Synchronization input terminal, 41... Endogenous image output terminal, 1:2・・・
Prediction mode input terminal. Patent applicant: Japan Victor Co., Ltd.

Claims (2)

[Claims] (1) Coding that skips and predicts frames and fields of a moving image, which includes a means for obtaining signals of multiple previously encoded frames and fields, and a method for obtaining signals of frames and fields to be predicted from the multiple signals. 1. A predictive encoding device comprising: a prediction means for creating a prediction signal closest to the signal. (2) In decoding that predicts frames or fields of a moving image by skipping over them, there is a method for obtaining signals of multiple previously encoded frames or fields, and a means for obtaining signals of multiple frames or fields that are to be predicted from the multiple signals. 1. A predictive decoding device comprising: prediction means for creating a predicted signal closest to the signal.