CN106303539A - A kind of unequal error transmission protecting based on H.264 video message type - Google Patents

Abstract

The present invention relates to a kind of transmission of video unequal error protection method based on H.264 video message content.The method according to the importance of real network packet loss environmental metrics H.264 video message different content and carries out classification to video streaming content according to this; set up redundancy bandwidth model of optimizing allocation; use uniform enconding that video flowing is carried out unequal error protection; use less bandwidth to achieve the protection of more preferable code stream, effectively reduce the transmission distortion of video.

Description

A kind of unequal error transmission protecting based on H.264 video message type

Technical field

The invention belongs to Network Video Transport Technology, a kind of based on H.264 video message type Error transmissions guard method.

Background technology

Although network video service is widely used and huge development, but, owing to transmission of video should All particularitys, this yet suffers from many urgently to be resolved hurrily and perfect problems.Existing Video coding mode For obtaining high compression rate, generally use based on transition coding and motion compensation framework, utilize estimation and fortune Dynamic compensation eliminates the time redundancy of interframe, utilizes transition coding and quantization operation to eliminate the spatial redundancy in frame. The realization of above method depends on substantial amounts of interframe, intraframe predictive coding, and incident is tight between data Weight dependency, it is very sensitive to network characteristics such as packet loss, time delay, shakes that this allows for network video stream. Therefore, the video data in network is carried out forward error correction, use certain mode to carry out error protection, produce A certain amount of redundancy deteriorates the video quality damage caused with opposing network environment, is necessary.

Video data bag in network carries different video datas, and these video datas are because of video content, volume The characteristics such as code mode, have again different importances, thus its quality being affected by network condition and causing are damaged Wound has again obvious difference.Even if the key component of video loses very small amount of data, it is also possible to causes and regards Frequently being decreased obviously of quality；Even and if unessential part loses more data, do not result in serious video yet Quality Down.Therefore, when carrying out error protection, carry out importance classification for network video data message, And carry out unequal error protection according to different grades of importance, it is a kind of effective way solving problem.At this Planting under application demand, the packet loss characteristic for network proposes a kind of video data bag importance criterion, becomes Primary and necessary problem.National inventing patent [the document 1: a kind of based on without rate modulation of Xing Chengwen et al. Stereo video transmission method, CN104394401A] study for H.264/MVC format video, the most just The Primary layer of Video coding and enhancement layer distinguish the importance of video message, divide the importance of message content not Enough fine.Therefore, need a kind of hierarchical approaches badly the importance of video message is effectively divided.

Video can cause the damage of video quality when compressed encoding, is referred to as code impairment.Due to Video coding Mode have a lot, regulatable parameter is also a lot, by the way of these are different and regarding of obtaining of parameter coding Frequency stream degree of injury is different, and its code check is the most different.Bit rate output is the highest, and video impairment is the least, and transport overhead is then The biggest；Otherwise, bit rate output is the lowest, and video impairment is the biggest, and transport overhead is the least.The bandwidth of wireless network Resource is very limited, and larger fluctuation likely occurs.National inventing patent [the document 2: of Shi Dongxin et al. Plant the Raptor code method of system unequal error protection, CN104320147A] use the Raptor in fountain codes Code carries out unequal error protection to network data, but owing to the complexity of fountain codes own is higher, the time delay caused Bigger, it is not suitable for RTP Transport System for Real-time, and after certain segment data carries out Raptor code coding, if Decoding unsuccessfully, the source data void in whole being protected, when tackling higher packet loss, performance can significantly decay. The national inventing patent of Dong Yuning et al. [document 3: live video communication method based on unequal error protection, CN103067719A] for the image sets (Group of Pictures, GOP) of H.264/AVC video sequence Structure has carried out importance classification, and the packet using forward error correction technique to constitute each video code flow is protected Protect, for the distortion in transmission, use error concealment methods to recover in decoding end, do not change Video coding Speed, therefore when channel bandwidth change, the waste of bandwidth resources may be caused because bandwidth is excessive, or due to Bandwidth is not enough and causes serious congestion packet loss, and, this method needs forward error correction and error concealing common Effect, i.e. transmitting terminal and receiving terminal to Video Applications system is all modified, and error concealing will also result in necessarily Time delay.Therefore, how to distribute limited bandwidth resources for transmitting video code flow and redundancy, with maximum journey Degree ground recovery video information, reduces coding, video quality damage that transmission brings, is one and very important asks Topic.

