JPS6373786A - Hierarchical burst communication system - Google Patents

Abstract

PURPOSE:To suppress the influence for the deterioration in a quality to a burst deficiency to a minimum by applying a hierarchization according to the contribution of the quality and a precedence according to the contribution of the quality and applying a proper redundancy suppressing technique in the respective hierarchies. CONSTITUTION:Input data 1a from a video data source 11 is divided into three hierarchies 2a-2c, for instance, different in the contribution for the quality in a hierarchization part 12 and a redundancy suppression encoding proper for the respective hierarchies in view of the repercussion of the deterioration in the quality is applied. The encoded data 3a-3c of the respective hierarchies 2a-2c is once decoded in a reverse hierarchization part 14 and the decoded data 19 is compared with the input data 1a in a quality decision part 15. The contribution to all the quality is decided, a signal relating to the precedence is fed to a precedence applying part 16, and a precedence applied packet or burst 21 is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] This invention relates to a burst communication system that encodes temporally changing multidimensional information such as a video signal and transmits it as a burst.

"Conventional technology" (g) Coding technology: Conventional high-efficiency coding technology can be roughly divided into predictive coding and transform coding. In these encoding technologies, the idea is to divide encoded data into layers according to their contribution to quality, and to apply redundancy suppression techniques that have different effects on quality deterioration when data is missing for each layer. There wasn't.

On the other hand, as a concept for layering video signals, there is a sequential transmission method based on a still image pyramid structure (Reference: 5
-L-Tanimoto*”1mageTransmi
ssi on with Gross Inf
organizationFirst', Computer G
RAPHICS AND IMAGE PROCESS
ing c+, pp72-76, 1979). For example, an original image of NXN pixels is divided into blocks of nxn pixels (N>n), and a new image (this is called a pyramid) with the average value of each block as the new pixel value is created one after another. In this system, the image is gradually made visible to the other party by transmitting from the pyramid that has the average value of the larger part, and gradually transmitting the values of the finer parts.

(() Burst and packet communication methods: In conventional packet switching methods, there was a concept of priority, such as giving higher priority to packets with strong real-time characteristics such as voice broadcasts, but There was no idea of categorizing them according to their contribution to the project and prioritizing them based on their contribution.

(Ushio) Disadvantages of conventional technology: Data is layered according to its contribution to quality, and redundancy suppression technology with different ripple effects of quality deterioration when data is missing is applied to each layer. Because it was not possible to transmit data by attaching it to each layered data, it is now possible to predict and control the impact on quality when decoding due to deterioration such as code errors in one line or packet loss. Therefore, in order to ensure a certain level of quality, it was necessary to anticipate the worst and use excessive error protection to perform inefficient transmission.Especially in the transmission of video signals. The effects of errors are more difficult to predict as the encoding algorithm becomes more complex. Therefore, various error protection and refresh techniques are used to prevent data errors or omissions from occurring in video bursts or packets (for transmission). This required complex control over two parts.

Furthermore, when throughput decreases due to traffic conditions that occur with a certain probability at each node in the network, attempting to control the quality deterioration caused by burst loss ζ requires complex exchanges such as control signals from the node to the information source. became.

As explained above, the conventional method basically does not hierarchize data according to the degree of contribution to quality and give priority to the degree of contribution to quality. However, no measures could be taken to minimize the impact.

"Means for Solving Problems" This invention appropriately hierarchizes data according to the degree of contribution of data quality.

Furthermore, from the perspective of quality deterioration spreading, a redundancy suppression method suitable for each layer is used.

Furthermore, appropriate priority is given to the data C2 of each layer.

In this way, it is possible to control and minimize the impact of quality degradation on bursts and packet loss due to transmission errors or network traffic conditions C2, which conventionally required complicated error control. . To enable packet transmission or burst transmission of video signals with simple control and efficiency.

