Professional Documents
Culture Documents
HEVC Comparison
HEVC Comparison
HEVC Comparison
265
Video Codecs
Comparison
Video group head: Dr. Dmitriy Vatolin
Project head: Dr. Dmitriy Kulikov
Measurements & analysis: Mikhail Erofeev,
Vladimir Yanushkovsky
Free version
Codecs:
H.265
Non H.265
InTeleMax TurboEnc
SIF Encoder
x264
Contents
1 Acknowledgments
2 Disclaimer
3 Overview
3.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Codecs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 Desktop Comparison
10
10
5.1.1
RD curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
5.1.2
Encoding Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
5.1.3
Speed/Quality Trade-Off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
5.1.4
Bitrate Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
5.1.5
13
5.2 Universal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
5.2.1
RD curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
5.2.2
Encoding Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
5.2.3
Speed/Quality Trade-Off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
5.2.4
Bitrate Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
5.2.5
20
5.3 Ripping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
5.3.1
RD curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
5.3.2
Encoding Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
5.3.3
Speed/Quality Trade-Off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
5.3.4
Bitrate Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
5.3.5
26
6 Conclusion
29
29
6.1.1
Fast Transcoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
6.1.2
Universal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
6.1.3
Ripping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
6.1.4
Overall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
33
6.2.1
Fast Transcoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
6.2.2
Universal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
6.2.3
Ripping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
Overall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7 Participants Comments
36
37
37
37
37
37
A Sequences
39
39
A.2 Bunny . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
41
A.4 Concert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
42
43
A.6 Developers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
A.7 Fire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
45
46
A.9 Hockey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
A.10 Kremlin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
49
50
51
52
53
A.16 Road . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
55
A.18 Sita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
A.19 Trigans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
A.20 Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
58
59
B Codecs
60
60
B.2 f265 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
62
63
64
65
66
67
B.9 SIF-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
69
B.11 x264 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
B.12 x265 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
72
C.1 RD Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
72
72
72
74
76
76
76
D.1.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
79
ACKNOWLEDGMENTS
The Graphics & Media Lab Video Group would like to express its gratitude to the following companies for providing the codecs and settings used in this report:
InTeleMax, Inc.
Intel Corporation
Ittiam Sysytems (P) Ltd.
Strongene Ltd.
System house Business partners company
SIF Encoder developper team
The WebM Project team
x264 developer team
x265 developer team
The Video Group would also like to thank these companies for their help and technical support during the tests.
DISCLAIMER
This is free version of the report which contains only overall results for desktop and server platform and detailed
analysis of codecs performance on one video sequence for desktop platform. For each individual sequence detailed plots and quality scores computed using various quality metrics please refer to Enterprise version of this
report (http://compression.ru/video/codec_comparison/hevc_2015/#version_comparison).
3
3.1
OVERVIEW
Sequences
Sequence
Number of frames
Frame rate
Resolution
1.
Apple Tree
338
30
1920 1080
2.
Bunny
600
24
1920 1080
3.
City Crowd
763
30
1920 1080
4.
Concert
1533
25
1920 1080
5.
Day Cars
1299
25
1920 1080
6.
Developers
1500
30
1280 720
7.
Fire
601
25
1920 1080
8.
Golden Statue
1993
30
1920 1080
9.
Hockey
1000
25
1920 1080
10.
Kremlin
1899
25
1920 1080
11.
Market Walk
688
30
1920 1080
12.
Mountain View
398
30
1920 1080
13.
Night Cars
1305
25
1920 1080
14.
Pine Tree
1130
30
1920 1080
15.
River Boats
1061
30
1920 1080
16.
Road
877
25
1920 1080
17.
Shake Walk
805
25
1920 1080
18.
Sita
1000
25
1920 1080
19.
Trigans
10500
30
1920 1080
20.
Water
1209
25
1920 1080
3.2
Codecs
Codec
Developer
Version
1.
InTeleMax TurboEnc
InTeleMax, Inc.
3.0
2.
0.2
3.
Intel
4.
Intel
5.
6.
1_14_8_06
7.
10.
11.
x264
12.
x265
MulticoreWare, Inc.
1.5+460-ac85c775620f
8.
9.
4.1
The main goal of this report is the presentation of a comparative evaluation of the quality of new HEVC codecs
and codecs of other standards using objective measures of assessment. The comparison was done using settings
provided by the developers of each codec. Nevertheless, we required all presets to satisfy minimum speed requirement on the particular use case. The main task of the comparison is to analyze different encoders for the
task of transcoding videoe.g., compressing video for personal use.
4.2
Testing Rules
5
5.1
DESKTOP COMPARISON
Fast Transcoding
5.1.1 RD curves
The plot below contains RD-curves for one test video sequence. For codecs performance on other video sequences please refer to Enterprise version of this report http://compression.ru/video/codec_comparison/
deo
deo
i
i
v
v
/
/
.ru
.ru
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
0.98
0.96
0.94
0.92
eo
eo
d
d
i
i
v
v
/ encoder
.ru/
.rutime
SHBP H.265
Real
n
n
o
o
i
i
ssMSS HEVC Software
ess
eIntel
r
r
Intel MSS HEVC GAcc
p
p
.com
.com x265
x264
w
w
Ittiam HEVC Hardware Encoder
ww
ww
0.9
0.88
0.86
0
6
7
8
Bitrate, Mbps
10
11
12
Better quality
hevc_2015/#version_comparison.
13
10
ssio
e
r
p
60
Encodeing speed (fps)
ide
v
/
u
.r
.com
w
w
50
essi
r
p
om
pr
m
o
w.c
w.c
w
w
0
vide
/
u
r
on.
i
s
s
e
ssio
e
r
p
.com
w
w
vide
/
u
r
on.
