the h.265/mpeg-hevc standard
DESCRIPTION
The H.265/MPEG-HEVC Standard Benjamin Bross, Fraunhofer HHI Annual Member Meeting - October 16TRANSCRIPT
Fraunhofer
Heinrich Hertz Institute
Fraunhofer
Heinrich Hertz Institute
Fraunhofer Heinrich Hertz Institute, Einsteinufer 37, 10587 Berlin www.hhi.fraunhofer.de
The H.265/MPEG-HEVC Standard
IMTC Annual Member Meeting – Redmond, WA, October 16, 2014
Fraunhofer
Heinrich Hertz Institute
© 20.10.2014 2
High Efficiency Video Coding (HEVC)
HEVC 1.0 – Low Delay Performance
HEVC 2.0 – The Extensions
HEVC Evolving fast
Summary & Outlook
Outline
Benjamin Bross
Fraunhofer
Heinrich Hertz Institute
© 20.10.2014 3
• Most recent joint video coding standard from ITU VCEG and ISO/IEC
MPEG: ITU-T Rec. H.265 ISO/IEC 23008-2 MPEG-H Part 2 HEVC
• Fraunhofer HHI contributed key techniques (Coding Block and
Residual Quadtree, Prediction Block Merging, Transform Coding,
CABAC,...)
• 50% bitrate reduction compared to H.264/AVC High Profile
• Version 1 finalized in Jan. 2013 with 3 profiles (all 4:2:0):
Main (8bit), Main 10 (10 bit), Main Still Picture
High-Efficiency Video Coding (HEVC)
Benjamin Bross
Fraunhofer
Heinrich Hertz Institute
©
HEVC 1.0 – Performance
20.10.2014 4 Benjamin Bross
What is the coding efficiency for different coding applications?
• Entertainment for TV/Movies (Broadcast, VoD, Storage)
Hierarchical B-pictures, structural delay, 1s random access
• Interactive for Videoconferencing
No picture reordering, low delay
Fraunhofer
Heinrich Hertz Institute
©
Several studies reporting bitrate savings of H.265/HEVC HM
Reference Encoder for the same objective/subjective quality:
HEVC 1.0 – Performance
20.10.2014 5 Benjamin Bross
AVC HP VP9 AVC HP VP9 Encoder
Sequences Objective
[PSNR] Subjective
[MOS] Objective
[PSNR] Subjective
[MOS] Objective
[PSNR] Objective
[PSNR] AVC VP9
[1] Ohm2012 35% 49% 40% JSVM JCT-VC
[2] Grois2013 39% 43% x264 WebM JCT-VC
(Class A,B,E, F)
[3] Rerabek2014 39% 53% 36% 49% JM WebM 4K
[4] Grois2014 41% 33% x264 WebM JCT-VC
(Class E)
Entertainment (Random Access) Interactive (Low Delay)
Fraunhofer
Heinrich Hertz Institute
©
Several studies reporting bitrate savings of H.265/HEVC HM
Reference Encoder for the same objective/subjective quality:
HEVC 1.0 – Performance
20.10.2014 6 Benjamin Bross
AVC HP VP9 AVC HP VP9 Encoder
Sequences Objective
[PSNR] Subjective
[MOS] Objective
[PSNR] Subjective
[MOS] Objective
[PSNR] Objective
[PSNR] AVC VP9
[1] Ohm2012 35% 49% 40% JSVM JCT-VC
[2] Grois2013 39% 43% x264 WebM JCT-VC
(Class A,B,E, F)
[3] Rerabek2014 39% 53% 36% 49% JM WebM 4K
[4] Grois2014 41% 33% x264 WebM JCT-VC
(Class E)
Entertainment (Random Access) Interactive (Low Delay)
Fraunhofer
Heinrich Hertz Institute
©
HEVC 1.0 – Low Delay Performance
20.10.2014 7 Benjamin Bross
Low Delay performance study from Grois et al [4]:
• Tested publicly available encoders:
• H.265/HEVC HM reference encoder
• H.264/AVC x264 r2334
• VP9 WebM v1.2.0-3088-ga81bd12
• IPPP coding structure (I-picture followed by P-pictures)
• Three JCT-VC 720p60 class E test sequences
Fraunhofer
Heinrich Hertz Institute
©
HEVC 1.0 – Low Delay Performance
20.10.2014 8 Benjamin Bross
• Encoder settings for a fair comparison:
• x264 settings for optimized PSNR
• VP9 settings recommended by Google
• VP9 not optimized for 1-pass additional 2-pass results provided
• Constant quantizer with QP 22, 27, 32, 37 Bitrates from 0.2 to 4 Mbps
• Bitrate savings in Bjøntegaard Delta (BD) rate based on PSNR for the 4 rate points
Pre-print of a paper to be published in SPIE Proceedings, vol. 9217, Applications of Digital Image Pro-
cessing XXXVII
very close, while the difference in BD-BR savings [24] of the VP9 encoder vs. x264 encoder for the FourPeople video
sequence are 2.5% in favour of VP9.
In addition, Table 7 presents detailed experimental results with regard to the HEVC bit-rate savings for the same ob-
jective quality, such as the PSNRYUV.
Figure 2. 2-pass Encoding Mode: R-D curves and corresponding bit-rate saving plots for the tested Class E sequences, which
are representing different video conferencing scenarios.