Embodiment First Embodiment FIG. 1 shows a first embodiment of the present invention. A source 11 of video data to be transmitted (for simplicity, data for one frame is assumed here) is connected to a layering section 12 (; each layered output thereof is supplied to an encoding section 13. Encoding Each layered encoded data in the section 13 is decoded and reconstructed into the original data in the dehierarchization section 14.The quality determination section 15 evaluates the contribution of the layered data to the encoding quality and the overall The output side of the encoding unit 13 is also connected to a priority bursting or packetizing unit 16 to grasp the quality.

Input data 1a from the video data source 11 is sent to the layering unit 12.
Qualitatively (- divided into two or more hierarchies with different degrees of contribution, in this example three hierarchies 2a, 2b, and 2C. Furthermore, these stratified data 2a.

2b and 2c are subjected to redundancy suppression coding associated with the respective layers in the coding unit 131. Each redundancy suppression technique is selected from the viewpoint of the ripple effect of quality deterioration when data is lost. For example, three hierarchies 2a, 2b,
2c is a layer 2a with a low quality contribution, a layer 2c with a high quality contribution, and a layer 2b with an intermediate quality contribution, then the layer 2a has a low priority when transmitting, and is prepared for data loss. , for this layer 2a, we do not use a coding method that has a high degree of deterioration, such as interframe prediction, and use the 5th layer 2.
Since C is considered to have a high priority, an encoding method that fully incorporates effects such as prediction and motion compensation is adopted.

Encoded data 3a+ of each layer 23a 2 b+ 2 c
3b and 3c are once decoded by the dehierarchization unit 14.

The decoded data 19 is compared with the input data 1a in the quality determination section 15. At this time, the de-hierarchization unit 14 decodes the data of each layer independently using the control signal 17, determines the degree of contribution to the overall quality, and adjusts the encoding parameters as necessary. conduct. Final ('-Group section 16 with superior priority: Signal 1 related to priority C
8 is sent from the quality determination unit 15, and the packet or burst 21 is output as a packet or burst (-) with a priority given to it.

In this way, by layering according to the degree of contribution of quality and giving priority according to the degree of contribution of quality, and by applying appropriate redundancy suppression technology that has different ripple effects such as errors in each layer, The influence of quality deterioration due to burst loss can be suppressed to a minimum of C2. Furthermore, it is possible to control video quality in burst communications independently and relatively easily at the network and news source.

In the embodiments of the present invention, there are many combinations of hierarchization methods, combinations with encoding methods, priority assignment methods, etc. (1) of these will be explained below.

No. 29 shows a second embodiment of the present invention. This embodiment shows a two-film structure C2 as an example of a multi-stage structure. First, the prediction signal from the prediction circuit 25 is subtracted by the subtraction circuit 24 of the first predictor, and the correlation component of the input image signal 23 is removed. The output is quantized by a quantization circuit 26 as an output belonging to the first layer, and is given a priority l and burst-formed by a bursting circuit 27. This component is always transmitted to the receiver, and has a certain image quality. The output of the quantization circuit 26 is also input to the local decoder 28, where it is dequantized and input to the prediction circuit 25C.

'As a second stage, the difference between the first stage residual component from the look-up circuit 24 and the output of the quantization circuit 26 is obtained by the subtraction circuit 29, and then the difference is calculated by the first stage prediction circuit 25 and the quantization circuit 26. circuit 28
The prediction circuit 31 predicts this difference signal using the preliminary information of The quantization circuit 33 quantizes the signal, and the bursting circuit 34 assigns priority 2 and bursts it as a second layer signal.

This priority level 2 component is a component that further improves image quality. A multi-stage predictor like this! By using this method to hierarchize and at the same time give appropriate priority, it becomes possible to suppress the influence of quality deterioration and to perform burst transmission efficiently.

