JP4643988B2 - Video on demand server system and method - Google Patents

Description

  The present invention relates to a technique for efficiently providing video-on-demand while maintaining fixed channel bandwidth limitations.

  Currently, Video on Demand (VoD) systems that support viewers to watch desired programs (programs) when they want to watch are generally of a constant bit rate (CBR). Transmit video. Transmission of constant bit rate (CBR) video cannot achieve the same efficiency as transmission of variable bit rate (VBR) video for the equivalent long-term average bit rate. Transmitting multiple pre-recorded variable bit rate (VBR) video programs over a certain bandwidth channel means that the total instantaneous bit rate required by each program is This often causes problems because it can exceed the total effective bit rate. On the other hand, broadcast applications using real-time encoders are statistical in order to take advantage of instantaneous bit rate variations between multiple programs transmitted on the same carrier or transponder. Use multiplexing. The encoder operates in real time so that the system limits the total combined bit rate for all programs and does not exceed channel capacity beyond a predetermined time window. Each video program has an associated complexity indicator. The central controller dynamically adjusts the bit rate assigned to each video program based on the relative complexity.

  Currently, there exists a device known as a “video translator” that functions to reduce the bit rate of a video stream within a common compression standard such as MPEG-2. Current translators often create problems when converting between fixed and variable bit rates. This type of translator can also cause problems when trying to convert the bit rate into a bit stream with many scene changes or multiple I-pictures.

  Scalable video coding, which allows video (video) signals to be split into a base layer (base layer) and one or more enhancement layers (enhancement layer), is a bit rate Can be addressed. Video coding is scalable including spatial, SNR (signal to noise ratio), temporal, data partitioning, fine grain scalability (FGS), frequency scalability, etc. There are various methods. The MPEG-2 and MPEG-4 video compression standards include various scalability schemes. Using scalable coding requires that both the transmitter and receiver have equivalent capabilities to implement different coding algorithms, which adds additional complexity. Thus, there is a need for variable bit rate video-on-demand management technology that overcomes the problems of the prior art.

(Summary of Invention)
Briefly stated, the present invention provides a method for transmitting multiple pre-coded programs having different bit rates across a fixed bandwidth channel. For each program, at least two, preferably a plurality of different bit rate representations are generated. Control information is provided in each of a plurality of consecutive time windows for each bit rate representation. The control information provides a bit rate and quality measure during each time window. Control information is selected to maximize the quality of the selected representation during each time window of the bit rate representation for each program so that the total available channel capacity is not exceeded during that time window. Can do.

In disclosing the technology according to the present invention, the following technical terms are useful.
T is the time interval for which system optimization is intended.
C is the total effective channel capacity in the time window T.
P is the total number of programs.
pε (0, P−1) is an index of a specific program.
N [p] is the total number of representations of program p.
n [p] ε (0, N [p] −1) is an index of a specific representation of the program p.
r [p, x] is the bit rate of the representation x of program p at T.
q [p, x] is the quality of the representation x of program p at T.

FIG. 1 illustrates the present invention for transmitting precoded constant bit rate (CBR) and / or variable bit rate (VBR) audio / video programs over a fixed bandwidth channel 11 having full channel capacity C. 1 represents a first preferred embodiment of a video on demand (VoD) server basic configuration 10 according to FIG. The server basic configuration 10 includes a plurality of multi-rate stream generators 12 ₀ , 12 ₁ ,... 12 _P-1 . Here, P is an integer greater than zero corresponding to the number of independent input streams. The stream rate generators 12 _{0 to} 12 _P-1 pre-code one of programs corresponding to program 0, program 1... Program P-1 at a plurality of different bit rates. In other words, each multi-rate stream generator pre-codes a program that produces at least two corresponding different bit rate representations. A representation may include a particular stored coded bit stream, or a combination of various portions of different coded bit streams. Storage device 14 is generally a single disk drive, RAID in (Redundant Array of Inexpensive Discs) or more RAID format, precoded replicator presentation generated by the multi-rate streams generator ₁₂ 0 _{~12 P-1} It plays a role to remember.

