Studies On Adaptive Video Quality Control Based On IP Network

The applications based on Internet video has seen increased popularity. "Groupdecision supporting platform" requires high quality video communication. Videocontains large amount of information, while Internet is best-effort service and has noguarantee of video quality. Video system for transmission needs to provide goodvisual quality based on current network condition. Besides coding efficiency,IP-network oriented video communication has higher requirements: encodingmethod evolved from fixed bit-rate suitable for storage to variable bit-rate; handlingof packet loss and flow control. This paper studies the methodology to control thevideo quality and provide good images to end-user under the condition of IP networkenvironment change. The following is some of the achievements and innovations:1. Propose and design an adaptive video quality control mechanism for IPnetworkQoS in video data flow includes media QoS and network QoS. The former isrelated to encoding rate and quantization factor in the encoder; and the latter isrelated to network latency and packet loss rate. This paper proposes a mechanism toadaptively control the video quality when the environment changes in the IP networksuch as network flow rate and congestion. This mechanism is consisted of a feedbacksystem with network monitors in both encoder and decoder. Network parameter ismonitored in the decoder to adjust the buffer in the receiver and improve theplayback quality; in the encoder, through the Similar Laplacian distortion modelproposed in this paper, the network feedback parameters are used to adaptivelycontrol the target bit rate and quantization factor of the video frame and micro-block,reduce congestion, and improve media QoS.2. Propose and design a Similar Laplacian distortion modelBit rate control in video compression is to estimate a set of parameters toachieve minimum distortion under the limit of compression rate. In this paper,Laplacian rate distortion model is derived based on the probability distribution of theparameters aider DCT quantization, and the Laplacian PDF. With this model, thisdesign implements a data rate control function for video frame and micro-block layerin the transmitter, transmitter buffer over-flow/under-flow equation, and usingproportion interception algorithm in FGS enhanced layer. As shown in theexperiment results, the encoding stability and video quality can gain 32.96％byusing SL model. 3. Propose and implement an iterative stability approaching algorithm fornetwork bandwidth calculation, and weighted congestion-adjustablealgorithmDue to randomness of network flow and congestion, without QoS mechanism,the video quality at receiver can experience large fluctuation over the networktransmission. Traditional AIMD algorithm monitoring the network congestion canalso have large fluctuation in video quality. The iterative stability approachingalgorithm proposed in this paper, periodically generates test packet between the twoends, iteratively calculates the network flow change based on the packet loss, andmakes the target bit rate change close to the maximum available bandwidth smoothlyand quickly. Experiment result shows that this algorithm has good convergence andsmoothness. Considering the friendliness of TCP, congestion control needs to bedesigned for UDP. Packet loss rate at the receiver can be used to measure thecongestion. The ratio factor weighting can be used to adjust the transmission rate toavoid disadvantage of AIMD. Target bit rate in iterative stability approachingalgorithm, under congestion control is suitable for network environment. Experimentresult shows that this congestion algorithm has quick response and is TCP-friendly.4. Propose an error control in code retransmission for FGS and regularlayer, and a robust packaging method.This paper proposes a re-transmission strategy in basic and enhanced layer ofFGS and a decomposition of enhanced layer into multiple packages using bit-planecoding. A new UDP packet header is also proposed for efficient and robusttransmission of video data. Compared to regular layered encoding method, this paperproposes error recovery method which is different form ARQ. This method reducedthe data computation quantity 21.8％and is suitable for quick re-transmission. Arobust packaging method(1 FlexMux) after source encoding in encoder, utilizes therelationship between RTP packet and Microblock size to reduce the correlationbetween groups and increase the robustness of I frame.Experiment results of the whole system show that video average PSNR gains1.22dB—2.31dB by using the video quality control mechanism in this paper, undersevere congestion (average packet loss rate of 15％-20％) and same video test sets.