The output of the quadrupling path 33 is dequantized by the local decoder 35 and supplied to the prediction circuit 31. Burst circuit 27.3
The four outputs are burst multiplexed by the output circuit 36 and output to the transmission line 37. At this time, bursts with priority 1 are always sent, but bursts with priority 2 are sent when there is a margin on the transmission line 37C.

Section 3 (8) shows a third embodiment of the present invention, and is a case where two-stage hierarchization is performed. Input image signal 23
is divided into NXN subblocks (for example, 4X4, etc.) by a frame memory part 41, and a two-dimensional orthogonal transform circuit 42 performs D CT (Direct
An orthogonal transformation is performed in a two-dimensional domain using an orthogonal transformation such as Co51ne Transform (Discrete Cosine Transform). The converted conversion coefficients are sequentially converted from the general C2 low-frequency components in the conversion section 43, and the bursting circuit 27 converts the converted coefficients up to a certain range into the first one (here, the generated low-frequency components and the remaining (divided into two layers with high frequency components)
Priority (- 1 burst as multiple bursts) The low-frequency transform coefficients of the passing sub-blocks are grouped together as one burst and given priority 1. Furthermore, the high-frequency component!: Even if there is, the quantization unit 4.1 The high-frequency transform coefficients of multiple sub-blocks are sequentially quantized, and the high-frequency transform coefficients of multiple sub-blocks are combined into one burst as a burst belonging to the second priority in the bursting circuit 34, and the burst is given priority 2. After this, the burst is transmitted by burst multiplex transmission. Route 37 (- performs hierarchical transmission of the sending video signal.

Those with priority l are always sent to the receiving side, and a certain level of image quality is always guaranteed. Another priority ingredient! =, it is possible to transmit and receive high quality video signals as long as the burst multiplex transmission line 37 has a margin.

FIG. 4 shows a modification of the third embodiment of the invention.

This is shown in the case where two-stage hierarchization is performed.

In the same manner as described above, the input video signal 23 is divided into sub-blocks in the NXN sub-blocking frame memory part 41, and then orthogonally transformed in a two-dimensional head area in a two-dimensional orthogonal transform circuit 42 for each sub-block. .

The generated orthogonal transform coefficients are first encoded in the quantization unit 45 using, for example, vector quantization (for example, 6 discrete cosine transform vector quantization (DCT-VQ)). Television Society Journal vol. 1.3
9. NG, 10 (1985)). This generated portion is given a priority l by the bursting circuit 27 and is burst-formed. Furthermore, in order to enable high-quality transmission, the component that is transmitted by 1 at the tip order 1 is subtracted from the orthogonal transform coefficient in the two-dimensional head area by the subtraction circuit 46, and the residual component C2 is A quantization unit 47 performs quantization. This component is given priority 2 in the bursting circuit 34 as a component that further improves the image quality by C2, and a burst is formed. priority l
(: Guarantees that the corresponding components are always in the second layer of receiver C,
Images with a quality that meets a certain standard are always provided.

Moreover, the priority level 2 component makes it possible to obtain a high-quality image as long as there is a margin in the transmission path 37. By layering using orthogonal transformation in this way, it is possible to use a redundancy compression method suitable for each layer, and at the same time, by giving consecutive priorities, the effect of quality deterioration is suppressed and burst transmission is performed. can be carried out efficiently.

The fifth factor represents the fourth embodiment of the present invention. The input video signal 23 is subtracted by the predicted value from the interframe prediction circuit 49 in a subtraction circuit 48, and the residual signal is subjected to two-dimensional orthogonal transformation in an orthogonal transformation circuit 42. The converted coefficients are hierarchically divided into a generated component and a remaining component by a generated component quantization section 51 and a remaining component quantization section 52, and the generated components are further quantized, and then sent to a bursting circuit 27 in order of priority l.
is given and becomes a burst. This quantized product is dequantized by the local decoder 53 and supplied to the interframe prediction circuit 49 to perform interframe prediction. Coefficients other than the generated coefficients are quantized by the quantization unit 52 using the generated coefficients, and are further given priority 2 and burst-coded by the bursting circuit 34. These layered and bursted signals are sent to a burst multiplex transmission line 37. Those with priority level 1 are always sent to the receiving side, and it is guaranteed that a certain level of image quality will be transmitted. Also priority 2
The images are sent to the receiving side as long as there is room on the transmission path 37, making it possible to transmit images of higher quality. In this way, by orthogonalizing the residual signal, layering it, and assigning an appropriate priority to each layer, it is possible to efficiently perform burst transmission while suppressing the effects of quality deterioration. Become.

