Limits...
Protection of HEVC video delivery in vehicular networks with RaptorQ codes.

Piñol P, Martínez-Rach M, López O, Pérez Malumbres M - ScientificWorldJournal (2014)

Bottom Line: In this paper we focus on HEVC video coding standard streaming in vehicular networks and how it deals with packet losses with the aid of RaptorQ, a Forward Error Correction scheme.As vehicular networks are packet loss prone networks, protection mechanisms are necessary if we want to guarantee a minimum level of quality of experience to the final user.We have run simulations to evaluate which configurations fit better in this type of scenarios.

View Article: PubMed Central - PubMed

Affiliation: Physics and Computer Architecture Department, Miguel Hernández University, Avenida de la Universidad, s/n, 03202 Elche, Spain.

ABSTRACT
With future vehicles equipped with processing capability, storage, and communications, vehicular networks will become a reality. A vast number of applications will arise that will make use of this connectivity. Some of them will be based on video streaming. In this paper we focus on HEVC video coding standard streaming in vehicular networks and how it deals with packet losses with the aid of RaptorQ, a Forward Error Correction scheme. As vehicular networks are packet loss prone networks, protection mechanisms are necessary if we want to guarantee a minimum level of quality of experience to the final user. We have run simulations to evaluate which configurations fit better in this type of scenarios.

No MeSH data available.


Related in: MedlinePlus

Percentage of bitrate increase (without FEC protection) for different number of slices per frame. (a) LP mode. (b) AI mode. (HEVC raw bitstream//HEVC + RTP header//HEVC + RTP header + fragmentation header.)
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4127264&req=5

fig2: Percentage of bitrate increase (without FEC protection) for different number of slices per frame. (a) LP mode. (b) AI mode. (HEVC raw bitstream//HEVC + RTP header//HEVC + RTP header + fragmentation header.)

Mentions: The sequence chosen for the tests is RaceHorses, one of the test sequences used by JCT-VC for HEVC evaluation in common test conditions [23]. It has a resolution of 832x480 pixels and a frame rate of 30 frames per second. We have encoded it at 1, 2, 4, 8, 13, and 26 slices per frame (slc/frm). For the encoding process we have used a value of 37 for the quantization parameter (QP). For that value we obtain a mean PSNR value of 32.12 dB for AI mode (1 slc/frm) and a value of 30.19 dB for LP mode (1 slc/frm). Video quality is higher in AI mode but bitrate in LP mode is much lower. As predictions cannot cross slice boundaries, when we split each frame into several slices we are reducing coding efficiency because we cannot use information of nearby areas if they do not belong to the slice. An example of this penalization is that slices cannot use intraprediction between slices (in AI mode) and slices cannot use predictions for motion vectors (in LP mode). Figure 2 shows the percentage of increment of the encoded sequence size at different number of slices per frame compared to 1 slc/frm. It shows the percentage of increment of HEVC raw bitstream and also the percentage of increment after adding the RTP headers to each slice. If the size of one slice (including its RTP header) is greater than the network MTU, then it will be divided into some fragments. If one of the fragments of a slice gets lost, then the whole slice will be discarded because it will be undecodable. So the rest of the fragments of the slice are automatically discarded. We identify every fragment of a slice with a header in order to know if a slice has received all its fragments. In Figure 2, data labeled as “TOTAL” shows the bitrate increment with respect to 1 slc/frm when both RTP and fragmentation headers are included. As slices generated by LP mode are much smaller than slices generated by AI mode (because of LP mode coding efficiency), the overhead introduced by RTP and fragmentation headers is much greater for LP mode, and consequently the same happens to the total percentage of increment in the bitstream. For example, encoding at 13 slc/frm increases the bitstream around a 20%, and encoding at 26 slc/frm increases it around a 40% for LP mode. But for the same number of slices per frame, in AI mode the increments are around 8.6% and 12.8%, respectively.


Protection of HEVC video delivery in vehicular networks with RaptorQ codes.

Piñol P, Martínez-Rach M, López O, Pérez Malumbres M - ScientificWorldJournal (2014)

Percentage of bitrate increase (without FEC protection) for different number of slices per frame. (a) LP mode. (b) AI mode. (HEVC raw bitstream//HEVC + RTP header//HEVC + RTP header + fragmentation header.)
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4127264&req=5

fig2: Percentage of bitrate increase (without FEC protection) for different number of slices per frame. (a) LP mode. (b) AI mode. (HEVC raw bitstream//HEVC + RTP header//HEVC + RTP header + fragmentation header.)
Mentions: The sequence chosen for the tests is RaceHorses, one of the test sequences used by JCT-VC for HEVC evaluation in common test conditions [23]. It has a resolution of 832x480 pixels and a frame rate of 30 frames per second. We have encoded it at 1, 2, 4, 8, 13, and 26 slices per frame (slc/frm). For the encoding process we have used a value of 37 for the quantization parameter (QP). For that value we obtain a mean PSNR value of 32.12 dB for AI mode (1 slc/frm) and a value of 30.19 dB for LP mode (1 slc/frm). Video quality is higher in AI mode but bitrate in LP mode is much lower. As predictions cannot cross slice boundaries, when we split each frame into several slices we are reducing coding efficiency because we cannot use information of nearby areas if they do not belong to the slice. An example of this penalization is that slices cannot use intraprediction between slices (in AI mode) and slices cannot use predictions for motion vectors (in LP mode). Figure 2 shows the percentage of increment of the encoded sequence size at different number of slices per frame compared to 1 slc/frm. It shows the percentage of increment of HEVC raw bitstream and also the percentage of increment after adding the RTP headers to each slice. If the size of one slice (including its RTP header) is greater than the network MTU, then it will be divided into some fragments. If one of the fragments of a slice gets lost, then the whole slice will be discarded because it will be undecodable. So the rest of the fragments of the slice are automatically discarded. We identify every fragment of a slice with a header in order to know if a slice has received all its fragments. In Figure 2, data labeled as “TOTAL” shows the bitrate increment with respect to 1 slc/frm when both RTP and fragmentation headers are included. As slices generated by LP mode are much smaller than slices generated by AI mode (because of LP mode coding efficiency), the overhead introduced by RTP and fragmentation headers is much greater for LP mode, and consequently the same happens to the total percentage of increment in the bitstream. For example, encoding at 13 slc/frm increases the bitstream around a 20%, and encoding at 26 slc/frm increases it around a 40% for LP mode. But for the same number of slices per frame, in AI mode the increments are around 8.6% and 12.8%, respectively.

Bottom Line: In this paper we focus on HEVC video coding standard streaming in vehicular networks and how it deals with packet losses with the aid of RaptorQ, a Forward Error Correction scheme.As vehicular networks are packet loss prone networks, protection mechanisms are necessary if we want to guarantee a minimum level of quality of experience to the final user.We have run simulations to evaluate which configurations fit better in this type of scenarios.

View Article: PubMed Central - PubMed

Affiliation: Physics and Computer Architecture Department, Miguel Hernández University, Avenida de la Universidad, s/n, 03202 Elche, Spain.

ABSTRACT
With future vehicles equipped with processing capability, storage, and communications, vehicular networks will become a reality. A vast number of applications will arise that will make use of this connectivity. Some of them will be based on video streaming. In this paper we focus on HEVC video coding standard streaming in vehicular networks and how it deals with packet losses with the aid of RaptorQ, a Forward Error Correction scheme. As vehicular networks are packet loss prone networks, protection mechanisms are necessary if we want to guarantee a minimum level of quality of experience to the final user. We have run simulations to evaluate which configurations fit better in this type of scenarios.

No MeSH data available.


Related in: MedlinePlus