Each of the multi-rate stream generators 12 _{0 to} 12 _P-1 generates control information for each time window T for each of the corresponding programs ₀ to _P-1 . The control information for each time window T for each representation of each program includes the bit rate of the representation, the peak signal-to-noise ratio (PSNR) of the representation, etc. Includes display of quality indicators. A central statistical multiplexer (stat mux) 16 receives control information for each representation. The stat mux 16 maintains a full bit rate at or below the full capacity C of the channel 11 and, for each successive time window T, for each program to maximize the quality of the transmitted program. Select a representation.

を満たす他の方法が存在する。一般に、特定のプログラムの再生が始まる時期に関しては、事前の知識がない。従って、最も低いビット・レート・リプリゼンテーションのビット・レート制限が、特定のプログラムの全ての時間ウィンドウＴについて同じになるように要求することにより、最も容易に容量制限を実施することができる。 The lowest bit rate representation for each program must not exceed the default value so that the sum of the lowest bit rates for all programs does not exceed the channel capacity for each interval T. . In this way, at least one representation of each program is transmitted via the channel 11. This limitation is achieved by requiring each program's lowest bit rate representation to have a peak bit rate below C / P. Limit for all time windows T of all programs

There are other ways to satisfy. In general, there is no prior knowledge as to when a particular program begins to play. Thus, capacity limitation can be most easily implemented by requiring that the lowest bit rate representation bit rate limit be the same for all time windows T of a particular program.

は全てのプログラムの全ての時間ウィンドウＴに対して満たされている状態でなければならない。ミニマックス（ｍｉｎｉｍａｘ）処理法は、それぞれのｐ∈（０，Ｐ−１）に対してｎ［ｐ］を選択することにより制限を満たすことができ、最低限の品質プログラムの品質を最高にする（即ち、最大歪曲を最小化する）ことができる。 The steps for maximizing the overall quality can be accomplished in various ways. In all cases, restrictions

Must be satisfied for all time windows T of all programs. The minimax processing method can satisfy the restriction by selecting n [p] for each pε (0, P−1), and maximizes the quality of the minimum quality program. (Ie, minimizing maximum distortion).

また、以下の方法における個々のプログラム品質の積を最適化することも可能である。

異なる視聴者に対して異なるクラスのサービスを提供するために、加重平均が使用される Minimax processing is performed by sorting the control information for each representation of each program, so that the quality and bit rate index increase monotonically as the index increases. The change in bit rate for each step (referred to as the “delta (Δ)” bit rate) is then stored. A statistical mux 16 calculates the total capacity S, starting with the lowest index for each program representation. The stat mux 16 selects the lowest quality program representation and sees if its delta (Δ) bit rate addition to S exceeds C. If the representation addition does not exceed C, the index for that representation is incremented and the procedure is repeated. Once C is exceeded, the next lowest quality representation is checked to see if it exceeds C when its delta bit rate is added to S. The procedure is repeated until the increment in any representation exceeds the channel capacity C. Alternatively, the overall quality of all programs can be optimized by maximizing the sum of the individual program qualities, including solving the following limited optimization problem.

It is also possible to optimize the product of individual program qualities in the following method.

Weighted average is used to provide different classes of service to different viewers

Multirate stream generators 12 _{0 to} 12 _P-1 can use various methods or combinations of methods to form multiple representations of each program. In all cases, the inner coding (I) for each representation at the beginning of each time window T so that a different representation is selected without variation for each time window T. Random access points such as frames must occur. Each representation uses either constant bit rate (CBR) or variable bit rate (VBR) coding as long as the capacity of the lowest bit rate representation meets the total channel capacity limit can do.

In the first method, the multi-rate video encoder encodes each program at various bit rates in several independent bit streams. Each different bit rate bit stream functions as a different representation. FIG. 2 shows a basic configuration 100 for implementing such a method. As shown in FIG. 2, the multi-rate video encoder 110 functions to encode a corresponding program and generates a plurality of different rate representations. Each of the independent transport packetizers 112 ₀ , 112 ₂ ... 112 _P-1 packetizes the corresponding representation. Before the multi-rate coding performed by multi-rate encoder 110 and the packetization performed by packetizers 112 ₀ -112 _P-1 are stored in storage device 140 that is structurally similar to storage device 14 of FIG. Once done, it doesn't have to be done in real time.