20
40
30
ide
v
/
u
.r
Faster
70
ww
5
6
7
8
Bitrate, Mbps
10
11
12
13
x265
Detailed descriptions of the speed/quality trade-off graphs can be found in Appendix C. Sometimes, codec
results are not present in the particular graph owing to the codecs extremely poor performance. The
codecs RD curve has no intersection with the references RD curve.
The speed/quality trade-off graphs simultaneously show relative quality and encoding speed for the encoders
tested in this comparison. x264 is the reference codec, for which both quality and speed are normalized to unity
for all of the graphs.
11
0.9
1
1.1
1.2
deo
i
v
/
.ru
1.3
deo
i
v
/
.rutime encoder
SHBP H.265
nReal
o
i
Intel
MSS
HEVC
Software
s
s
preIntel MSS HEVC GAcc
ssio
e
r
mp
1.4
.com
w
ww
.co
ww
1.5
0.7
0.75
0.8
0.85
Better
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
x265
x264
Ittiam HEVC Hardware Encoder
0.9
0.95
1
1.05
Relative Encoding Time
1.1
1.15
1.2
Better
0.8
mp
o
c
.
w
1.2
vide
/
u
r
on.
i
s
s
e
mp
o
c
.
w
ww
ww
1.4
vide
/
u
r
on.
i
s
s
e
1.6
ide
v
/
u
.r
ion
s
s
e
pr
1.8
.com
w
w
2.2
0.8
0.85
0.9
0.95
pr
m
o
c
w.
ww
1
1.05
1.1
Relative Encoding Time
eo
d
i
v
.ruReal/ time encoder
SHBP H.265
n
o
i
s MSS HEVC Software
sIntel
1.15
1.2
1.25
1.3
FIGURE 4: Speed/quality trade-offusecase Fast Transcoding, Apple Tree sequence, Y-SSIM metric
12
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
1.1
1.05
deo
i
v
/
.ru
ssio
e
r
omp
0.95
w.c
w
w
0
deo
i
v
/
.ru
n
ssio
e
r
omp
Better
1.15
w.c
w
w
5
6
7
8
Target bitrate, Mbps
Better
10
11
12
x265
Note that each number in the tables below corresponds to some range of bitrates (see Appendix C). Unfortunately, these ranges can differ signicantly because of differences in the quality of compared encoders.
This situation can lead to some inadequate results when three or more codecs are compared. Please see
Section C.4 for explanation how to read table and plot below.
13
x265
x264
100% t
58% p
58% p
69% p
66% o
67% q
210% o
100% t
100% t
120% r 106% p
120% r
210% o
100% t
100% t
120% q 106% p
120% r
x265
189% q
87% s
87% s
100% t 92% q
103% s
x264
186% r
97% t
97% t
117% s 100% t
114% s
180% q
86% s
85% s
102% r 91% p
100% t
Condence
0%
50%
100%
TABLE 3: Average bitrate ratio for a xed qualityusecase Fast Transcoding, Y-SSIM metric
Figure below depicts the data from the table above. Each line in the gure corresponds to one codec. Values
on the vertical axis are the average relative bitrates compared with the codecs along the horizontal axis. A
lower bitrate indicates better relative results.
14
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
o
vide
/
u
r
on.
essi
r
p
m
w.co
pr
m
o
w.c
Itt
ia
H mH
ar E
En dw VC
co ar
de e
r
64
H In
EV te
C lM
G SS
Ac
c
65
ww
BP
tim H.
e 265
en R
co ea
de l
r
EV
C Int
So el
ft MS
w S
ar
e
ww
50%
x2
100%
vide
/
u
r
on.
i
s
s
e
x2
150%
SH
Relative bitrate
200%
Better
Codec
SHBP H.265 Real time encoder
x264
x265
FIGURE 6: Average bitrate ratio for a xed qualityusecase Fast Transcoding, Y-SSIM metric
For visual purposes we show the same plot without SHBP H.265 Real time encoder below.
15
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
vide
/
u
r
on.
essi
r
p
m
w.co
90%
x2
H In
EV te
C lM
G SS
Ac
c
65
pre
m
o
w.c
w
w
vide
/
u
r
on.
ssi
EV
C Int
So el
ft M
w SS
ar
e
ww
Itt
ia
H mH
ar E
En dw VC
co ar
de e
r
100%
64
110%
x2
Relative bitrate
120%
Better
Codec
Intel MSS HEVC Software
Ittiam HEVC Hardware Encoder
x265
x264
FIGURE 7: Average bitrate ratio for a xed qualityusecase Fast Transcoding, Y-SSIM metric, without SHBP
H.265 Real time encoder
5.2
Universal
Now we move to Universal use case which imposes weaker speed requirements (codecs should process at least
10 frames per second on reference video sequence).