Summary of the invention

It is an object of the invention to provide the transmission protection of a kind of unequal error based on H.264 video message type Method, it is possible to realize the unequal error protection of video flowing, obtains optimal video transmission quality.

The technical solution realizing the object of the invention is: a kind of not equal difference based on H.264 video message type Wrong transmission protecting, is first according to real network packet loss environmental metrics H.264 video message different types of heavy The property wanted also carries out classification according to this to video streaming content, i.e. by resolving video flowing, according to the frame type of video packets Classification is carried out with image sets GOP position；Then set up the redundancy bandwidth model of optimizing allocation of uniform enconding；? Adjusting the code check of source coding and channel coding on the basis of this, chnnel coding uses uniform enconding to carry out video flowing Unequal error protection, finds the channel width allocation proportion that the minimum of video quality decay is corresponding, obtains network Random loss rate is the best video quality that the receiving terminal between 1% to 10% receives.

Compared with prior art, its remarkable advantage is the present invention: (1) uses based on frame type and GOP position Video message importance hierarchical approaches, obtain the importance ratio of video message, and grind for network packet loss rate Studying carefully its importance, stage division is more reasonable, more conforms to real network environment.(2) uniform enconding pair is used Code stream is protected, and sets up redundancy bandwidth model of optimizing allocation based on uniform enconding, obtains under different packet loss rate The redundancy bandwidth allocation proportion of different importance classifications, receiving terminal recovers video quality and is significantly improved so that nothing Line network video stream is preferably protected under random loss environment.(3) unequal error protection method is being determined On the basis of, solve network bandwidth Data Rate Distribution problem between message sink coding and uniform enconding, obtain optimum Allocation proportion, it is possible to use less bandwidth to complete the preferable protection of video flowing, be that a kind of efficiency is high, in real time The guard method that property is good, applied widely.

Below in conjunction with the accompanying drawings the present invention is described in further detail.

Accompanying drawing explanation

Fig. 1 is the flow chart that simulation controls packet loss.

Fig. 2 is that the present invention is based on frame type and the flow chart of the unequal error protection of GOP position.

Detailed description of the invention

Present invention unequal error transmission protecting based on H.264 video message type, including three parts: H.264 video flowing importance classification based on frame type and GOP position, redundancy bandwidth based on uniform enconding Model of optimizing allocation is set up and the distribution of signal source and channel bandwidth, and specific implementation process is:

1, H.264 video flowing is carried out importance classification based on frame type and GOP position

First, by resolving the dissimilar of message entrained by code stream differentiation different pieces of information bag: in compression H.264 In video flowing, frame of video is divided into I frame, P frame, B frame three basic forms of it according to generating mode and referential, And with GOP for absolute coding unit.I frame uses intraframe coding method, i.e. encodes without reference to other frames； P frame is forward predictive coded, and reference frame can be I frame, P frame, and plays out according to coded sequence；B Frame uses bi-directional predictive coding, refers to playing sequence I frame before it, afterwards, P frame (B frame itself Typically not there is referential).So, in H.264 video flowing, the mistake of I frame can cause and generate with reference to it The mistake of P, B frame, and the mistake of P frame can cause the mistake of the B frame generated with reference to it, the mistake of B frame is then Do not interfere with other frames (not there is referential because of it).From informative angle it can be seen that the importance of I frame Higher than P frame, and the importance of P frame is higher than B frame.And in each gop, the first frame is I frame；P frame Generating with reference to I frame and GOP position P frame before it, position P frame earlier above has higher reference Property, therefore P frame is reduced successively by its GOP position importance from front to back；B frame is then with reference to I frame and P Frame generates, and owing to B frame itself does not have referential, therefore its position is unrelated with its importance.

Secondly, H.264 video flowing is divided into different importance rates, I according to frame type and GOP position Frame data are one-level, and B frame data are one-level, and P frame data is depended on by importance after forward direction according to its GOP position Secondary reduction, is divided into different importance rates.Resolve video flowing, according to above-mentioned stage division classification, according to respectively Importance rate video packets loses the impact on video quality, obtains between its Loss Rate and video quality decay For linear relationship, I frame packet importance is high, and B frame packet importance is low, the packet of P frame by its GOP position by After forward direction, importance successively decreases.

