JPH046931A - Packet transmission system - Google Patents

Abstract

PURPOSE:To enhance the compensation capability at packet abort by applying packet processing to each picture information of a block. CONSTITUTION:When picture information of a block is inputted and stored in a memory 1b and a packet in the not consecutive input waiting state exists, a processor 1a stores the picture information to a packet awaiting the input. On the other hand, when the input waiting is consecutive, the picture information is stored in a new packet. When the processor 1a discriminates it that the picture information to be stored is stored entirely in the packet and no room exists in the information storage section, the packet is outputted to an ATM communication network 4. When all picture information of one block reach, the picture information of the block is outputted to a moving picture decoding section 3. According to this system, since the packet processing is executed to each picture information of the block and the information is continuously sent in the unit of blocks, number of blocks effected by packet abort is suppressed to one or two blocks.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention disassembles variable-length image information, assembles and transmits fixed-length pieces, and decomposes the fixed-length pieces on the receiving side. to assemble a variable length image signal) (related to the Soto transmission system).

[Conventional technology] Asynchronous transfer mode (ATM) is scheduled to be adopted in high-speed broadband integrated communication networks, which are being standardized.
The TM communication network accommodates a wide variety of information such as images, audio, data, control/management signals, etc. in fixed-length packets called cells, and transmits information at high speed and flexibly responds to temporal changes in the amount of information. This makes it possible to transmit data in accordance with the following. ATM communication networks are optimal for image information that is higher in quality than current television signals because it has the characteristics of high speed and time-varying video rates.

By the way, in the encoding method of image information, one image (
frame) into multiple blocks, and for each block,
Processing such as predictive encoding (for example, intraframe DPCM), transform encoding (for example, discrete cosine transform), unification, entropy encoding (run-length Huffman encoding), etc. is performed sequentially. As a result of information compression through such encoding,
The amount of encoded information differs from block to block.

Therefore, in order to transmit image information through an ATM communication network, the interface section disassembles variable length image information and assembles fixed length packets, and also divides the fixed length packets and assembles variable length image information. processing is required.

Further, if the amount of information concentrated per unit time in a communication network switch or multiplexing device exceeds the processing capacity of the device, packets are discarded.

Conventionally, as methods for converting between variable-length image information and fixed-length packets (packet assembly method and packet disassembly method) that take packet discard compensation into consideration, the following documents have been proposed.

Ryozo Kishimoto, Nari Irie, Yasuyuki Okumura, rHDTV Digital Transmission System Co., Television Society Technical Report, Vol.
, 13. No. 51, pp19-24 Although this document describes the case of an adaptive subband encoding method for HD and TV signal transmission, it can also be applied to the case where image information is encoded block by block. , the versatility is not lost even when this document is used for comparison.

Next, the packet assembly method (image information decomposition method) proposed in this document will be briefly explained.

First, encoded image information for each block is divided into fixed lengths without considering boundaries between blocks and blocks. next,
As a packet discard compensation method, a random error correction code is added to the divided fixed-length image information, and then, as shown in Figure 2, byte-by-byte interleave conversion is performed between multiple packets and sotos, and the information storage part of the packet is to be accommodated.

FIG. 2A shows an example in which the packet length is 40 bytes and the interleaving period is 20 packets before interleaving. FIG. 2('B) shows a packet after interleaving, in which two bytes are collected from each packet before conversion to form one packet.

A packet in which the error correction code has been added and interleaved information is inserted into the information storage section is output to the communication network.

When disassembling packets and assembling image information, first,
As shown in FIG. 2C, received packets are deinterleaved, and then bit errors are detected and corrected based on random error correction codes to assemble variable length image information. At this time, for example, as shown by "x" in FIG. 2(B),
When the first packet is discarded, the first two bytes of each packet after the deinterleaving process become errors as shown in FIG. 2(C) through the deinterleaving process.

This error can be corrected if it is within the capabilities of the error correction code.

In this way, compensation for packet discard is taken into consideration.

"Problems to be Solved by the Invention" However, this conventional method mainly has the following four problems.

(1) Without considering the boundaries between blocks, auxiliary information (
Block position information, sequence number error correction code, etc.) and main image information (prediction error, e.g. orthogonal transform coefficients such as discrete cosine transform coefficients, motion vectors, etc.) are divided into fixed lengths and interleaved transform is performed. Therefore, packets that should not be discarded and packets that are relatively resistant to discarding occur accidentally, and discard priority control cannot be applied.