16
pr
.com
w
w
.com
w
w
0.9
es
mpr
.co
w
w
vide
/
u
r
.
n
ssio
e
r
mp
.co
w
w
InTeleMax TurboEnc
SHBP H.265 Real time encoder
Strongene Lentoid HEVC Encoder
Intel MSS HEVC Software
Intel MSS HEVC GAcc
x265
SIF Encoder
x264
Ittiam HEVC Software Encoder
Ittiam HEVC Hardware Encoder
vide
/
u
r
.
sion
0.85
0.8
ide
v
/
u
n.r
o
i
s
res
7
8
9
Bitrate, Mbps
10
11
12
13
Better quality
0.95
ide
v
/
u
n.r
o
i
s
es
14
eo
d
i
v
.ru/
35
.com
w
ww
30
25
ion
s
s
e
pr
.com
w
ww
eo
d
i
v
.ru/
ion
s
s
e
pr
deo
i
v
/
.ru
20
ion
s
s
e
mpr
o
c
.
w
ww
15
10
0
InTeleMax TurboEnc
Intel MSS HEVC GAcc
Ittiam HEVC Software Encoder
deo
i
v
/
.ru
Faster
40
ion
s
s
e
mpr
o
c
.
w
ww
7
8
9
Bitrate, Mbps
10
11
12
13
14
17
2.5
InTeleMax TurboEnc
SHBP H.265 Real time encoder
Strongene Lentoid HEVC Encoder
Intel MSS HEVC Software
Intel MSS HEVC GAcc
x265
SIF Encoder
x264
Ittiam HEVC Software Encoder
Ittiam HEVC Hardware Encoder
0.8
vide
/
u
r
.
n
ssio
e
r
mp
.co
w
ww
ww
0.7
vide
/
u
r
.
ssio
e
r
mp
o
c
.
w
0.6
.com
w
ww
ww
ide
v
/
u
n.r
o
i
s
res
Better
ess
r
p
m
w.co
1.5
ide
v
/
u
r
ion.
0.9
1
1.1
Relative Encoding Time
1.2
1.3
1.4
Better
deo
i
v
/
.ru
n
ssio
e
r
mp
w.co
w
w
InTeleMax TurboEnc
SHBP H.265 Real time encoder
Strongene Lentoid HEVC Encoder
Intel MSS HEVC Software
Intel MSS HEVC GAcc
x265
SIF Encoder
x264
Ittiam HEVC Software Encoder
Ittiam HEVC Hardware Encoder
0.8
0.9
1.1
eo
d
i
v
.ru/
n
o
i
ress
.com
w
ww
0.7
n
ssio
e
r
mp
w.co
w
w
eo
d
i
v
.ru/
n
o
i
ress
.com
w
w
deo
i
v
/
.ru
1.2
1.3
1.4
1.5
Relative Encoding Time
1.6
1.7
1.8
1.9
FIGURE 11: Speed/quality trade-offusecase Universal, Apple Tree sequence, Y-SSIM metric
18
1.5
deo
i
v
/
u
on.r
essi
r
p
m
w.co
ww
0.5
Better
deo
i
v
/
n.ru
o
i
s
es
pr
m
o
w.c
w
w
5
6
7
8
Target bitrate, Mbps
10
11
Better
InTeleMax TurboEnc
12
19
InTeleMax
TurboEnc
SHBP H.265
Real time
encoder
Strongene
Lentoid HEVC
Encoder
Intel MSS
HEVC
Software
Intel MSS
HEVC GAcc
x265
SIF Encoder
x264
Ittiam HEVC
Software
Encoder
Ittiam HEVC
Hardware
Encoder
InTeleMax
TurboEnc
100% t
55% r
35% n
33% n
31% n
30% n
80% q
34% o
34% n
35% o
SHBP H.265
Real time
encoder
197% n
100% t
61% p
57% p
53% o
51% o
167% l
61% o
N/A a
60% p
Strongene
Lentoid HEVC
Encoder
334% k
215% o
100% t
95% s
88% s
88% s
279% j
96% q
98% s
102% s
346% k
219% o
107% s
100% t
94% t
94% s
299% j
102% q
104% s
108% s
374% j
243% n
115% r
108% s
100% t
100% s
326% i
108% p
111% s
116% s
x265
367% j
237% n
115% r
108% r
101% s
100% t
319% i
109% p
112% s
116% s
SIF Encoder
134% q
65% r
38% n
37% n
33% n
33% n
100% t
39% o
36% n
38% n
x264
330% m
203% q
107% t
101% t
93% t
94% t
276% l
100% t
104% s
108% s
Ittiam HEVC
Software
Encoder
344% k
N/A a
103% r
98% r
91% s
90% s
295% i
99% p
100% t
104% s
Ittiam HEVC
Hardware
Encoder
330% k
215% o
99% s
94% s
87% s
87% s
281% j
95% p
96% t
100% t
Condence
0%
50%
100%
TABLE 4: Average bitrate ratio for a xed qualityusecase Universal, Y-SSIM metric
20
400%
o
ss
re
p
m
w.co
w
w
ide
v
/
u
r
ion.
200%
ssio
e
r
mp
o
c
.
w
100%
vide
/
u
r
.
vide
/
u
r
.
Better
ess
r
p
.com
w
w
300%
Relative bitrate
ide
v
/
u
r
ion.
n
ssio
e
r
mp
w.co
w
w
ww
Itt
x2
64
ia
m
w HE
ar V
e C
En S
co oft
de r
Itt
ia
m
H
a H
En rdw EVC
co ar
de e
r
r
En
co
de
SI
F
x2
6
I
EV nte
C lM
G SS
Ac
c
H
In
Tu Tele
rb M
oE ax
SH
nc
BP
tim H.
e 265
en R
co ea
de l
r
Le S
nt tr
oi on
d g
En HE en
co VC e
de
H
r
EV
C Int
So el
ft MS
w S
ar
e
0%
Codec
InTeleMax TurboEnc
Intel MSS HEVC GAcc
Ittiam HEVC Software Encoder
FIGURE 13: Average bitrate ratio for a xed qualityusecase Universal, Y-SSIM metric
For visual purposes we show the same plot below without InTeleMax TurboEnc, SIF and SHBP encoders.