Again, different type of messages is independently controlled packet loss, obtain between packet loss and video quality decay Relation.When H.264 video flowing being carried out importance classification based on frame type and GOP position, first have to Distinguish the dissimilar of message entrained by different pieces of information bag by resolving code stream, then different types is independently controlled Packet loss processed, determines the relation between its packet loss and video quality decay.H.264 video is by network abstraction, layer NAL (Network Abstract Layer, NAL) unit forms, and NAL unit is divided into NAL head and former Beginning syllable sequence load (Raw Byte Sequence Payload, RBSP), type is the RBSP of video data band It is divided into again slice header (Slice Header) and strip data (Slice Data), the frame type of video message (I, P, B frame) it is stored in NAL head, and its GOP location number can be by Slice Header and sequence parameter set (Sequence Parameter Set, SPS), picture parameter set (Picture Parameter Set, PPS) are altogether Learn with resolving.

H.264 video flowing is when carrying out network transmission, is encapsulated as UDP bag, carries out sliced transmission in transport layer, It is that MTU, MTU refer to MTU (Maximum that transport layer carries out the greatest length of burst Transmission Unit, MTU), Network Packet Loss is also carried out as unit.Therefore, true for simulation Network Packet Loss phenomenon, use X264 encoder carry out Video coding time, i.e. according to MTU size limit Carry out burst coding, that is each NAL unit size is less than MTU, it is achieved application layer video packets is to net The mapping one to one of network packet.Thus, thus control to abandon video packets and can simulate control packet loss.? On the basis of this, the video packets number of different importance rates is added up, according to packet loss arrange generation with Machine ordered series of numbers, selects to abandon the video packets that number is present level video packets number and the product of packet loss, i.e. The video quality decay that under available different packet loss rate, the video packets independence packet loss of different importance rates brings；

When controlling packet loss, first resolve H.264 video flowing and obtain frame type and the GOP position of video packets, Video packets is carried out point according to the P frame data that content is I frame data, B frame data and different GOP position Level；Add up the number of each scalable video packet in video sequence and it is carried out label, being answered according to packet loss The grouping number lost the random number series producing label in the total packet number of current classification, by random number series The packet of label abandons, repeatedly and using average as result.Packet loss is set as 1% to 10%, every 1% Stepping.

Finally, use FFmpeg to be restored video after being decoded, the quality recovering video is regarded with original Frequently quality versus, uses Y-PSNR (Peak Signal Noise Ratio, PSNR) evaluation methodology assessment Video quality；Again video directly encoded and decode, and contrasting with original video, obtaining decoding video Quality.Decoding video and recovery video quality are asked poor, offsets the compression artefacts of video, obtain random loss and make The transmission distortion become.Thus, the pass between available different brackets video packets packet loss and video quality decay It is curve, and the video quality attenuation differences caused by different brackets packet loss accounts for the most superfluous as redundancy code check needed for it The reference of remaining bandwidth ratio.Thus can obtain each grade video packets Loss Rate and video quality decay between linearly Relation, each grade has visibly different slope because of its importance difference.

The idiographic flow of said process sees Fig. 1.Message sink coding uses X264 encoder, uses code stream FFmpeg decoder carries out once local decoding, obtains local marred video.To H.264 solve by code stream again Analysis, identifies the perceptual importance grade of NAL unit message content, adds up dissimilar (frame type, GOP Position) number of NAL unit (i.e. video packets)；Random sequence is produced according to set packet loss, The NAL unit of each type of random drop independently；FFmpeg is used to carry out 264 impaired video flowings After video decoding, according to each NAL unit sequence order in 264 video flowings, use and replicate former frame Method compensates the frame lost, the impaired yuv video after being registrated；By this locality YUV marred video be subject to Damage yuv video and carry out PSNR evaluation with source video for reference video respectively, obtain local evaluation result and lose Bag evaluation result, asks difference to obtain its PSNR decay, obtains the video quality damage that Network Packet Loss causes.