FIG. 6 shows a fifth embodiment of the invention. Video signal 23
is divided into N x N sub-blocks by the frame memory part 41, and the sub-blocks are divided into 2 by the orthogonal transform circuit 42.
Orthogonal transformation is performed in the dimensional domain. The transformed orthogonal transform coefficients are generally divided into a low frequency component, which is a generated component, and a remaining high frequency component. Of these, the generated portion is transmitted with priority 1 and always reaches the recipient. Therefore, it is possible to perform interframe prediction only on this generated component. Therefore, the subtraction circuit 48. The inter-frame prediction circuit 49, the q-coding unit 51, and the local decoder 53 perform inter-frame prediction using only the generated components, quantize the prediction error component, give priority l to the bursting circuit 27, convert it into bursts, and send it out. . The remaining two components C are adaptively quantized using the generated components in the quantization unit 52, are given priority 2 in the bursting circuit 34, are burst-formed, and are sent to the transmission path. The generated portion is a burst that always reaches the receiving end, and it guarantees a certain level of image quality to the receiving end. Furthermore, as long as there is a margin in the transmission path 37, the burst with priority level 2 will reach the receiving side C2, making it possible to transmit high-quality images. In this way, hierarchization is performed using orthogonalization, and predictions are made that are appropriate for each layer in terms of the spread of quality deterioration.

At the same time, by assigning appropriate priority, it becomes possible to efficiently perform a bursting circuit that suppresses the influence of quality deterioration.

Example of band division FIG. 7 shows a sixth embodiment of the present invention. The video signal 23 is decomposed into a plurality of band signals by temporal or time-spatial band pass filters 61 to 64, and here, as an example, a case where the video signal 23 is decomposed into four signals is shown.

This filtering is performed in the two-time direction or in the time-space direction.

The filter 61 outputs the lowest frequency component, and the filter 64 outputs the highest frequency component. Next, the number signal is thinned out in the time direction or the time-space direction depending on the direction of filtering by thinners 65 to 68. Each output signal of the decimators 65 to 68 is subjected to redundancy compression by prediction, conversion, etc. in encoding units 71 to 74, and then quantized. Here, the high frequency components are as described below (:
Since it is given a low priority, there is a high possibility that data will be lost. For this reason, when predictive coding is applied, the prediction gain is made small so that the effects of omissions are quickly attenuated.

Each encoded output of the encoders 71 to 74 is sent to a burst generator 75.
It is given priority and bursted.

Here, priority is given from the following two viewpoints.

a. The frequency characteristics of vision have temporal and spatial band-pass characteristics. Therefore, the contribution of each frequency component to the quality is different, and even if high frequency components are missing, it is subjective (:
No major quality deterioration occurs. Therefore, high frequency components (
: gives lower priority.

b. Frequency components with large energy will cause large distortion if they are missing. Therefore, component C with large energy is given high priority.

The decoding unit 76 decodes the encoded data from the encoding units 71 to 74 into a signal when there is no data loss and a signal when data loss is assumed. Quality control section 15
Now, the signal decoded by the decoder 76 and the input video signal 23
The quantization characteristics of the encoding units 71 to 74 are controlled by comparing the quality with the quantization characteristics of the encoders 71 to 74 (priority assignment of the Niburst converting unit 75 is controlled).