21
ide
v
/
u
r
ion.
vide
/
u
r
.
n
ssio
e
r
omp
ia
H mH
ar E
En dw VC
co ar
de e
r
64
Itt
x2
x2
65
w.c
w
w
H In
EV te
C lM
G SS
Ac
c
EV
C Int
So el
ft M
w SS
ar
e
ww
Le S
nt tr
oi on
d g
En HE en
co VC e
de
r
85%
vide
/
u
r
.
ssio
e
r
mp
o
c
.
w
90%
.com
w
w
100%
95%
ide
v
/
u
n.r
o
i
s
s
pre
Itt
105%
ess
r
p
.com
w
w
Relative bitrate
110%
ia
m
w HE
ar V
e C
En S
co oft
de r
115%
Better
Codec
Strongene Lentoid HEVC Encoder
x264
x265
FIGURE 14: Average bitrate ratio for a xed qualityusecase Universal, Y-SSIM metric, without SIF Encoder,
SHBP H.265 Real time encoder, InTeleMax TurboEnc
5.3
Ripping
This year we have intensive competition in Desktop-Ripping nomination of our comparison, since there are 10
participants.
22
pr
.com
w
w
.com
w
w
0.9
ide
v
/
u
n.r
o
i
s
res
InTeleMax TurboEnc
SHBP H.265 Real time encoder
Strongene Lentoid HEVC Encoder
Intel MSS HEVC Software
Intel MSS HEVC GAcc
x265
SIF Encoder
VP9 Video Codec
x264
Ittiam HEVC Software Encoder
f265 H.265 Encoder
vide
/
u
r
.
sion
0.85
es
mpr
.co
w
w
n
ssio
e
r
mp
.co
w
w
0.8
vide
/
u
r
.
w
4
8
10
Bitrate, Mbps
12
14
Better quality
0.95
ide
v
/
u
n.r
o
i
s
es
16
30
ww
20
pr
m
o
w.c
deo
i
v
/
n.ru
o
i
s
es
pr
m
o
w.c
ww
eo
d
i
v
.ru/
ssio
e
r
p
10
0
0
.com
w
w
2
InTeleMax TurboEnc
Intel MSS HEVC GAcc
x264
vide
/
u
r
on.
i
s
s
e
ide
v
/
u
.r
Faster
40
ssio
e
r
p
.com
w
w
w
6
8
10
Bitrate, Mbps
12
14
16
Intel MSS HEVC Software
VP9 Video Codec
23
ess
r
p
m
w.co
1.5
.com
w
ww
ww
ide
v
/
u
n.r
o
i
s
res
InTeleMax TurboEnc
SHBP H.265 Real time encoder
Strongene Lentoid HEVC Encoder
Intel MSS HEVC Software
Intel MSS HEVC GAcc
x265
SIF Encoder
VP9 Video Codec
x264
Ittiam HEVC Software Encoder
f265 H.265 Encoder
vide
/
u
r
.
vide
/
u
r
.
ssio
e
r
mp
o
c
.
w
2.5
n
ssio
e
r
mp
.co
w
ww
ww
Better
ide
v
/
u
r
ion.
3.2 3.4
Better
deo
i
v
/
.ru
n
ssio
e
r
mp
w.co
w
w
w.co
w
w
eo
d
i
v
.ru/
ion
s
s
e
pr
.com
w
w
w
6
deo
i
v
/
n.ru
o
i
s
es
mpr
0.2
0.4
0.6
InTeleMax TurboEnc
Intel MSS HEVC GAcc
x264
0.8
ion
s
s
e
mpr
w.co
w
w
1.2
1.4
1.6
1.8
2
Relative Encoding Time
eo
d
i
v
.ru/
2.2
2.4
2.6
2.8
FIGURE 18: Speed/quality trade-offusecase Ripping, Apple Tree sequence, Y-SSIM metric
24
.com
w
w
1.5
ion
s
s
e
pr
.com
w
ww
o
ww
0.5
0
mp
o
c
.
w
InTeleMax TurboEnc
Intel MSS HEVC GAcc
x264
vide
/
u
r
on.
i
s
s
e
Better
ion
s
s
e
pr
ide
v
/
u
.r
vide
/
u
r
on.
i
s
s
e
mp
o
c
.
w
ww
4
5
6
7
8
Target bitrate, Mbps
10
11
Better
ide
v
/
u
.r
12
25
InTeleMax
TurboEnc
SHBP H.265
Strongene
Real time
Lentoid HEVC
encoder
Encoder
Intel MSS
HEVC
Software
Intel MSS
HEVC GAcc
x265
SIF Encoder
VP9 Video
Codec
x264
Ittiam HEVC
f265 H.265
Software
Encoder
Encoder
InTeleMax
TurboEnc
100% t
57% r
32% n
27% m
32% n
26% m
80% p
N/A a
33% n
N/A a
46% p
SHBP H.265
Real time
encoder
189% n
100% t
N/A a
N/A a
53% o
N/A a
161% m
N/A a
56% o
N/A a
83% o
Strongene
Lentoid HEVC
Encoder
360% j
N/A a
100% t
84% t
100% s
79% s
310% i
84% r
98% p
96% s
146% m
Intel MSS
HEVC
Software
435% i
N/A a
120% q
100% t
118% q
93% r
390% g
98% r
115% o
114% r
179% k
Intel MSS
HEVC GAcc
357% j
243% n
102% s
85% s
100% t
80% s
314% i
84% q
99% p
96% s
150% m
x265
432% i
N/A a
128% r
108% t
126% q
100% t
382% g
106% r
122% o
121% r
187% k
SIF Encoder
133% q
68% r
35% n
28% l
34% n
27% l
100% t
29% j
37% o
32% n
54% q
VP9 Video
Codec
N/A a
N/A a
122% o
103% q
120% o
95% q
388% f
100% t
115% m
116% p
179% i
x264
351% l
229% p
105% t
89% t
103% t
83% s
295% k
89% r
100% t
99% s
153% p
Ittiam HEVC
Software
Encoder
N/A a
N/A a
105% r
89% s
105% r
83% r
338% h
87% q
103% o
100% t
156% l
f265 H.265
Encoder
255% o
158% s
71% r
60% q
72% r
58% q
203% o
61% o
73% r
69% r
100% t
Condence
0%
50%
100%
TABLE 5: Average bitrate ratio for a xed qualityusecase Ripping, Y-SSIM metric
26
deo
i
v
/
.ru
deo
i
v
/
.ru
eo
Vi
d
9
x2
64
ia
m
w HE
ar V
e C
En S
co oft
de r
f2
65
En H.