2, redundancy bandwidth model of optimizing allocation based on uniform enconding is set up

The principle of uniform enconding is that through certain linear combination, m source block is generated n redundant data Block, sends to receiving terminal jointly by m+n data block, as long as receiving terminal receives any m number therein According to block, by its linear relationship, m source block correctly can be recovered.Uniform enconding based on video packets Principle be that the packet of m source is generated n redundancy packets through linear combination, will be grouped by source and redundancy packets M+n packet of composition is common to be sent to receiving terminal, any m in decoder can have correctly received individual with Upper packet, when i.e. lost packets number is not more than n, m source video packets can be recovered completely；When decoder just The packet count really received is less than m, then redundancy packets therein all abandoned, and only source packet is decoded. According to the principle of above-mentioned packet-based uniform enconding, with different importance scalable video packet packet loss and reception The linear relationship of end video quality decay is reference, can set up redundancy bandwidth model of optimizing allocation.

Setting video stream is classified to t grade according to message importance, and j-th stage source grouping number is m_j, redundancy Packet count is n_j, the packet loss of packet-video quality attenuation curve slope is k_j, then being located at network packet loss rate is p's Under the conditions of, the probability of j-th stage single source packet loss is remaining in himself loss, and this grade m_j+n_jIn-1 packet, number of missing (number of missing is i) is more than n_jThe total probability of-1 is p_{R, j},

It is represented by

$p_{r,j}=p\·\underset{i=n_j}{\overset{m_j+n_j-1}{\mathrm{\Σ}}}{C}_{m_j+n_j-1}^i{p}^i{(1-p)}^{m_j+n_j-1-i}$

WhereinRefer to probability.

Then the overall video distortion under conditions present is

$\begin{array}{c}\mathrm{\ΔPSNR}=\underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}{k}_j{p}_{r,j}\\ =\underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}{k}_{j}p\·\underset{i=n_j}{\overset{m_j+n_j-1}{\mathrm{\Σ}}}{C}_{m_j+n_j-1}^i{p}^i{(1-p)}^{m_j+n_j-1-i}\end{array}$

Therefore, redundancy code check assignment problem can be expressed as optimization problem

$\left\{\begin{array}{c}\begin{array}{cc}\min & \underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}\left[k_{j}p\underset{i=n_j}{\overset{m_j+n_j-1}{\mathrm{\Σ}}}{C}_{m_j+n_j-1}^i{p}^i{(1-p)}^{m_j+n_j-1-i}\right]\end{array}\\ \\ \underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}{n}_j=n\\ n_j\≥0\end{array}\right.$

Wherein n is redundancy packets number, (is approximately with source video packets greatest length for the total redundancy bandwidth of channel MTU) ratio.The solution of optimization problem is tried to achieve by search, under the conditions of available different packet loss rate, and channel Optimization redundancy adding proportion needed for each importance rate during bandwidth difference redundancy ratio.Such as, by Live Tr.yuv video in Video Database normal video storehouse according to The gop structure of " IBBBBBPBBBBBPBBBBBPBBBBBPBBBBBP " uses X264 burst to compile Code is used for testing for source video flowing, then it can be divided into I, B, P according to importance stage division₁、P₂、P₃、P₄； When channel width is 1000kbps, set redundancy bandwidth as the 10% of channel width, to above-mentioned Optimized model Solve the ratio such as table of redundancy interpolation code check and the total redundancy code check that can obtain each importance classification under different packet loss rate Shown in 1, with ratio in table as reference when taking uniform enconding that video packets is protected.

During table 1 10% redundancy, test video different brackets video sequence amount of redundancy accounts for channel total amount of redundancy ratio

3, the distribution of signal source and channel bandwidth

Message sink coding is both needed to consume certain bandwidth with chnnel coding (for uniform enconding in the present invention), and information source is compiled Code code check is the highest, then the source video quality delivering to network is the highest；Chnnel coding code check is the highest, then can be in network Video the best protection is provided.Under conditions of limited bandwidth, and too high message sink coding code check can cause letter Road encoder bit rate is too low, it is impossible to what video quality was brought by opposing packet loss has a strong impact on；Too high chnnel coding code Rate can cause message sink coding code check too low, and source video quality is too low.