(2) The ability to compensate for packet discards is determined by the length of the random error correction code provided and the packet period at which interleaving conversion is performed. If information is lost, all packets belonging to the period of interleaving conversion may not be available for assembling image information.

In particular, as the interleaving period becomes longer, the possibility of uncompensable discards decreases; however, once consecutive discards occur, the number of packets for which discards cannot be compensated increases.

(3) It is known that image information has considerable redundancy, and depending on the movement of the image, there may be no change in the central main information (prediction error information or motion vector) within the block. Whether the amount of information can be compressed by not sending it or not, this information compression is not considered.

Conventionally, methods have been proposed that only transmit the fact that the block is an invalid block, or no compensation for discarding the transmission of an invalid block has been considered.

(4) To open up the ability to compensate for packet discards,
Increasing the packet period for performing interleaving conversion leads to an increase in delay time, and increasing the number of error correction codes reduces transmission efficiency.

The present invention has been made in consideration of the above points, and is intended to provide a packet transmission system with a higher compensation ability when packets are discarded.

[Means for Solving the Problem] In order to solve the problem, in the present invention, one image is divided into a plurality of blocks, and variable length encoded image information is decomposed into fixed length coded image information. In a packet transmission method in which packets are assembled and output to a communication channel, and fixed-length packets from the communication channel are disassembled to assemble variable-length image information, the image information of blocks is arranged by constituent elements with different meanings of the information, and the blocks are divided into blocks. We decided to perform packetization processing on a unit-by-packet basis.

In such processing of image information units of blocks, when part or all of the image information of the block to be processed can be stored in the information storage section of a packet that already contains all or part of the image information of other blocks. In addition, it is desirable that the packet contains image information of the block to be processed.

In addition, when the image information of a block to be accommodated in a packet is larger than the available capacity of the information storage section of the packet and is accommodated in multiple packets, the boundaries between the constituent elements are made clear, and the information of the same constituent elements is It is preferable to accommodate all the information in the same packet to prevent the information of the same configuration T from being divided into two and distributed to different packets.

Furthermore, when assembling a packet from image information including block validity/invalidity information, it is preferable that the number of invalid blocks preceding the block to be packetized is also included in the packet.

Even if two or more blocks of image information are stored in one packet, it is preferable not to store two blocks of information that are consecutive on the image in one packet. If there is a margin in the block and the image information of another block is to be waited for, and the waiting time exceeds a predetermined time, it is preferable to transmit the packet while there is margin.

[Operation] The concept of block is an IR concept in the encoding system, but in the present invention, it is also adopted in packetization to execute packetization processing in block units, and the number of blocks affected by packet discard is reduced. expected to reduce the amount.

Even if the packetization process is performed for each block of image information, if there is room in the packet, it is preferable to insert multiple blocks of image information to improve transmission efficiency.

When the image information of a block is distributed into two or more packets, the packet discard resistance of each component can be made equal by distributing it as a unit of component. When information of the same component is divided into two and distributed to different packets,
It is twice as likely to be affected by packet discard than other components.

When performing invalid block/valid block determination processing to convert encoded image information into buckets, regardless of whether invalidity is inserted into the packet or not, the block before the block to be processed is It is preferable to include the number of invalid blocks in the image information of the block to be processed and packetize it to make it easier to determine invalid blocks.

Even if two or more blocks of image information are stored in one packet, if two consecutive blocks of information on the image are not stored in one packet, there will be scattered blocks with missing information as packets are discarded, causing them to be discarded. It becomes easier to process.

To prevent delays in packetization processing, when waiting for image information from other blocks, set a permissible time, and when the waiting time exceeds the set time, the packet can be transmitted while there is a margin. preferable.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 3 shows a device configuration related to a packet transmission method according to an embodiment.

In Figure 3, packet assembly/disassembly device (PAD) 1
A moving image encoding section 2 and a moving image decoding section 3 are provided in association with each other. The packet assembly/disassembly device 1 assembles fixed length packets shown in FIG. 1, which will be described later, from variable length image information provided by the video encoding unit 2 and auxiliary information generated internally as a derivative thereof. , for example, ATM
It is transmitted to the communication network 4. In addition, packet assembly/disassembly equipment 1
The fixed length packet given from the ATM communication w44 is decomposed as shown in FIG.