The dotted line shows the control.

By assigning priority and transmitting in this way,
It is possible to visually reduce the impact of data loss with less deterioration in quality. Also, the effect of missing data: distortion caused by two is filtered and smoothed, block-like distortion when using orthogonal transformation, and pre-Wl encoding? There is no sudden C2 occurrence when there is a distortion like when using it. Moreover, the distortion is attenuated by the time constant of the filter. Therefore, the distortion becomes less noticeable. In addition, it is possible to control the degree of data loss by controlling the priority assignment.

FIG. 8 shows a seventh embodiment of the invention. The video signal or the conversion coefficient signal 77 obtained when the video signal is converted is processed through quantization units 81 to 81 which perform quantization and redundancy compression provided in series.
At 84, it becomes a step-by-step (: @ child.

Here, a case where four-stage quantization is performed is shown as an example, and each quantized residual signal 85 to 87 of quantization units 81 to 83 is shown.
are input signals to the quantizers 82 to 84, respectively. The quantization results 91 to 94 of the quantization units 81 to 84 approximate the human input signal 77 in stages.

The quantization results 91 to 94 are given higher priority as the quantization data of the quantization unit is closer to the input signal 77, which has a greater influence on the degree of quantization, and is burst-formed by the burst-formation unit 95.

In this example, the quantization results show data 91.92.93.
A higher priority is assigned in the order of 94. The decoding unit 96 decodes a signal when there is no data loss and a signal when data loss is assumed from the data 91 to 94. The quality control unit 15 compares the quality of the decoded signal and the input signal 77 and controls the priority assignment. The dotted line indicates this.

By assigning priority and transmitting in this way,
The effect of missing data can be suppressed by small-amplitude objects C2, and distortion can be reduced to random noise C2. Furthermore, since the quantized data given a high priority need hardly be considered for the influence of deletion, prediction and run-length encoding 1: Therefore, high redundancy compression can be performed. In this example, the quantization unit 81 compresses the redundancy of the quantized data using prediction, for example. Furthermore, it is possible to easily control the degree of data loss by controlling the priority assignment.

FIG. 9 shows an eighth embodiment of the invention. The input video signal 23 is thinned out by thinning devices 96 to 98 and decomposed into several sample data. This example shows the case where three sample data (-) are decomposed. An example of the sampling pattern is shown in FIG. 96a, 97a, 98a) do not overlap, and the original signal can be reproduced by overlapping these samples. Also, sample data 9
7a, even if this data is missing, it can be interpolated with less distortion using sample data 96a and 98a on both sides. Here sample data 98a*97a
, 96a are data of the 1st, 2nd and 3rd hierarchy, respectively.

These sample data are encoded in order from the lowest layer (=redundancy compressed).The first layer data 98a is predictively encoded in the prediction encoding unit 101.For example, the prediction of the difference between frames Next, the second layer data 97a is predictively encoded in the encoding section 103 using the first layer data decoded by the decoding section 102.

Further, the third hierarchical data 96a is stored in the decoding units 102 and 104.
The encoder 105 performs predictive encoding using the first and second layer data respectively decoded in .

In the bursting unit 95, the lower the data in one hierarchy, the higher the priority is given to the data, and the data is bursted. Here, the highest priority is given to the hierarchical data 98a, and the lowest priority is given to the first hierarchical data 96a. In the decoding section 106, the encoding section 1
From each encoded data of 01, 102, and 105, a signal with no data loss and a signal with data loss assumed are decoded. Quality system? 5 section 15 compares the quality of the decoded signal and the input signal 23, and then outputs the bursting section 95.
control the prioritization of The dotted line indicates this.

From f2, assigning priority and transmitting in this way,
Even if low-priority data is missing, the impact can be kept local. In addition, the interpolation allows video to be played back with less distortion, and subjectively, quality deterioration can be kept to a small level.

Furthermore, it is possible to easily control the degree of data loss by controlling the priority assignment.