co 26
de 5
r
co
d
En
VP
Itt
er
65
w.co
w
w
In
Tu Tele
rb M
SH
oE ax
BP
nc
tim H.
2
e 65
en R
co ea
de l
r
Le S
nt tr
oi on
d g
En HE en
co VC e
H
de
EV
r
C Int
So el
ft M
w SS
ar
e
H In
EV te
C lM
G SS
Ac
c
0%
w.c
w
w
n
ssio
e
r
mp
SI
F
ssio
e
r
omp
100%
200%
Co
de
300%
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
x2
Relative bitrate
400%
Better
Codec
InTeleMax TurboEnc
Intel MSS HEVC GAcc
x264
FIGURE 20: Average bitrate ratio for a xed qualityusecase Ripping, Y-SSIM metric
For visual purposes we show the same plot below without SIF, SHBP and TurboEnc encoders.
27
120%
65
En H.
co 26
de 5
r
f2
ia
m
w HE
ar V
e C
En S
co oft
de r
c
de
VP
Vi
de
o
Co
64
w.co
w
w
65
w.co
w
w
Le S
nt tr
oi on
d g
En HE en
co VC e
de
r
H
EV
C Int
So el
ft MS
w S
ar
e
60%
n
ssio
e
r
mp
x2
80%
deo
i
v
/
n.ru
o
i
s
es
mpr
vide
/
u
r
.
100%
Itt
140%
H In
EV te
C lM
G SS
Ac
c
Relative bitrate
160%
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
x2
180%
Better
Codec
Strongene Lentoid HEVC Encoder
VP9 Video Codec
x265
x264
FIGURE 21: Average bitrate ratio for a xed qualityusecase Ripping, Y-SSIM metric, without SIF Encoder,
SHBP H.265 Real time encoder, InTeleMax TurboEnc
28
6
6.1
CONCLUSION
Desktop Comparison
ww
120%
ide
v
/
u
.r
ion
s
s
e
r
100%
mp
o
c
.
94%
w
ww
.com
w
w
Itt
x2
64
SH
EV
C Int
So el
ft MS
w S
ar
e
90%
I
EV nte
C lM
G SS
Ac
c
94%
ide
v
/
u
.r
ion
s
s
e
pr
108%
65
100%
110%
x2
110%
BP
tim H.
e 265
en R
co ea
de l
r
ww
130%
mp
o
c
.
w
Better
mp
o
c
.
w
deo
i
v
/
n.ru
o
i
s
es
ia
H mH
ar E
En dw VC
co ar
de e
r
140%
152%
deo
i
v
/
n.ru
o
i
s
es
150%
Codec
FIGURE 22: Average bitrate ratio for a xed qualityusecase Fast Transcoding, Y-SSIM metric.
29
Better
deo
i
v
/
n.ru
o
i
s
es
163%
x2
64
ia
m
w HE
ar V
e C
En S
co oft
de r
Le S
nt tr
oi on
d g
En HE en
co VC e
d
Itt er
ia
m
H
a H
E rd EV
SH nc wa C
BP ode re
tim H. r
e 265
en R
co ea
de l
r
SI
F
En
co
de
r
104%
In
Tu Tele
rb M
oE ax
nc
pr
105%
m
o
w.c
w
w
101%
Itt
EV
C Int
So el
ft MS
w S
ar
e
100%
essi
r
p
com 98% 100%
91% w.92%
ww
I
EV nte
C lM
G SS
Ac
c
150%
deo
i
v
/
u
on.r
200%
65
250%
292%
eo
eo
d
d
i
i
v
v
ru/
.ru/
.254%
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
x2
300%
Codec
FIGURE 23: Average bitrate ratio for a xed qualityusecase Universal, Y-SSIM metric.
30
In
Tu Tele
rb M
oE ax
nc
I
EV nte
C lM
G SS
Ac
c
Le S
nt tr
oi on
d g
En HE en
co VC e
f2 der
65
En H.
co 26
SH
de 5
BP
r
tim H.
e 265
en R
co ea
de l
r
SI
F
En
co
de
r
H
x2
6
c
EV
I
n
C t
So el
ft MS
Itt
w S
ia
ar
m
e
w HE
ar V
e C
En S
co oft
de r
H
Co
de
de
100%
178%
deo
i
v
/
136%
n.ru
o
i
i
s
s
res97% 100% 101% 102%mpres
p
m
.co 87%
.co
82% w87%
w
ww
ww
vide
/
u
r
on.
Vi
150%
x2
6
200%
VP
9
250%
eo
eo
d
d
i
i
v
v
u/
.ru/
.r269%
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
Better
306%
300%
Codec
FIGURE 24: Average bitrate ratio for a xed qualityusecase Ripping, Y-SSIM metric.