The packet assembly/disassembly device 1 executes a conversion process between variable-length image information and fixed-length packets using software, and is composed of a processor 1a and a memory 1b. Although not shown in the figure, the video encoding unit 2
, a moving image decoding section 3, and an input/output section for communicating with the ATM communication network 4.

Further, the processor 1a has a built-in timer counter (not shown) for monitoring a predetermined time allowed to form one packet.

In the case of this embodiment, the moving image encoding unit 2 converts human image information into Y component (luminance component), U component (first color difference component),
For each component, one image is divided into multiple blocks, and motion-compensated interframe predictive coding or intraframe predictive coding is adaptively performed for each block. , it is assumed that orthogonal transform (for example, discrete cosine transform) and entropy encoding are performed on the obtained prediction error information.

Further, the video encoding unit 2 detects whether a block is valid or invalid. Here, an invalid block is an orthogonal transform coefficient of a prediction error, a motion vector in interframe predictive coding, or an average value in intraframe predictive coding (
This is a block in which the subtraction reference value) is all 0. By not transmitting detailed information on invalid blocks, it is possible to compress the amount of image information when there is little movement in the image.

Therefore, from the video encoding unit 2 to the packet assembly/disassembly device 1
The types of image information given to the block include orthogonal transform coefficients (more precisely, entropy-encoded coefficients) of the prediction error of each Y, U, and V component, and the encoding method selected for the block by adaptive control. A coding method identifier that identifies whether the is interframe predictive coding or intraframe predictive coding, and a motion vector when interframe predictive coding is selected or a motion vector when intraframe predictive coding is selected. There is an average value of pixel values within a block as a subtraction reference value, block validity/invalidity information, etc.

The packet assembly/disassembly device 1 assembles packets not only from the image information provided from the video encoding unit 2, but also by using auxiliary information derived from the image information such as sequence numbers, error correction codes, and block position information. Assemble.

FIG. 4 shows the packet format according to this embodiment.

The packet 10 is composed of one ATM header section and one information field section 20.

In the ATM header section 1, information such as a personal path identifier, virtual cell identifier, header error checker, payload type, etc. is inserted as in the conventional case.

The information field section 20 is further divided into an adaptation header section (AAL header section) 21 and an information storage section 22 .

The adaptation part 21 indicates whether the frame number, sequence number, error correction code in the information field part 20, and image information in the information storage part 22 are composed of only that packet or continue with the next packet. An identifier is inserted.

Information of each block is stored in the information storage unit 22, and one piece of block information 30 is stored in the information storage unit 22 consecutively, grouped into constituent elements having different meanings of information. FIG. 4 shows that the information 30 of one block is 1
The arrangement of the components when fully contained within the packet (10) is also shown. FIG. 5 shows the arrangement of components when one block of information 30 spans two packets. Note that when the above-mentioned encoding method is adopted, in reality, the average of the variable information amount of each flock is approximately equal to the length of the information storage section 22, and the information 30 of one block is It is often accommodated in two packets.

The arrangement order of each component of one piece of block information 30 is basically as shown in FIG. 4. That is, the position information 30a of the relevant block, the number of invalid blocks before the relevant block 30b, the identifier 30c that specifies the encoding method applied to the relevant block, and the orthogonal transform coefficients 30d of prediction errors of Y, U, and V components. , a motion vector or intra-block average value 30e, and a specific pattern (EOB) 30f instructing the end of one block information 30, in this order.

Note that the orthogonal transformation coefficient 30d is the Y component coefficient 3OdY,
The I component coefficient is 30 dU, the V component coefficient is 30 dV, and there is a bit pattern ESB at the boundary that makes the boundary clear.
is inserted. Hereinafter, the coefficients of each component themselves will also be referred to as constituent elements.

When one piece of block information 30 is stored in the information storage section 22 of two packets, the information storage section of the second packet that stores the latter half of the information contains the position information 30a of the block. , the number of invalid blocks 30b before the block and an identifier 30c for specifying the encoding method applied to the block are stored even if they overlap with the first packet containing the first half of the information. Furthermore, information on components that cannot be accommodated is accommodated in the information accommodating section of the second packet from the beginning.

Therefore, for example, as shown in FIG. 5A, the orthogonal transform coefficient 3 of the U component is
When 0dU can no longer be accommodated, instead of accommodating the information from the middle in the information storage section 222 of the second packet, as shown in FIG. 5(B), the position information 30a of the block, Number of invalid blocks before 30
b and the identifier 30c that specifies the encoding method applied to the block are repeatedly stored in the information storage unit 222 of the second packet, and then the orthogonal transform coefficient 30dU of the U component is stored.
is accommodated from the beginning.