Therefore, under conditions of channel total bandwidth is certain, adjust the code check of source coding and channel coding, find Optimum allocation proportion, optimum allocation proportion refers to the allocation proportion that minimum that video quality is decayed is corresponding. Concrete methods of realizing is the guard method first determining unequal error, then in the condition that channel total bandwidth is certain Under, under 1% to 10% (every 1% stepping) network packet loss rate, change the code check of source coding and channel coding, Message sink coding code check is changed to 100% (every 2% stepping) from the 80% of network total bandwidth；By such as the survey of Fig. 2 Examination obtains the optimal message source and channel Data Rate Distribution ratio under the conditions of 1% to 10% packet loss.To tr.yuv video Testing, the distribution of optimal message source and channel Data Rate Distribution ratio is as shown in table 2.Wherein R_totalFor channel total bandwidth, R_sFor message sink coding code check, R_rFor chnnel coding code check.

Optimal signal source and channel Data Rate Distribution distribution under table 2 different packet loss rate

Verify through above-mentioned test, the packet loss condition within 10% of the unequal error protection method in the present invention Under, consume the network bandwidth within 10% and effectively reduce the distortion of receiving terminal recovery video, it is provided that preferably video The transmission quality assurance.And, the present invention has good real-time and portability, it is adaptable to different networks, Can be effectively used for actual application.For any video, all its compression can be obtained by analysis within a certain period of time Characteristic also provides good protection accordingly, is effectively improved recovery video quality.

Wherein, the idiographic flow of unequal error protection sees Fig. 2.Wherein, Video coding uses X264 coding Device, video decoding uses FFmpeg decoder.Other types packet in addition to video band comprises important letter Breath, do not carry out packet loss, be directly over single passage use socket carry out UDP packing deliver to receiving terminal, This passage named ELSE passage；The video packets of each importance rate is then according to redundancy bandwidth optimized distribution mould Type is distributed the redundancy packets adding proportion obtained and is carried out uniform enconding, and uses socket to carry out UDP packing Sending, each passage is I passage (being used for transmitting I packet) according to the video packets type command transmitted, P₁, P₂..., P_n(for transmitting the P packet of each GOP position).Other in addition to ELSE passage are led to Track data bag carries out random loss；Video packets after packet loss delivers to the decoding of receiving terminal laggard line linearity, and It is ranked up according to its video sequence order, uses FFmpeg to be decoded the yuv video that is restored.YUV Video encoded generation H.264 video flowing, the latter generates yuv video through decoding.During evaluation, still With original yuv video for reference to calculating the PSNR value recovering yuv video, obtaining video compress and packet loss The total distortion caused.

Understand after tested, on the basis of video message importance classification research based on frame type and GOP position Uniform enconding unequal error protection strategy than equalization protection, guard method performance based on frame type has substantially Raising.

Claims (4)

1. a unequal error transmission protecting based on H.264 video message type, it is characterised in that first First according to the real network packet loss environmental metrics different types of importance of H.264 video message and according to this to video Stream content carries out classification, i.e. by resolving video flowing, according to frame type and the image sets GOP position of video packets Put and carry out classification；Then set up the redundancy bandwidth model of optimizing allocation of uniform enconding；Adjust information source on this basis Coding and the code check of chnnel coding, chnnel coding uses uniform enconding that video flowing is carried out unequal error protection, seeks The channel width allocation proportion that the minimum of looking for video quality to decay is corresponding, obtaining network random loss rate is 1% The best video quality that receiving terminal between 10% receives.

Unequal error based on H.264 video message type the most according to claim 1 transmission protection side Method, it is characterised in that H.264 video flowing is carried out importance classification step based on frame type and GOP position As follows:

First, by resolving the dissimilar of message entrained by code stream differentiation different pieces of information bag: in compression H.264 In video flowing, frame of video is divided into I frame, P frame, B frame three basic forms of it according to generating mode and referential, And with GOP for absolute coding unit；I frame uses intraframe coding method, i.e. encodes without reference to other frames； P frame is forward predictive coded, and reference frame is I frame and playing sequence P frame before it, and raw according to coding One-tenth order plays out；B frame uses bi-directional predictive coding, with reference to playing sequence at its forward and backward I frame and P Frame generates；

Secondly, H.264 video flowing is divided into different importances etc. according to the GOP position of frame type and P frame Level, I frame data are one-level, and B frame data are one-level, P frame data according to its GOP position by important after forward direction Property reduce successively, be divided into different importance rates, resolve video flowing, according to above-mentioned stage division classification, I Frame packet importance is high, and B frame packet importance is low, and the packet of P frame presses its GOP position by importance after forward direction Successively decrease；