  Selection of a real-time output representation stored in the storage device 140 is made in response to a signal received from the selector block 150. Selector block 150 identifies the stored output representation in response to a specific representation request from statistical multiplexer (stat mux) 160. Upon receipt of the signal from selector block 150, storage device 140 provides the selected packetized representation to stat mux 160 for output.

  In the second method, parts of several different bit streams are combined to generate an additional representation of the program. By having a random access point in the same place as all the bitstreams of the program (as happens naturally with scene changes), or by using a fixed GOP (Group-Of-Pictures) structure The combination of compressed data from different bit rate streams at the access limit forms a new representation. As long as each representation can be generated from the stored data, it is not necessary to store each representation separately. Below, T is ½, the data is encoded at 30 fps, I frames are inserted every 15 frames, and three different bit rate bit streams (bit stream 0, bit Consider an example of generating stream 1 and bit stream 2). Each bit stream constitutes an independent representation. Additional representations using bit stream 0 for the first 15 frames and bit stream 1 for the second 15 frames and others may be formed.

  If a random access point with transport packets is deployed, the server base configuration 100 of FIG. 2 works well in advance when transport packetization occurs. FIG. 3 shows a VoD server base configuration 200 that is more suitable when random access points do not necessarily have to be placed with transport packets. Referring to FIG. 3, the VoD server basic configuration 200 includes a multi-rate video encoder 210 that generates a plurality of different bit rate representations for storage in a storage device 240 similar to the storage devices 14 and 140. Contains. The representation stored in the storage device 140 exists as a non-packetized bit stream.

  Selection of a representation for real time output stored in the storage device 240 occurs in response to a signal received from the selector block 250. Essentially similar to selector block 150 of FIG. 2, selector block 250 is for stored outputs in response to a specific representation request from static mixer / multiplexer 260. Identify the representation of. Upon receipt of the signal from selector block 250, storage device 240 provides packetized representations for selection to static mixer / multiplexer 260. The packetizer 212 packetizes the output stream of the static mixer / multiplexer 260 for output (not shown) on the channel.

In addition, the representation includes non-video codec (not shown) and the like including images that are not used as reference images for predicting other images such as MPEG-2 B-frames and JVT non-memory images. Also formed by switching between stored bit streams for reference images. Each non-reference image may be selected from a different stored bit stream that does not affect the quality of the subsequent coded image since the non-stored image is not used for prediction. The complete deletion of the non-reference image of the representation is also performed. Each non-reference image can be switched individually, or non-reference image groups can be switched together, realizing as many full bit rate representations as possible. The multi-rate stream signal generators 12 _{0 to} 12 _P-1 of FIG. 1 transmit to the stat mux 16 of FIG. 1 to select possible representations to select those with significant differences in quality and bit rate. Choose to limit the number of. Each representation need not be stored separately in the storage device. However, to facilitate the generation of representations when needed, a table showing a list of locations and the length of each coded frame is stored. It is also possible to store multiple bit rate bit streams for stored image only applications, or store only a single version of the reference image (I and P), greatly reducing storage capacity.

  For the video on demand (VoD) server basic configurations 100 and 200 of FIGS. 2 and 3, the statistical multiplexer (stat mux) 160 and the static mixer / multiplexer 260 are selected representations respectively. A bit stream corresponding to is generated. If the individual image is not an arranged transport packet, the basic configuration 200 of FIG. 3 performs transport packetization after generating a bit stream for a specific representation. On the other hand, the VoD server basic configuration 100 of FIG. 2 functions without any problem.