Next, a process of disassembling image information and assembling such a packet will be described according to the process flowchart of FIG.

Regarding the packet assembly process, FIG. 1 shows the process for image information of ]- blocks. This illustration was adopted because the feature of this embodiment is the packetization processing in units of blocks. The block concept is a concept adopted in the image signal encoding method, and is not a concept for packetization, but the block processing concept in the encoding method is directly applied to packetization. This is due to consideration of packet discard resistance.

When the image information of a block is input and stored in the memory 1b, the processor 1a first determines whether there is a packet with room (empty) in the information storage section waiting for input of the image information ( Steps 101.102). Note that, as described later, there may be a plurality of packets waiting for input.

If there is no packet waiting for input, the image information is stored in a new packet (step 104).

On the other hand, if there is a packet waiting for input, it is determined whether the position on the image of the block related to the image information contained in the input waiting packet is continuous with the position of the block related to the input image information. If there are packets waiting for input that are not consecutive, image information is stored in the packets waiting for input (step 103.
105).

On the other hand, if the image information is continuous, the image information is stored in a new packet (step 104).

If the process moves from step 103 to step 104 in this way, there will be a plurality of packets waiting for input.

The reason why consecutive blocks of image information are not stored in the same packet is to intersperse blocks where information is missing due to packet discards, so that discard compensation can function effectively and deterioration in playback image quality can be kept to a minimum. This is because he tried to do so.

When the accommodating process for a new packet (step 104) or the accommodating process for a packet waiting for input (step 105) is completed, it is determined whether or not some of the image information of the block cannot be accommodated in the information accommodating section of the packet. is determined (step 106).

If there is information that cannot be accommodated, the packet that has been accommodated is output to the ATM communication w44 by excluding the information of the component that was partially accommodated, and the head of the component that could not be accommodated is output. is detected, the process returns to the accommodating process for the new packet (step 104), and the new packet is accommodated again starting from the detected head information (step 1).
07.108).

At this time, the position information of the block, the number of invalid blocks before the block, and an identifier that specifies the encoding method applied to the block are always inserted. In this way, when dividing one block of information into two packets, an identifier identifying the block position information, the number of invalid blocks before the block, and the encoding method applied to the block is inserted into both packets. This is done so that the information on the components inserted in the packet can be effectively used without requiring reception of the other packet as a condition.

When the processor 1a determines in step 106 that all of the image information to be stored cannot be accommodated in the packet, the processor 1a determines whether there is enough room in the information storage unit (step 109). If there is no margin, the packet is output to the ATM communication network 4 (step 110).

On the other hand, if there is room in the information storage unit, the process for the image information of the block currently being processed is temporarily finished, the packet is placed in a packet waiting for input, and the image information of the subsequent block is input. (Step 101)
(step 111).

At this time, use the built-in time counter to judge whether or not the waiting time exceeds the set time.
, if the set time is exceeded, the bucket t~ is output to the ATM communication network 4 even if there is room in the information storage unit. This determination of the elapse of the set time is to prevent delays in packetization.

Next, the process of disassembling received packets and assembling image information will be described with reference to FIG. In addition, Figure 1 shows
Regarding packet disassembly processing, processing for one packet is shown.

When a packet is input, it is detected whether or not it has been discarded (steps 121 and 122). If the information has been discarded, compensation processing for the discarded information is performed (step 123).

Next, it is detected whether all the image information of one block has arrived, and if it has arrived, the image information of that block is output to the moving image decoding unit 3 (steps 124 and 125).
. If all the image information for one block has not arrived, the process enters a state of waiting for input of the next packet (step 12).
6). At this time, monitor whether or not a certain set time is exceeded,
If it exceeds the block, compensation processing for the unarrived information is performed, and the image information of the block is output to the moving image decoding unit 3 (steps 127 and 128).

Note that the position of the image block for image reproduction can be determined directly from the position information 30a of the image block. The occurrence of packet discard is detected from the missing number in the sequence number. The position of the discarded image information block is determined from the position information 30a of the block of the packet that has not been discarded, when the information of the block spans a plurality of packets.

When the image information of the discarded block is included in only one packet, the position of the discarded packet is then determined from the number 30b of invalid blocks included before the block in the auxiliary information of the received packet.

Therefore, according to the above embodiment, the following effects can be expected.

