JPH07154799A - Moving picture encoding method - Google Patents

Abstract

PURPOSE:To improve the encoding efficiency by using the frame of the channel having a minimum prediction error of every macro block out of the frame of the channel to be subjected to predictive encoding and frames of a pair of channels adjacent to this channel to perform movement compensation prediction. CONSTITUTION:A mode discriminating circuit 16 compares original picture data inputted from a first input terminal 19 with predictive error picture data obtained by the difference between original picture data and current decoded data using the data of the preceding frame and sends original picture data or predictive error picture data to an encoding part in the next stage in accordance with the comparison result. This encoding part consists of a DCT circuit 3, a quantization circuit 4, and a VCL 5, and prediction is performed with a P or I picture frame by switching of a switch 6, and the output of the VLC 5 is sent to a comparator 8, and it is discriminated whether the frame has a minimum movement compensation predictive error or not. Data of the frame having it is stored in a buffer 9 and is fed back to the circuit 4, and it is used as a spare of the frame to be obtained, and a moving picture signal is outputted from a buffer 9.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coding method in the case where an image signal is composed of a large number of channels such as a multi-view type three-dimensional moving image.

[0002]

2. Description of the Related Art A conventional moving picture coding method is described, for example, in "International Standard for Multimedia Coding", edited by Hiroshi Yasuda, published by Maruzen P1.
As disclosed in 26-156, since the moving image is transmitted by one channel, there is no disclosure of a method or a problem in the case of encoding a plurality of channels.

[0003]

[Problems to be Solved by the Invention]
If the moving image coding method of the channels is to be extended to the multi-lens type three-dimensional as it is, each frame corresponding to the multi-lens requires a frame for performing the intra-frame coding in order to perform the coding individually, It can be predicted that the total amount of data increases as the number of channels increases, and as a result, the transmission efficiency of image signals deteriorates. Therefore, a coding method with a high compression rate is required.

The problem to be solved by the present invention is to realize a more advanced moving image compression method with a simple structure in view of the problems in such a simple extension of the prior art.

[0005]

SUMMARY OF THE INVENTION The present invention is a method for encoding a multi-channel image signal that constitutes a three-dimensional moving image, and is used when time-axis direction predictive encoding is performed for each frame of each channel. This is a method of performing motion compensation prediction using a frame of a channel to be predictively coded and a frame of a pair of channels adjacent to the frame of a channel in which a motion compensation prediction error for each macroblock is the smallest.

[0006]

With the above configuration, when the predictive coding in the time axis direction is performed, the motion compensation prediction error is reduced, the amount of data to be transmitted is reduced, and as a result, the prediction error data coding efficiency is improved.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the moving picture coding method of the present invention will be described in detail below with reference to the drawings.

2 to 4 are block diagrams showing an embodiment of a multi-view 3D inter-picture prediction method. In these figures, I and I ′ are I-pictures (Intra-coded Picture:
Intra-coded image), B is B picture (Bidirectio)
nally Predictive-coded Picture: P is a P-picture (Predictive-coded Picture:
Forward predictive encoded image).

First, FIG. 1 is a block diagram for explaining a forward prediction method. Here, for a P picture frame for which forward prediction is performed on an arbitrary channel chN, channel chN and adjacent channels chN-1, chN + 1
Each I picture frame immediately preceding in time is used to perform prediction in macroblock units, and the I picture frame of the channel having the smallest motion compensation error out of the three prediction results obtained is used. Perform inter-frame prediction.

It should be noted that 16 × 16 pixels are set as one macroblock.

Next, FIG. 2 is a block diagram for explaining a first embodiment of the bidirectional prediction method. Any channel here
For a B picture frame for bidirectional prediction in chN, prediction is performed in macroblock units using data of I picture frame and P picture frame immediately before and after each of channels chN, chN-1, chN + 1.
Inter-frame prediction is performed using the I picture frame and the P picture frame of the channel having the smallest motion compensation prediction error among the three prediction results obtained.