31
vide
/
u
r
on.
SH
BP
tim H.
e 265
en R
co ea
de l
r
I
EV nte
C lM
G SS
Ac
c
ssi
e100%
r
p
96% com
w.
w
w
EV
C Int
So el
ft MS
w S
ar
e
65
100%
essi
r
p
om 93%
92% w.c
ww
vide
/
u
r
on.
64
120%
x2
140%
x2
160%
Better
163%
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
Codec
FIGURE 25: Average bitrate ratio for a xed qualityY-SSIM metric.
32
6.2
Server Comparison
110%
ssi
e
r
mp
de
i
v
/
100%n.ru
io
s
res
mp
o
c
.
ww
BP
tim H.
e 265
en R
co ea
de l
r
SH
Itt
w
5
EV
C Int
So el
ft MS
w S
ar
e
89%
Le S
nt tr
oi on
d g
En HE en
co VC e
de
r
o
87% w.c
w
w
90%
95%
x2
6
100%
deo
i
v
/
u
on.r
97%
x2
6
120%
ia
m
w HE
ar V
e C
En S
co oft
de r
130%
140%
Better
144%
deo
deo
i
i
v
v
/
/
.ru
.ru
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
Codec
FIGURE 26: Average bitrate ratio for a xed qualityusecase Fast Transcoding, Y-SSIM metric.
33
100%
pr
m
o
w.c
w
w
deo
i
v
/
n.ru
o
i
s
es
Better
de
r
En
co
SI
F
SH
BP
tim H.
e 265
en R
co ea
de l
r
93%
ia
m
w HE
ar V
e C
En S
co oft
de r
Itt
EV
C Int
So el
ft MS
w S
ar
e
essi
r
p
om
91%
87% w.c 90%
w
w
5
100%
vide
/
u
r
on.
151%
x2
6
150%
Le S
nt tr
oi on
d g
En HE en
co VC e
de
r
200%
x2
6
235%
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
Codec
FIGURE 27: Average bitrate ratio for a xed qualityusecase Universal, Y-SSIM metric.
34
200%
164%
In
Tu Tele
rb M
oE ax
nc
65
En H.
co 26
SH
de 5
r
BP
tim H.
2
e 65
en R
co ea
de l
r
SI
F
En
co
de
r
f2
4
x2
6
EV
C Int
So el
ft MS
w S
Itt
ar
ia
e
m
w HE
ar V
e C
En S
co oft
de r
Le S
nt tr
oi on
d g
En HE en
co VC e
de
r
Co
de
o
de
Vi
essi87%
r
p
com 80%
74% w.77%
ww
x2
6
100%
eo
d
i
v
/
123%
n.ru
o
i
s
100%
res
92%
p
m
.co
w
ww
vide
/
u
r
on.
150%
VP
9
250%
Better
268%
eo
eo
d
d
i
i
v
v
ru/
.ru/
.236%
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
Codec
FIGURE 28: Average bitrate ratio for a xed qualityusecase Ripping, Y-SSIM metric.
35
Better
BP
tim H.
e 265
en R
co ea
de l
r
SH
ia
m
w HE
ar V
e C
En S
co oft
de r
mp
o
c
.
w
ww
92%
deo
i
v
/
100%n.ru
o
s
es i
Itt
91%
x2
6
vide
/
u
r
on.
Le S
nt tr
oi on
d g
En HE en
co VC e
de
r
80%
essi
r
p
om85%
84% w.c
ww
EV
C Int
So el
ft MS
w S
ar
e
100%
120%
140%
x2
6
152%
eo
eo
d
d
i
i
v
v
.ru/
.ru/
n
n
o
o
i
i
ess
ess
r
r
p
p
.com
.com
w
w
ww
ww
Codec
FIGURE 29: Average bitrate ratio for a xed qualityY-SSIM metric.
36
PARTICIPANTS COMMENTS
7.1
1. We noted that the VP9 encoder in the test is running with parameter (--cpu-used=1). This setting trades off
some compression performance for the encoding speed. One can use parameter (--cpu-used=0) to improve
the compression performance by 8%-10%.
2. Although the test environment allows up to 32 cores, our current VP9 encoder is implemented to support
up to 8 cores for encoding.
3. It is denitely possible to incorporate more computing resources to signicantly speedup the VP9 encoding
process while achieving the same compression statistics. As an open source project, we certainly welcome
contributions from the video coding community to make it happen.
4. Regarding the rate control variance noted on page 25, the VP9 encode settings are optimized for VoD. This
use case values a high compression ratio over precisely matching target bitrates, so the encoder allows the
actual bitrate to be within 50% of the target. The acceptable range of variance could be reduced, with
some loss in compression performance.
7.2
Ittiam is happy that its encoder has been rated as pareto optimal in 5 out of 6 cases (across desktop and server),
and, faster by 40+% than the reference x264 encoder in the fast transcoding and universal use-cases. We would
like to point out that the congurations provided by us were not tuned for SSIM, but more for subjective viewing.
Also, in cases other than ripping, our encoder was congured for single pass and CBR, while 2-pass VBR encoding
(as done for x264) would have further improved the compression/BDRATE. The metric chosen Y-SSIM fails to take
the chroma quality into account and we hope that MSU will factor this in their next round of comparisons. In our
own comparisons against publicly available encoders, our chroma quality has been better than these encoders.
For short sequences, the look-ahead congured tends to be comparable and reects as a reduced throughput.
Also, we feel that conguring the desktop in high performance mode could have further differentiated our encoder.
7.3
x265 is developed to optimize subjective visual quality, not objective quality metrics. As such, our objective quality measurements are not as high as they would be if we optimized for PSNR or SSIM.