(1) Since packetization processing is executed for each block of image information and the block is continuously transmitted in units of blocks, the number of blocks affected by packet discard can be kept to 1 or 2 blocks.

Incidentally, in consideration of transmission efficiency, it has been proposed to group multiple blocks and perform packetization processing on the group's image information as a unit. big.

(2) When the image information of a certain block spans multiple packets, the boundaries between the components of the image information are made clear, and the main information of the block (orthogonal transform coefficients, motion vectors, etc.)
Since none of the packets span two packets, the packet discard resistance can be made equal for each packet.

(3) By packetizing the number of invalid blocks before the block to be processed as auxiliary information, the amount of information can be compressed without reducing the discard compensation ability.

(4) Since it is possible to accommodate a plurality of block information in one packet, it is possible to minimize the reduction in efficiency due to bucketing in units of blocks.

(5) Since two consecutive blocks of information on the image are not accommodated in the same packet, it is possible to easily compensate for packet discard.

(6) If the information that should be included in the packet does not arrive even after the set time has elapsed, the information is transmitted even if there is room in the information storage section of the packet, so it is possible to prevent an extreme increase in delay time. .

Note that this device can be used as an interface section when a communication network that uses a transmission mode using variable-length packets is connected to a communication network that transmits fixed-length packets, such as an ATM communication network. The concept of the packet transmission system, which is the name of the invention, includes such cases.

Further, the encoding method of input image information is not limited to the embodiment described above. For example, it can also be applied when packetizing image information from an encoding unit that does not perform adaptive control, such as an encoding unit that performs only interframe encoding or an encoding unit that performs only intraframe encoding. I can do it. It can also be applied to encoding systems that vary quantization characteristics, and the quantization characteristics are also a component of image information obtained by employing such encoding systems.

A video encoding unit or a video decoding unit, and a packet assembly/decoding unit.
The transmission format of the image information to and from the decomposition device may be multiplex transmission of the constituent elements, or may not be multiplex transmission.

[Effects of the Invention] As described above, according to the present invention, since packetization processing is performed for each block of image information, a packet transmission method with a higher compensating ability when packets are discarded than in the past can be achieved. can be realized.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing packet assembly processing and disassembly processing according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of main processing of a conventional method, and FIG. 3 is a block diagram showing a device configuration according to the above embodiment. , FIG. 4 is an explanatory diagram of the packet format of the above embodiment, and FIG. 5 is an explanatory diagram of the method of accommodating information in the information accommodating section of the packet. DESCRIPTION OF SYMBOLS 1...Packet assembly/disassembly device, 1a...Processor, 1b...Memory, 2...Moving image encoding unit, 3.
...Moving image decoding unit, 4...ATM communication network, 22...
・Packet information storage section. Fig. 4 221 First shaft information storage section 221 First shaft of the device according to the embodiment ° Information storage section of Caret 2 Explanation diagram of separation method when spanning across Bakeat Figure 5

Claims (6)

[Claims] (1) One image is divided into multiple blocks, and variable-length encoded image information is encoded for each block, and fixed-length packets are assembled and output to a communication channel, and the fixed-length packets from the communication channel are A packet transmission method for disassembling and assembling variable-length image information, characterized in that blocks of image information are arranged by constituent elements having different meanings of information, and packetization processing is performed on a block-by-block basis. (2) Part or all of the image information of the block to be processed,
2. The method according to claim 1, wherein if the image information of the block to be processed can be accommodated in the information accommodating section of a packet that already accommodates all or part of the image information of the other block, the image information of the block to be processed is accommodated in the packet. Packet transmission method. (3) When the image information of a block to be accommodated in a packet is larger than the available capacity of the information storage section of the packet and is accommodated in multiple packets, the boundaries between the constituent elements are made clear, and the boundaries between the same constituent elements are 3. The packet transmission system according to claim 1, wherein all information is accommodated in the same packet to prevent information on the same component from being divided into two and distributed into different packets. (4) When assembling a packet from image information including block validity/invalidity information, the number of invalid blocks preceding the block to be packetized is also accommodated in the packet. Packet transmission method described in . (5) The packet transmission method according to any one of claims 1 to 4, characterized in that two blocks of information that are consecutive on the image are not accommodated in one packet. (6) Even if the image information of a block is accommodated, there is still room in the information storage section of the packet, and when the image information of another block is carried over, if the waiting time exceeds a predetermined time, 2. The packet transmission method according to claim 1, wherein a packet is transmitted.