  In accordance with another aspect of the invention, scalable video coding is possible through the use of a scalable video encoder (not shown) to form the base layer and one or more enhancement layers. Any form of scalability that uses a base layer for motion, such as frequency scalability, FGS, SNR scalability, or temporal scalability, with reference picture prediction compensated to avoid variability can be used. The lowest bit rate representation corresponds to the base layer. The peak bit rate of the base layer must be such that the channel capacity C is not exceeded for each time window T when the base layer bit rates of all programs are summed. Other higher bit rate representations correspond to the base layer plus some portions of the enhancement layer. This method requires that the video decoder support the type of scalability that is used. The VoD server basic configuration 100 of FIG. 2 is used to encompass all enhancement layers in a particular representation. When a part of the enhancement layer is used in the presentation, the VoD server basic configuration 200 of FIG. 3 is used.

  In the VoD server basic configuration 100 of FIG. 2, encryption is added to the transport packetization process performed in non-real time, and all encrypted transport packets are stored and transmitted. For systems that use encryption to manage content rights, no real-time decryption or encryption is required. The VoD server basic configuration 200 of FIG. 3 requires encryption to be performed with real time transport packetization.

  In general, a VoD program includes audio and video. Since the audio bit rate is generally much lower than the video bit rate and is usually a fixed bit rate, the statistical multiplexers (stat muxes) 16 and 160 of FIGS. Fewer advantages are gained by selecting from different precoded audio bit streams using the static mixer / multiplexer 260. Users are more likely to notice voice quality switching while viewing a single program being delivered. Assuming that a single bit rate audio stream is also transmitted for each program, the total channel capacity C is considered to be completely consumed by the video channel. Alternatively, the above method can also be applied to audio, where multiple representations of audio programs having different bit rates and qualities are generated.

  A video player (not shown) that operates according to the basic configuration of the VoD server of the present invention includes a video decoder (not shown) and several storage devices for buffering. For a particular program received, the data reaches a non-uniform data rate, provided that the data corresponding to the time unit T arrives at any point in the T time window. The video player must have the ability to buffer and delay T. The upper limit on the maximum bit rate transmitted for each given representation of the program for time window T may be predetermined to limit decoder complexity and decoder buffer size. The decoding buffer is generally a requirement of the video decoder, such as a specific condition of the setting buffer size at the MPEG-2 level, and a combination of T and bit rate that satisfies it is selected. The

  Track fast forward and fast reverse for each program are also stored on the VoD server. They are stored at a single bit rate or at various bit rates and can be adapted as is usually done for play programs.

  The determination of the time window T length is made based on several factors. T needs to be large enough to obtain a statistical multiplexing increase for the most part, but for start-up and normal playback (play) and trick playback (trick play) in video players. For switching between, it must be small enough to limit the delay. T must be small enough to satisfy the decoder requirements.

  From the foregoing, systems and methods for efficiently providing video on demand while maintaining fixed channel bandwidth limitations are disclosed.

FIG. 1 shows a block diagram of a first embodiment of a basic configuration of a video on demand server according to the present invention. FIG. 2 shows a block diagram of a second embodiment of the basic configuration of the video on demand server according to the present invention. FIG. 3 shows a block diagram of a third embodiment of the basic configuration of the video on demand server according to the present invention.

Claims (16)

A method for transmitting a plurality of precoded programs having different bit rates across a fixed bandwidth channel, comprising:
Generating at least two different bit rate representations for each program, the step of generating the at least two different bit rate representations comprising: Generating the lowest bit rate replication with a rate for each program, wherein C is the total channel capacity at time T and P is the total number of programs ;
Providing control information in each of a plurality of successive time windows T for each representation of each program, wherein the control information for each successive window is for the representation of the corresponding program. Providing the control information indicating a bit rate and a quality indicator ;
During each time window T, by selecting the bit rate replicator presentation for each program, to maximize the quality indicator of the selected replicator presentation without exceeding the total available capacity of the channel, step met The step is limited during all time windows