Configuration Example Using Motion Compensation) FIG. 11 shows a ninth embodiment of the present invention, which shows a system combined with an interframe predictive coding system using orthogonal transform. This example is an example in which a motion compensation circuit is combined with an interframe predictive coding method using orthogonal transform, and uses two-stage hierarchization, that is, method I shown in FIG. 5: Addition of motion compensation. This is the configuration.

First, motion detection is performed in the motion detection circuit 111 using the input video signal and the locally decoded signal, and this is used as a sub-input to the predictor. Motion compensation is performed on the output. The prediction after this motion compensation is subtracted from the input video signal 23 by a subtraction circuit 48, and the residual signal is divided into subblocks.

Furthermore, a two-dimensional orthogonal transformation is performed using DCT or the like, and the quantization section is used to hierarchize the resulting component and the remaining components according to the size C' of the component in the transform output, and to quantize the component. It will be held at 51.52. The quantized output of the generated portion is locally decoded and input to the motion detection circuit 112. In addition, the burst forming unit 113 forms a burst with priority 1 from the generated orthogonal transform coefficients and the motion compensation vector, and the remaining components (- are given priority 2 to form a burst).

A certain level of image quality is guaranteed to receiver 1 by the generated component and the motion compensation vector. Furthermore, as long as there is room on the transmission path, bursts with priority level 2 are transmitted to the receiver, making it possible to transmit high-quality images. In this way, even in a method using motion compensation, layering is used, and by giving appropriate priority to the motion vector and the encoded output of each layer, burst transmission can be efficiently performed while suppressing the effects of quality deterioration. becomes possible.

"Effect of the invention" As explained above, according to the present invention, multi-stage prediction is possible.

Using methods such as , orthogonal transformation, and band division, the data to be transmitted is divided into layers with different quality contributions.

Appropriate priorities should be given to the quality contribution (elimination) and bursting should be performed, and the quality contribution of these layers (i.e., appropriate priority should be given to the By applying redundancy suppression techniques.

(1) Transmission errors and traffic conditions (: Therefore, it is possible to control quality deterioration due to burst loss that occurs within the network, and to realize a burst communication system that minimizes the deterioration.

(2) In order to control the quality deterioration mentioned above, there is no need to form a feedback system between the clear information source and the network, such as layering and prioritization in the clear information source, and burst processing according to the priority (:) in the network. (: Each can be realized independently.

this! Second, a more efficient burst communication method with higher real-time performance can be realized.

[Brief explanation of the drawing]

The first factor is a block diagram showing a first embodiment of this invention, FIG. 2 is a block diagram showing a second embodiment of this invention using multi-stage prediction, and FIG. 3 is a block diagram showing this invention using orthogonal transformation. the third
FIG. 4 is a block diagram showing an example of the third embodiment.
- Figure 5 is a block diagram showing a modified example of layering in
FIG. 6 is a block diagram showing a fifth embodiment of the present invention in which inter-frame prediction is performed after orthogonal transformation, and FIG. 7 is a block diagram showing a sixth embodiment of the present invention using band division. A block diagram showing an example, FIG. 8 is a block diagram showing a seventh embodiment of the present invention using stepwise quantization for layering, and FIG. 9 is a block diagram showing a seventh embodiment of the present invention using layering by sampling pattern. A block diagram showing the eighth embodiment, FIG. io is a diagram showing an example of sample points in each decimator in FIG.
The figure is a block diagram showing a ninth embodiment of the present invention to which motion compensation is applied. Patent Applicant: Nippon Telegraph and Telephone Corporation Agent Taku Kusano / +71 K.O. 2 K.Y.

