SVC Evaluation


To evaluate the H.264/SVC transmission over different kinds of networks, especially ad hoc networks, we proposed a evaluation framework SVCEval . It is based on the SVC reference software JSVM and make use of the Qualnet simulator for the network simulation.

The SVCEval is written in Java.  An introduction of the framework is available at P156 9.1.2 SVCEval Evaluation Framework of MP-OLSR_thesis.pdf. The details of parameters for QualnetTraceParser is at P183 Appendix D of MP-OLSR_thesis.pdf.


  • A sample of the configuration is presented in P161 of MP-OLSR_thesis.pdf. Please pay attention to the parameter SliceMode and SliceArgument, which defines the size of the packet.
  • In the bitstream trace, there are some 9-byte packets that are commonly not useful for the decoding. So it’s better to remove those packets during the video transmission simulation. Check P150 Section 8.2.2 Packet priority in H.264/SVC of MP-OLSR_thesis.pdf for more info.
  • In the current simulation, an old version of JSVM (9.8) is used. This is at the moment, the error concealment function are removed in the new version. If the user wants to use the latest version, please make sure that the error concealment function are included.
  • In JSVM 9.8, there are still some bugs exist. So for the last several frames of the decoding, there are probably some errors occur. Just ignore them.

Source code is available at


or on GitHub


At the video sender, the YUV file is encoded by the JSVM encoder, and the .264 bitstream is generated. Then the BitStreamExtractor is used to generate the bitstream trace from the given bitstream. It is a text file which specifies the parameters of each packet inside the bitstream. Those parameters include the start position of the packet inside the bitstream, the length of the packets, the values of dependency id (LId), temporal level (TId) and quality level (QId) for the packet, the type of the packet and two flag which indicate whether the packet is discardable or truncatable. A Traffic Generator is written to generate the input traffic trace file for Qualnet simulation, which includes mainly the packet rate and size. The simulator will take the traffic trace file and run the simulation according the configuration of scenarios to simulate different kinds of networks.

At the video receiver, a packet trace file is produced by the simulator. The packet trace file records all the operation on each packet in each node and each layer (so normally hundreds of MBytes). A QualnetTraceParser is developed to analyse the trace to detect which packets are lost and which packets are properly received. For real-time transmission, we can set a delay threshold to discard the packets that timed out. The trace parser can generate the distorted bitstream trace, and then we use the BitstreamExtractor and JSVM decoder to have the distorted YUV file after the video transmission. Then we can evaluate the quality of video with different metrics such as PSNR or Mean Opinion Score (MOS).


FECSim is an FEC codec simulator to analysis the performance of the Unequal Error Protection of  FEC coding in H.264/SVC network transmission.

An introduction of the FECSim can be found at P160 of  MP-OLSR_thesis.pdf . A short manual is at P185, Appendix E of  MP-OLSR_thesis.pdf.


It mainly consists of two parts: in the sender, the QualnetFECTrafficGen  plays as a FEC encoder; in the receiver, the QualnetFECTraceParser  plays as a FEC decoder.


The FEC traffic generator simulates the behavior of FEC encoder. It reads the original bitstream packet trace and the configuration of the priority FEC:

t1 2 3 2 s

t2 2 3 2 s

The configuration means that for temporal layer t1  and t2 , the encoder will buffer 2 data packets, and generate 3 coded projections with the same size. In the receiver, 2 projections in 3 are needed to recover the original packets. The s  means that systematic code is used.


The FEC trace parser simulates the behavior of FEC decoder. It reads the Qualnet packet trace and generates the distorted bitstream trace based on the FEC trace (generated by QualnetFECTrafficGen ).

Related publications


Radu, Dan; Avram, Camelia; Aştilean, Adina; Parrein, Benoît; Yi, Jiazi

Acoustic noise pollution monitoring in an urban environment using a VANET network Inproceedings

2012 IEEE International Conference on Automation Quality and Testing Robotics (AQTR), pp. 244-248, IEEE, Cluj-Napoca, 2012, ISBN: 978-1-4673-0701-7.

Abstract | Links | BibTeX

Radu, Dan; Yi, Jiazi; Parrein, Benoit

QoE enhancement for H.264/SVC video transmission in MANET using MP-OLSR protocol Inproceedings

ISIVC 2012, the 6th International Symposium on signal, Image, Video and Communications, pp. 1-4, Valenciennes, France, 2012.

Abstract | Links | BibTeX


Yi, Jiazi; Parrein, Benoit; Radu, Dan

Multipath routing protocol for manet: Application to H.264/SVC video content delivery Inproceedings

Wireless Personal Multimedia Communications (WPMC), 2011 14th International Symposium on, 2011, ISBN: 1347-6890.

Abstract | Links | BibTeX