(However,
C is the total possible channel capacity in time frame T,
P is the total number of programs,
p∈ (0, P−1) is the index of a specific program,
N (p) is the total number of representations for program p,
n [p] ∈ (0, N [p] −1) is the index of a particular representation of program p,
r [p, x] is the bit rate of the representation x of program p during T)
Selecting a representation for each program that satisfies the steps , and
Having a method. The method of claim 1, wherein providing the control information further comprises setting a peak signal to noise ratio (PSNR) as a quality indicator represented in the control information. Each of the replicator presentation n Programs [p] to select ∈ a (0, N [P] -1 ), further comprising the step of maximizing the quality indicator of the program p, A method according to claim 2. (A) sorting information about each program with a bit rate and quality index that monotonically increases with the index value;
(B) storing, for each index value, each bit rate increment (delta) and quality indicator ;
Comprising the steps of starting with (c) the lowest index value,
(c1) calculating a total capacity S for a program representation previously selected for the index value;
(D) selecting a program representation with the lowest quality index ;
(E) checking if the bit rate increment of the program selected with the lowest quality metric exceeds the total channel capacity when added to the previously selected and added representation;
And if not,
(F) repeating the steps of increasing the index value, and (g) Step (c 1) ~ (f), The method of claim 3 further comprising the steps. The step of selecting includes

(However, q [p, x] is replicator is the quality indicator of presentation x program p during T) by solving, by selecting the replicator presentation for each of the programs, the quality of individual programs The method of claim 1, further comprising the step of maximizing the sum of . The step of selecting includes

(However, q [p, x] is replicator is the quality indicator of presentation x program p during T) by solving, by selecting the replicator presentation for each of the programs, the quality of individual programs The method of claim 1, further comprising the step of maximizing the product of: A system for transmitting a plurality of precoded programs having different bit rates across a fixed bandwidth channel,
Means for generating at least two different bit rate representations for each program, wherein the means for generating at least two different bit rate representations comprises a peak bit rate less than C / P. Generating the lowest bit rate representation having for each program, C is the total channel capacity at time T, and P is the total number of programs ;
Means for providing control information for each representation of each program in each of a plurality of successive time windows T, wherein the control information for each successive window is a bit for the representation of the corresponding program Means for providing said control information indicating a rate and a quality indicator ;
During each time window T, by selecting the bit rate replicator presentation for each of the programs, to maximize the quality indicator of the selected replicator presentation without exceeding the total available capacity of the channel, with means And the means is limited during all time windows

(However,
C is the total possible channel capacity in time frame T,
P is the total number of programs,
p∈ (0, P−1) is the index of a specific program,
N (p) is the total number of representations for program p,
n [p] ∈ (0, N [p] −1) is an index of a specific representation of the program p,
r [p, x] is the bit rate of the representation x of program p during T)
Means to select a representation for each program that satisfies
Having a system. Means for providing means and said control information to said generation as a whole, it has a plurality of multirate stream generators associated with respective one of a corresponding plurality of pre-coded programs, according to claim 7 system. Means for providing means and said control information to said generation as a whole, has a multi-rate video encoder for encoding at least two bit rates replicator presentation of each pre-coded program, according to claim 7 The system described in. The means for generating and the means for providing the control information are as a whole:
A multi-rate video encoder that encodes at least two bit rate representations of each precoded program;
A plurality of transport packetizer which serves to packetize the bit rate representation for each of the pre-coded programs,
8. The system of claim 7 , comprising: The system of claim 7 , wherein the means for selecting comprises a statistical multiplexer. The means for selecting is
A statistical multiplexer; and
A transport packetizer that packetizes the representation to be selected;
10. The system of claim 9 , comprising: The system of claim 7 , wherein the means for providing control information sets a quality indicator according to a peak signal to noise ratio (PSNR). Said means for selecting, in order to maximize the quality indicator program p, selects each of the program replicator presentation n [p] ∈ of (0, N [P] -1 ), according to claim 7 System. The means for selecting is

(Where q [p, x] is the quality index of the representation x of the program p during T), so that the total quality of each program is maximized for each program. The system of claim 7 , wherein a presentation is selected. The means for selecting is

(Where q [p, x] is the quality index of the representation x of the program p during T), so that the program of each program is maximized so that the product of the quality of each program is maximized. The system of claim 10, wherein a presentation is selected.

Images