Encoding speed has been greatly improved in the latest version of x265 (1.7 + 478), thanks to a combination
of algorithmic improvements and other performance optimizations.
7.4
Thank you for invitation to participate in video codec test. Wed like to inform you that SIF codec version we
presented is work-in progress. Version youve tested was very close to release which was made about 3 years
ago with very limited resources. SIF Codec core has a signicant potential for improvement and we are going to
37
38
SEQUENCES
Apple Tree
Resolution
19201080
Number of frames
338
Color space
YV12
30.0
39
A.2 Bunny
Sequence title
Bunny
Resolution
19201080
Number of frames
600
Color space
YV12
24.0
40
City Crowd
Resolution
19201080
Number of frames
763
Color space
YV12
30.0
41
A.4 Concert
Sequence title
Concert
Resolution
19201080
Number of frames
1533
Color space
YV12
25.0
42
Day Cars
Resolution
19201080
Number of frames
1299
Color space
YV12
25.0
43
A.6 Developers
Sequence title
Developers
Resolution
1280720
Number of frames
1500
Color space
YV12
30.0
44
A.7 Fire
Sequence title
Fire
Resolution
19201080
Number of frames
601
Color space
YV12
25.0
45
Golden Statue
Resolution
19201080
Number of frames
1993
Color space
YV12
30.0
46
A.9 Hockey
Sequence title
Hockey
Resolution
19201080
Number of frames
1000
Color space
YV12
25.0
47
A.10
Kremlin
Sequence title
Kremlin
Resolution
19201080
Number of frames
1899
Color space
YV12
25.0
48
A.11
Market Walk
Sequence title
Market Walk
Resolution
19201080
Number of frames
688
Color space
YV12
30.0
49
A.12
Mountain View
Sequence title
Mountain View
Resolution
19201080
Number of frames
398
Color space
YV12
30.0
50
A.13
Night Cars
Sequence title
Night Cars
Resolution
19201080
Number of frames
1305
Color space
YV12
25.0
51
A.14
Pine Tree
Sequence title
Pine Tree
Resolution
19201080
Number of frames
1130
Color space
YV12
30.0
52
A.15
River Boats
Sequence title
River Boats
Resolution
19201080
Number of frames
1061
Color space
YV12
30.0
53
A.16
Road
Sequence title
Road
Resolution
19201080
Number of frames
877
Color space
YV12
25.0
54
A.17
Shake Walk
Sequence title
Shake Walk
Resolution
19201080
Number of frames
805
Color space
YV12
25.0
55
A.18
Sita
Sequence title
Sita
Resolution
19201080
Number of frames
1000
Color space
YV12
25.0
56
A.19
Trigans
Sequence title
Trigans
Resolution
19201080
Number of frames
10500
Color space
YV12
30.0
57
A.20
Water
Sequence title
Water
Resolution
19201080
Number of frames
1209
Color space
YV12
25.0
58
A.21
Water Ripple
Sequence title
Water Ripple
Resolution
19201080
Number of frames
220
Color space
YV12
30.0
59
B
B.1
CODECS
InTeleMax TurboEnc
Encoder title
InTeleMax TurboEnc
Version
3.0
Developed by
InTeleMax, Inc.
Preset name
Encoder parameters
Desktop
Ripping
Universal
Server
Ripping
60
B.2
f265
Encoder title
f265
Version
0.2
Developed by
Preset name
Encoder parameters
Desktop,
Ripping
Server
bitrate=%BITRATE_KBPS1000%" -w %WIDTH%x%HEIGHT%
%SOURCE_FILE% %TARGET_FILE%
61
B.3
Version
Intel Media Server Studio 2015 R4 Professional Edition (release date: April 2015)
Developed by
Intel
Preset name
Encoder parameters
Desktop,
Ripping
Server
Fast
62
B.4
Version
Intel Media Server Studio 2015 R4 Professional Edition (release date: April 2015)
Developed by
Intel
Preset name
Encoder parameters
Desktop
Ripping
Universal
Fast
63
B.5
Version
1_14_8_06
Developed by
Preset name
Encoder parameters
Desktop,
Ripping
ihevce_x64_sw_eval.exe -c vid_enc_cfg_FirstPass_vbr_1080p_ES_ripping.txt
Server
Universal
Fast
64
B.6
Version
2_04_4_00
Developed by
Preset name
Encoder parameters
Desktop
Universal
ihevce_x64_hw_eval.exe -c
vid_enc_cfg_Cbr_1080p_MS_hw_universal.txt --src_width
%WIDTH% --src_height %HEIGHT% --tgt_width %WIDTH%
--tgt_height %HEIGHT% --num_frames_to_encode
%FRAMES_NUM% --src_frame_rate_num %FPS_NUM%
--src_frame_rate_denom %FPS_DENOM% --tgt_bitrate
%BITRATE_BPS% -i %SOURCE_FILE% -o %TARGET_FILE%
Fast
ihevce_x64_hw_eval.exe -c
vid_enc_cfg_Cbr_1080p_ES_hw_fast.txt --src_width
%WIDTH% --src_height %HEIGHT% --tgt_width %WIDTH%
--tgt_height %HEIGHT% --num_frames_to_encode
%FRAMES_NUM% --src_frame_rate_num %FPS_NUM%
--src_frame_rate_denom %FPS_DENOM% --tgt_bitrate
%BITRATE_BPS% -i %SOURCE_FILE% -o %TARGET_FILE%
65
B.7
Version
Developed by
Strongene Ltd.