Further, FIG. 3 is a block diagram for explaining a second embodiment of the bidirectional prediction method. Here, with respect to a B picture frame for which bidirectional prediction is performed in an arbitrary channel chN, image data of any one of I (or P) picture frames immediately preceding in time of each of channels chN, chN-1, and chN + 1, P immediately after time
(Or I) Bidirectional prediction is performed temporally in macroblock units using any one of the image data of (I) picture frames, and this prediction is performed for all 9 combinations as shown in Table 1.

[0013]

[Table 1]

Of the nine combinations obtained,
The inter-frame prediction is performed using the frames related to the combination having the smallest motion compensation error.

Next, each of the channels chN, chN-1, chN +
The moving picture coding apparatus used in 1 will be described with reference to FIG. Each channel chN, chN-1, chN +
It is assumed that both have the same device, and that the input data is further divided into macroblock units each consisting of 16 × 16 pixels.

In FIG. 4, the original image data input from the first input terminal 19 is recorded in the memory 18. This memory 16 has a storage capacity for storing data for one macroblock, and sends the data for each macroblock to the processing of the next stage.

Reference numeral 16 is a mode determination circuit, which has a first input terminal 1
9 is compared with the prediction error image data obtained from the difference between the original image data input from 9 and the current decoded data using the data of the previous frame, and whether the original image data or the prediction error image data is determined by this determination result. It is sent to the encoding unit at the next stage by switching the first switch 2.

Here, as shown in the description of FIGS.
Consider computing a picture frame and then a P or B picture frame.

The encoder is a DCT (Discrete Cosine Tran).
sform: Discrete cosine transform circuit 3, quantization circuit 4, V
In the case of a P or B picture frame which is composed of an LC (Variable Length Coder) 5 and is switched by the second switch 6, the output of the VLC 5 is input to the comparator 8 via the first buffer 7, and this comparator 8 , It is determined whether or not the frame has the smallest motion compensation prediction error.

The data of the frame determined to be the minimum is stored in the second buffer 9, and this data is fed back to the quantizing circuit 4 to be used for prediction of the frame to be obtained, and the second buffer is also used. 9
From the output terminal 22, the encoded moving image signal is output and transmitted or recorded.

When the frame is a P picture frame, the second buffer 9 determines which channel of the I frame the data output from the second buffer 9 is predicted by using the I frame. It is further sent to the motion compensation frame memory 13, and the decoded data corresponding to this is used when predicting a B picture frame.

On the other hand, when the data of the intra-coded I picture frame is output from the VLC 5, the second buffer 9 is directly switched by switching the second switch 6.
And the encoded moving image signal is output from the first output terminal 22.

The quantized moving image data output of the quantization circuit 4 is an inverse quantization circuit 10 and an inverse DCT circuit 11.
Is input to the adder 12 and becomes the locally decoded image data of the previous frame, and the motion compensation frame memory 13
Will be saved in.

The output from the frame memory 13 is the third
It is sent to the adder 12 via the switch 15. As described above, the data switched and transmitted by the first switch 2 is DCT-converted by the DCT circuit 3, and the DC obtained next.
The T coefficient is quantized by the quantizer circuit 4, then variable-length coded by the variable-length coding circuit 5, and the comparator 8 is quantized by the second switch 6.
Finally, the data is transmitted or recorded through the second buffer 6 depending on whether it follows the route passing through or passing through as it is.

At this time, the mode determination circuit 16 operates the first switch 2, the second switch 16 and the third switch 15 in an interlocking manner. Therefore, when the I picture frame is encoded, the original image data is directly passed from the first input terminal 19 through the first switch 2, and the locally decoded data from the motion compensation frame 13 is passed through the third switch 15. It is then fed back to the adder 9.

The motion compensation frame memory 13 is provided with a second input terminal 20 and a second output terminal 21 for performing input / output with another channel. Therefore, the data of the I and P picture frames of this channel is sent to the second input terminal of another channel via the second output terminal 21,
Conversely, the data from the second output terminal of the other channel is the second
It is taken into the input terminal 20.