Again, different type of messages is independently controlled packet loss, obtain between packet loss and video quality decay Relation, the most H.264 video flowing is when carrying out network transmission, is encapsulated as UDP bag, carries out burst in transport layer Transmission, burst greatest length is MTU MTU, and Network Packet Loss is also carried out as unit； When using X264 encoder to carry out Video coding, i.e. limit according to MTU size and carry out burst coding, H.264 Each network abstraction, layer NAL unit size of video flowing is less than MTU, it is achieved application layer video packets is to net The mapping one to one of network packet, thus controls to abandon video packets and can simulate control packet loss；Basis at this On, the video packets number of different importance rates is added up, generation random number series is set according to packet loss, Select to abandon the video packets that number is present level video packets number and the product of packet loss, i.e. can get not The video quality decay brought with the video packets independence packet loss of importance rates different under packet loss；

Finally, use FFmpeg Video Decoder to be restored video after being decoded, the matter of video will be recovered Amount and original video quality versus, use PSNR method assessment video quality；Again video is directly encoded And decoding, and contrast with original video, obtain decoding the quality of video；To decoding video and recovery video quality Ask poor, offset the compression artefacts of video, obtain the transmission distortion that random loss causes, thus obtain different important Property grade video packets packet loss and video quality decay between relation curve, and difference importance rate is lost The video quality attenuation differences that bag causes accounts for the reference of total redundancy bandwidth ratio as redundancy code check needed for it.

Unequal error based on H.264 video message type the most according to claim 1 transmission protection side Method, it is characterised in that setting up redundancy bandwidth model of optimizing allocation method based on uniform enconding is: setting video stream is pressed Importance rate according to message is classified to t grade, and j-th stage source grouping number is m_j, redundancy packets number is n_j, the packet loss of packet-video quality attenuation curve slope is k_j, then it is located under conditions of network packet loss rate is p, The probability of j-th stage single source packet loss is that himself is lost, and remaining m in this grade_j+n_j-1 packet Middle number of missing is more than n_jThe probability of-1 is p_{R, j}, it is expressed as

$p_{r,j}=p\·\underset{i=n_j}{\overset{m_j+n_j-1}{\mathrm{\Σ}}}{C}_{m_j+n_j-1}^i{p}^i{(1-p)}^{m_j+n_j-1-i}$

Then the overall video distortion under conditions present is

$\begin{array}{c}\mathrm{\ΔPSNR}=\underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}{k}_j{p}_{r,j}\\ =\underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}{k}_{j}p\·\underset{i=n_j}{\overset{m_j+n_j-1}{\mathrm{\Σ}}}{C}_{m_j+n_j-1}^i{p}^i{(1-p)}^{m_j+n_j-1-i}\end{array}$

Therefore, redundancy code check assignment problem is expressed as optimization problem

$\left\{\begin{array}{c}\begin{array}{cc}\min & \underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}\left[k_{j}p\underset{i=n_j}{\overset{m_j+n_j-1}{\mathrm{\Σ}}}{C}_{m_j+n_j-1}^i{p}^i{(1-p)}^{m_j+n_j-1-i}\right]\end{array}\\ \underset{j=0}{\overset{t-1}{\mathrm{\Σ}}}{n}_j=n\\ n_j\≥0\end{array}\right.$

Wherein n is redundancy packets number, for the ratio of the total redundancy bandwidth of channel Yu source video packets greatest length.

Unequal error based on H.264 video message type the most according to claim 1 transmission protection side Method, it is characterised in that unequal error protection process is: original yuv format video is used X264 encoder Carry out Video coding and obtain H.264 video flowing, in the wireless network random loss；Its in addition to video band He does not carry out packet loss at type packet, directly uses socket to carry out UDP packing and delivers to receiving terminal；Each importance The video packets of grade is then distributed the redundancy packets adding proportion obtained according to redundancy bandwidth model of optimizing allocation Carry out uniform enconding, and use socket to carry out UDP to transmit；According to current network packet loss rate to regarding Frequently packet carries out random loss；Video packets after packet loss delivers to the decoding of receiving terminal laggard line linearity, and It is ranked up according to its video sequence order, uses FFmpeg that H.264 video flowing is carried out video decoding, Video is recovered to yuv format；During evaluation, still regard with original yuv video for recovering YUV with reference to calculating The PSNR value of frequency, obtains the total distortion that video compress causes with packet loss.