Preset name
Encoder parameters
Desktop
Ripping
Universal
Server
Ripping
Universal
Fast
66
B.8
Version
Developed by
Preset name
Encoder parameters
Desktop,
Ripping
Server
Fast
67
B.9
SIF-1
Encoder title
SIF-1
Version
1.30.4
Developed by
www.mysif.ru
Preset name
Encoder parameters
Desktop
Ripping
Universal
Server
Ripping
Universal
68
B.10
Version
1.3.0
Developed by
Preset name
Encoder parameters
Desktop
Ripping
Server
Ripping
69
B.11
x264
Encoder title
x264
Version
Developed by
Preset name
Encoder parameters
Desktop,
Ripping
Server
Universal
Fast
70
B.12
x265
Encoder title
x265
Version
1.5+460-ac85c775620f
Developed by
Preset name
Encoder parameters
Desktop
Ripping
Universal
Fast
Server
Ripping
Universal
Fast
71
C FIGURES EXPLANATION
The main charts in this comparison are classical RD curves (quality/bitrate graphs) and relative bitrate/relative
time charts. Additionally, bitrate handling charts (ratio of real and target bitrates) and per-frame quality charts
were also used.
C.1
RD Curves
These charts show variation in codec quality by bitrate or le size. For this metric, a higher curve presumably
indicates better quality.
C.2
Relative bitrate/relative time charts show the dependence on relative encoding time of the average bitrate for
a xed quality output. The Y-axis shows the ratio of the bitrate of the codec under test to that of the reference
codec for a xed quality. A lower value (that is, the higher the value is on the graph) indicates a better-performing
codec. For example, a value of 0.7 means that codec under test can encode the sequence under test in a le that
is 30% smaller than that encoded by the reference codec.
The X-axis shows the relative encoding time for the codec under test. Larger values indicate a slower codec.
For example, a value of 2.5 means that the codec under test works 2.5 times slower, on average, than the reference
codec.
C.3
Graph Example
Figure 63 shows a case where these graphs can be useful. In the top left graph, it is apparent that the Green
codec encodes with signicantly better quality than the Black codec. On the other hand, the top right graph
shows that the Green codec is slightly slower. Relative bitrate/relative time graphs can be useful in precisely
these situations: it is clearly visible in the bottom graph that one of the codecs is slower, but yields higher visual
quality, and that the other codec is faster, but yields lower visual quality.
As a result of these advantages, relative bitrate/relative time graphs are used frequently in this report since
they assist in the evaluation of the codecs in the test set, especially when number of codecs is large.
A more detailed description of the preparation of these graphs is given below.
C.4
The rst step in computing the average bitrate ratio for a xed quality is inversion of the axes of the bitrate/quality
graph (see Figure 64b). All further computations are performed using the inverted graph.
The second step involves averaging the interval over which the quality axis is chosen. Averaging is performed
only over those segments for which there are results for both codecs. This limitation is due to the difculty of
developing extrapolation methods for classic RD curves; nevertheless, for interpolation of RD curves, even linear
methods are acceptable.
The nal step is calculation of the area under the curves in the chosen interpolation segment and determination of their ratio (see Figure 64c). This result is an average bitrate ratio for a xed quality for the two codecs. If
72
Better quality
0.96
0.94
50
40
30
0.92
0
4
6
8
10
Bitrate, Mbps
12
(b)
4
6
8
10
Bitrate, Mbps
12
1.02
Better
0.98
Faster
60
1.04
1.06
0.94
0.95
0.96
0.97
0.98
Relative Encoding Time
0.99
Faster
(c) Integral situation with codecs. This plot shows the situation more clearly
73
First codec
Second codec
(b)
(a) Source RD curves
S1
S2
S1
S2
C.5
While most gures in this report provide codec scores relative to reference encoder (i.e. x264) the Relative Quality Analysis sections show bitrate ratio with xed quality (see Section C.4) score for each codec pair. This might
be useful if one is interested in comparison of codec A relative to codec B only.
Below we show simplied example of Average bitrate ratio for a xed quality table for two codecs only:
74
A 100% t 75% e
B 134% e 100% t
Condence
0%
50%
100%
75
(2x y + C1 )(2xy + C2 )
,
(x + y + C1 )(x + y + C2 )
where
x =
i xi ,
(1)
(2)
i=1
v
uN
u
x = t
i (xi x ),
(3)
i=1
xy =
i (xi x )(yi y ).
(4)
i=1
Finally, C1 = (K1 L)2 and C2 = (K2 L)2 , where L is the dynamic range of the pixel values (e.g. 255 for 8-bit
grayscale images), and K1, K2 1.
The values K1 = 0.01 and K2 = 0.03 were used for the comparison presented in this report, and the matrix
lled with a value 1 in each position to form a lter for the result map.
For the implementation used in this comparison, one SSIM value corresponds to two sequences. The value is
in the range [1, 1], with higher values being more desirable (a value of 1 corresponds to identical frames). One of
the advantages of the SSIM metric is that it better represents human visual perception than does PSNR. SSIM is
more complex, however, and takes more time to calculate.
D.1.2 Examples
Figure 65 shows the example of an SSIM result for an original and processed (compressed with lossy compression)
image. The resulting value of 0.9 demonstrates that the two images are very similar.
1 Zhou Wang, Alan Conrad Bovik, Hamid Rahim Sheikh and Eero P. Simoncelli, Image Quality Assessment: From Error Visibility to Structural Similarity, IEEE Transactions on Image Processing, Vol. 13, No. 4, April 2004.
76
(a) Original
(b) Compressed
(c) SSIM
77
SSIM = 1
SSIM = 0.552119
(c)
SSIM = 0.958917
78
79