The fourth switch 17 feeds back the original image data twice when the input data is a P picture frame and twice or eight times when the input data is a B picture frame, so that the motion compensation is performed three times or nine times in total. Difference data between the data sent from the frame memory 13 and the original image is calculated. In addition, from the second buffer 9 to the memory 18
A signal indicating that the sending of the block is completed is sent to, and the data of the next block stored in the memory 18 is sent.

That is, when the frame is a P frame or a B frame, and when prediction is performed only in the forward direction or backward direction in terms of time, the data output from the motion compensation frame memory 13 is chN-1, respectively. chN, chN +
It is the predicted image data of the frame obtained for 1 and is fed back twice by using the fourth switch 17 in order to calculate the difference from the original image data for each. Therefore, the original image data is input to the subtractor 1 three times.

Similarly, the frame is a B frame, and FIG.
When performing the prediction as shown in Table 1, the data output by the motion compensation frame memory 13 is the predicted image data using the nine types of prediction shown in Table 1, and the difference from the original image data is subtracted. The fourth switch 17 is used to feed back the original image data eight times for calculation by the instrument 1. Therefore, the original image data is input to the subtractor 1 nine times.

In the structure of FIG. 4, in the case of FIG. 1, the I picture frame data of this channel and the adjacent channel are sequentially sent from the motion compensation frame memory 13 to the subtractor 1 to be encoded P picture frame. The difference data with is calculated.

In the case of FIG. 2, the differential data of the I picture frame and the P picture frame immediately before and after this channel and the adjacent channel are sequentially sent from the motion compensation frame memory 13 to the subtractor 1 for encoding. Difference data from the B picture frame is calculated.

Further, in the case of FIG. 3, from the motion compensation frame memory 13, one of the image data of the I (or P) picture frame immediately preceding in time of each of this channel and the adjacent channel, and the time of the image data. The difference data with any one of the image data of the P (or I) picture frame immediately after is calculated, and the difference data obtained by performing this for all combinations is sequentially sent to the subtractor 1 to be encoded B frame. The difference data with is calculated.

The original image data of the relevant frame of chN is sent from the output terminal 24 to the motion vector detection circuit 14 of chN-1 and chN + 1, and the original image data of the relevant frame of chN-1 and chN + 1. Is input to the motion vector detection circuit 14 through the input terminal 23.

In the motion vector detection circuit 14, if the frame is a P or B picture frame, chN-1
A motion vector is calculated for each macroblock from the image data of the previous frame and the subsequent frame of chN and chN + 1 and the image data of the frame. This motion vector is VLC
5 and the motion compensation frame memory 13.

[0035]

As described above, the present invention enables reduction of encoded data in encoding data having a plurality of channels such as a multi-lens type three-dimensional image.
This can be expected to improve the transmission efficiency.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a P picture frame encoding system to which a moving image encoding device according to the present invention is applied.

[Fig. 2] Fig. 2 is a block diagram illustrating a first encoding method of a B picture frame to which the moving image encoding device of the present invention is applied.

FIG. 3 is a block diagram illustrating a second B picture frame encoding system to which the moving image encoding device of the present invention is applied.

FIG. 4 is a circuit block diagram showing a hardware configuration of each channel that realizes the encoding method of FIGS.

[Explanation of symbols]

 1 Subtractor 2, 6, 15, 17 Switch 3 DCT circuit 4 Quantization circuit 5 VLC 7, 9 Buffer 8 Comparator 10 Inverse quantization circuit 11 Inverse DCT 13 Motion compensation frame memory 14 Motion vector detection circuit 16 Mode determination circuit 19 ~ 22 input (output) power terminal

Claims (1)

[Claims] 1. A method for encoding a multi-channel image signal forming a three-dimensional moving image, wherein a channel to be predictively encoded when performing time-axis direction predictive encoding for each frame of each channel. And a frame of a pair of channels adjacent thereto, the motion-compensated prediction is performed using the frame of the channel in which the motion-compensated prediction error for each macroblock is the smallest.