标签: multimedia processing

Electronic Information Engineering

Simulatio…

I. Introduction

With the development of techniques and the growing demand for Internet multimedia services, high-quality video transmission attracts more and more network researchers [1-3]. In order to achieve video transmission through wireless network, video should be coded and packetized. During the transmission, packet loss always happens. Packet loss is a kind of data loss, and the loss rate will influence the video transmission quality. Therefore, a question is purposed by researchers: how do the packet losses impact the video quality?

In this report, models and simulations are used to evaluate packet loss effect on video transmission in wireless networks. Due to the renovation of Internet multimedia services, we will focus on various types of video transmission on different network.

II. Literature Review

The literatures are divided into two parts: before 2018 and after 2018. Techniques develop every day, by comparing the literatures, we may find the enhancement of multimedia techniques.

A. Literature before 2018

As we mentioned before, videos need to packetized before transmission. Generally, video is segmented and packetized in two ways: fixed-size packets or a variable sized packet can contain one or more slices. Typical scenarios for fixed-size packetization are:

a) a packet contains part of one slice;

b) a packet contains the end of one slice and the beginning of another;

c) a packet contains a frame header.

And these will cause the loss of:

a) one slice;

b) two slices;

c) an entire frame [4].

A method to retrieve data to model the visibility of video transmission is subjective test. After getting the subjective evaluation of packet loss, statistical methods are used to model the visibility of video transmission. The first goal is to classify packet losses as visible or invisible, and the second goal is to predict the probability of packet loss visibility, which can be converted to a regression problem. Researchers mainly use CART and GLM to achieve the goal.

B. Literature after 2018

Since on the real network, it is impossible to quantitatively simulate the different degrees of network damage, researchers propose an advanced simulation tool-set which integrates EvalVid framework into NS-2 and simulate different level of network damage by changing the parameter QoS (Quality of Service) in the simulation experiment [5-6]. With the enhancements, the tool-set allows network-related researchers to evaluate real video streams using their proposed network designs or protocols as well as evaluate video quality of their designed video coding mechanisms using a more realistic network. The model is based on PSNR (Peak Signal to Noise Ratio) value. The main factors affecting the quality of MPEG2 video are encoding rate and packet loss [7].

Researchers presented three methods to estimate Mean Squared Error (MSE) due to packet losses directly from the video bitstream. No Parse uses only network level measurements (like packet loss rate), Quick Parse extracts the spatio-temporal extent of the impact of the loss, and Full Parse extracts sequence-specific information including spatio-temporal activity and the effects of error propagation [8]. They finally come to a conclusion that the most accurate method is Full Parse and the No Parse is the least accurate of our methods, which treats all losses identically and only estimates the MSE on the same time scale as the PLR (packet loss rate) [8].

According to the scatter plot of the relationship between the packet loss rate and the PSNR value, it is found that there is a cubic relationship between the packet loss rate and the PSNR value after linear, square, mixed, logarithmic, power, and cubic model matching [5].

III. Analysis

A. Literature before 2018

According to Kanumuri, Cosman and Reibman, the initial error caused by a packet loss propagates in space and time as a result of the video decoding algorithm. The exact error due to the packet loss can be completely described by:

a) the initial error for each macroblock in the lost packet;

b) the macroblock type;

c) motion information for subsequently received macroblocks.

The latter two control the temporal duration and spatial spread of the error. They apply Logistic Regression, a type of GLM, to the problem of estimating the probability that a packet loss is visible to an average viewer. Software R is used for the GLM model fitting. Notice that p-value is calculated during the experiment: if the p-value is less than 0.05, then the factor is significant at the 95% level.

B. Literature after 2018

The mean squared error (MSE) is used as a rough measure of video quality. The methodology the researchers apply is general for any motion-compensated video compression algorithm, although the specific application that need to be considered is the transport of MPEG-2 video when Transport Packets may be lost.

Network-specific factors include delay, delay variation, bit-rate, packet loss rate, and the latter two are most important [5].

During the simulation, after fitting the curve, it is found that the packet loss rate and the user experience quality QoE do show a univariate nonlinear relationship.

Comparing the two figures, it can be found that under the same packet loss rate, the user experience quality is different. This is because the content complexity of the two videos is different. Therefore, the video content complexity has an impact on the user experience quality. Observing the two result graphs in Figure 3 and Figure 4, it is found that in both graphs, as the packet loss rate increases, the QoE first increases, then decreases, and finally increases again, and there are two inflection points. And comparing the two figures, we can know that the less complex the video content, the worse the QoE will be. The x-axis of the two figures is the packet loss rate, and the y-axis is the PSNR value.

C. Comparison

We may easily find that in 2004, researchers mostly use “visibility” to estimate the simulation model of video transmission with packet loss, while “QoE” and “QoS” are used after 2018. GLM model is used in 2004, and researchers need to use software R to calculate p-value to determine whether the model is correct. However, model based on PSNR is used in “Mathematics Mapping Model of Network Video Packet Loss Rate and the Quality of Experience under the Internet of Things”, and cygwin + NS2 network simulation built on Windows system can provide better simulation. However, we need to notice that packet loss is one of the factors affects video transmission quality, up till now, all researches only discuss ideal conditions.

IV. Reference List

[1] C. H. Lin, C. H. Ke, C. K. Shieh, & N. K. Chilamkurti. “The Packet Loss Effect on MPEG Video Transmission in Wireless Networks”, Proceedings – 20th International Conference on Advanced Information Networking and Applications, Pages 565-570, May 2006.

[2] P. Seeling, M. Reisslein, and B. Kulapala. “Network Performance Evaluation Using Frame Size and Quality Traces of Single-Layer and Two-Layer Video: A Tutorial”, IEEE Communications Surveys and Tutorials, Vol. 6, No. 2, Pages 58-78, Third Quarter, 2004.

[3] O. Rose. “Statistical properties of MPEG video traffic and their impact on traffic modeling in ATM systems”, Report No. 101, Institute of Computer Science, University of Wurzberg, February 1995.

[4] S. Kanumuri & P. Cosman & A. Reibman & Vinay A. Vaishampayan. “Modeling Packet-Loss Visibility in MPEG-2 Video”, IEEE transactions on multimedia, Vol. 8, no. 2, April 2006.

[5] Y. B. Hou. “Mathematics Mapping Model of Network Video Packet Loss Rate and the Quality of Experience under the Internet of Things”, Software Engineering and Applications, Vol 08, Pages 131-140, June 2019.

[6] M. Siller & J. Woods. “QoE Improvement of Multimedia Transmission”, Proceedings of the IADIS International Conference, Vol 2, Pages 821-825, May 2014.

[7] M. Volk & J. Guna & A. Kos & J. Bester. “IPTV Systems, Standards and Archi-Tectures: Part II-Quality-Assured Provisioning of IPTV Services within the NGN Environment”, IEEE Communications Magazine, Vol 46, Pages 118-126, May 2008. 

[8] A.R. Reibman & V.A. Vaishampayan & Y. Sermadevi. “Quality Monitoring of Video over a Packet Network”, IEEE Transactions on Multimedia, Vol 6, Pages 327-334, April 2004.