Claims (9)

[Claims] (1) In a burst communication method in which a multidimensional information signal that changes over time is periodically or aperiodically bursted or packetized and transmitted according to its occurrence, the above information signal contributes to the encoding quality. has a hierarchization means for outputting encoded data hierarchized in accordance with the degree of deterioration, and redundancy suppression processing having different degrees of influence on quality deterioration when data is missing is applied to the hierarchized data, and A layered burst communication system that includes means for assigning priority to layered encoded data according to its quality contribution, forming bursts or packets, and transmitting the data. (2) The layering means subtracts the prediction signal generated by the first stage prediction circuit from the input video signal,
A circuit that obtains the first layered encoded data from the residual signal, and a predicted value generated by a second stage prediction circuit from the signal components that could not be encoded in the first layer. 2. The layered burst communication system according to claim 1, further comprising a circuit for obtaining second layered encoded data from the residual signal. (3) The layering means divides a video signal spanning multiple frames into blocks containing two or more pixels, performs orthogonal transformation for each block, and converts the orthogonal transformation coefficients according to the size of the components. 2. The layered burst communication system according to claim 1, further comprising means for classifying into a plurality of layers. (4) The layering means subtracts the prediction signal generated by the interframe prediction circuit from the input video signal,
It has means for orthogonally transforming the residual signal and classifying the orthogonal transform coefficients into a plurality of hierarchies according to the size of the component,
The layered burst communication system according to claim 1, further comprising a prediction circuit that generates an interframe prediction signal from the encoded data of the layered generation. (5) The layering means divides a video signal spanning multiple frames into blocks each including two or more pixels, performs orthogonal transformation for each block, and converts the orthogonal transformation coefficients according to the size of the components. Claim 1 characterized in that it has means for classifying into a plurality of hierarchies, and a circuit for subtracting predicted values generated by an interframe predictor for each of the hierarchies to obtain encoded data. Layered burst communication method. (6) The layering means decomposes the video signal, which is a three-dimensional time-spatial signal, into a plurality of band-pass signals using a filter having band-pass characteristics in time or time-spatial terms, and Treat the signal as layered data, giving higher priority to low-frequency components and components with large energy that make a large contribution to subjective quality, and to high-frequency components that make a small contribution and components with small energy. are given a low priority, redundancy compression technology that provides high efficiency is applied to the components given a high priority, and the ripple effect of quality deterioration due to data loss is applied to the components given a low priority. 2. The layered burst communication system according to claim 1, wherein the layered burst communication system is a means for obtaining a layered encoded output by applying a redundancy compression technique with a small degree of redundancy. (7) The layering means is a means for quantizing a video signal, a prediction residual signal, or a transform coefficient, and quantization units are connected in cascade in multiple stages so as to improve the quantization accuracy step by step;
It is a means to obtain hierarchical data of the quantized data of each quantizer, and it gives a higher priority to the quantized data of the first stage quantizer, which greatly affects the quantization accuracy. Apply a redundancy compression technology that provides high efficiency to the assigned quantized data, and apply a redundancy compression technology that reduces the impact of quality deterioration when data is missing to the quantized data that is assigned a low priority. A hierarchical burst communication system according to claim 1, characterized in that: (8) The above-mentioned layering means ensures that the sample points of the video signal do not overlap with each other, and by overlapping them, the original signal is restored, and even if the entire sample data of some of them is missing, other The signal is decomposed into N sets of sample data by sampling N sets of sample data so that they can be interpolated with less distortion, and each sample data is given the highest priority as hierarchical data. The first layer performs efficient redundancy compression using the first layer data of frames that have already been transmitted, and the n-th layer (1<n≦N
) layer is a means to perform redundancy compression using data from the n-1th layer to the first layer, and is characterized by giving higher priority to data in the lower numbered layer to form a burst. A hierarchical burst communication system according to claim 1. (9) A method of transmitting video signals using motion compensation vectors, in which a motion compensation vector is detected from the input video signal and hierarchized data with high priority, and based on this motion compensation vector, motion components are 2. The hierarchical burst communication system according to claim 1, further comprising means for calculating a residual component by removing .