Table 7. HEVC Bit-Rate Savings for Equal PSNRYUV (Compared to VP9 and x264 High Profile Encoders, 2-Pass Mode)
HEVC vs. VP9 (in %) HEVC vs. x264 (in %) Sequences/QPs 22 27 32 37 BD-BR 22 27 32 37 BD-BR
FourPeople 16.8 28.7 33.5 39.6 -31.1 30.1 34.2 31.2 40.3 -32.7
Johnny 21.4 34.5 32.0 36.8 -34.0 35.3 53.8 56.9 60.0 -50.6
KristenAndSara 19.3 31.7 36.0 36.8 -32.6 35.8 41.1 47.4 53.8 -43.4
Averages 19.2 31.6 33.8 37.7 -32.6 33.7 43.0 45.2 51.4 -42.2
Total Average 30.6 -32.6 43.3 -42.2
Fraunhofer
Heinrich Hertz Institute
©
HEVC 1.0 – Low Delay Performance
20.10.2014 9 Benjamin Bross
• For same objective quality (PSNR), VP9 and x264 have significant bitrate overhead:
• VP9 48% more than HM
• x264 73% more than HM
• 2nd pass encoding does not gain much for low delay
• Runtime differences by factor ~103 between x264 and reference encoders (HM and VP9)
1
10
100
1000
10000
0 20 40 60 80E
nc
od
ing
Sp
ee
du
p
Bitrate overhead to HEVC HM in BD-rate [%]
VP9 (1-pass)
VP9 (2-pass)
x264 (1-pass)
x264 (2-pass)
Fraunhofer
Heinrich Hertz Institute
©
HEVC 1.0 – Low Delay Performance
20.10.2014 10 Benjamin Bross
Some thoughts on the software encoders used in the comparison...
• Only publicly available free encoders (reproducibility)
• 2-pass encoding in VP9 faster than 1-pass encoding
• All three encoders represent different degrees of software optimizations
• Extremely slow reference implementations for HEVC and VP9
• (Commercial) real-time encoders optimized for low delay are expected to achieve the coding efficiency of reference encoders much faster than in the past HEVC evolving fast
Fraunhofer
Heinrich Hertz Institute
©
What comes after version 1?
The following HEVC extensions are already finalized or to be finalized:
• Range Extensions in April 2014
Higher bit-depths (>10bit), More chroma formats (4:4:4, 4:2:2),...
• Multiview (MV) in July 2014
• Scalable Coding in July 2014
• 3D-HEVC (MV + Depth Data) to be finalized in 2015
HEVC 2.0 – The Extensions
20.10.2014 11 Benjamin Bross
Fraunhofer
Heinrich Hertz Institute
©
HEVC Evolving fast (2013)
20.10.2014 12 Benjamin Bross
2015
January
JCT-VC Meeting,
Geneva
HEVC version 1 finalized
April.
NAB,
Las Vegas
September.
IBC,
Amsterdam
HHI 4K/UHD live
SW decoder
HHI HD live SW encoder
2013 2014
Fraunhofer
Heinrich Hertz Institute
©
HEVC Evolving fast (2014)
20.10.2014 13 Benjamin Bross
2015
September.
IBC,
Amsterdam
2013 2014
February,
MWC,
Barcelona
HHI 4K/UHD live SW
encoder
HHI HEVC over LTE
HHI Encoding in 4K/UHD 50p
live soccer broadcast
April,
Allianz
Arena,
Munich
WWDC,
Cupertino
iPhone 6 with
H.265/HEVC En-
/Decoder
Fraunhofer
Heinrich Hertz Institute
©
• Several studies confirmed the 50% bitrate reduction compared to
H.264/AVC with the HM reference encoder (same subjective quality)
• Significant bitrate reductions compared to H.264/AVC and VP9 also
reported for low delay applications
• Only 1.5 years after finalization of HEVC, real-time encoders are
approaching coding efficiency of the HM encoder
• Fraunhofer HHI continues research on HEVC real-time solutions like
4K/UHD high-efficiency encoder and HD low-delay encoder
HEVC Summary & Outlook
20.10.2014 14 Benjamin Bross
Fraunhofer
Heinrich Hertz Institute
©
[1] Ohm et al, “Comparison of the Coding Efficiency of Video Coding Standards –
Including High Efficiency Video Coding (HEVC) ”, IEEE Trans. CSVT, Dec. 2012
[2] Grois et al, “Performance comparison of H.265/MPEG-HEVC, VP9, and
H.264/MPEG-AVC encoders”, PCS 2013
[3] Rerabek et al, “Comparison of compression efficiency between HEVC/H.265 and
VP9 based on subjective assessments”, SPIE Proc. 9217, 2014
[4] Grois et al, “Comparative Assessment of H.265/MPEG-HEVC, VP9, and
H.264/MPEG-AVC Encoders for Low-Delay Video Applications”, SPIE Proc. 9217,
2014
The H.265/MPEG-HEVC Standard
20.10.2014 15 Benjamin Bross
Fraunhofer
Heinrich Hertz Institute
©
Thank you very much!
Further Information:
hevc.hhi.fraunhofer.de
The H.265/MPEG-HEVC Standard
20.10.2014 16 Benjamin Bross