3/20/2017
Monitoring for High Dynamic
Range and Wide Colour
MARCH 2017
Gord Langdon
Video Applications Engineer – Canada
[email protected]
4K/UltraHD
MORE PIXELS BUT WHO CARES?
• What’s more impressive?
◦
◦
◦
◦
2048x1080 2K
4096x2160 4K
3840x2160 UHD
High Dynamic Range
Wider Colour Gamut
Immersive Audio
High Frame Rate?
1920x1080
1280x720
720x480
720x576
2
1
3/20/2017
How Does Workflow Change for HDR & WCG?
• Unprecedented demand for more content to more devices
creates new pain opportunities and challenges for us
◦ ‘The trick is optimizing the content for every platform’
◦ Camera Acquisition
▪ RAW, Log and LUTs
◦ Post-Production and Delivery
▪ Colour grading, mastering
• Are ‘golden eyes’ and experience enough?
◦ What tools could help you to stay ahead of the curve?
3
How Does Workflow Change for HDR & WCG?
• In Production, how do you verify scene lighting,
camera dynamic range (15 to 16 stops), specular
highlights, etc.?
• How do you determine the average light (18%
grey) level in HDR capture (APL too high for
display)?
• In Post, how do you verify that the SDR colour
grade does not have a significantly different ‘look’
than the HDR?
• How do you verify that the HDR delivery coding
(PQ, HLG, Dolby Vision) is correct for the target
devices?
4
2
3/20/2017
Colour Gamuts
Anatomy of the Eye’s Receptors
• Rods
◦ Sensitive to blue-green light
◦ Used for vision in dark/dim conditions
• Cones - 3 Types sensitive to:
◦ long wavelengths (red)
◦ medium wavelengths (green)
◦ short wavelengths (blue)
http://webvision.med.utah.edu/index.html
6
3
3/20/2017
Colour Models – CIE Chromatcity
NTSC
(1953)
BT2020
BT709/sRGB
~SMPTE C
BT2020
D65 white
NTSC
(1953)
BT709/sRGB
(70% NTSC)
~SMPTE C
D65 white
CIE-1976 chart
(More perceptually uniform than CIE-1931)
CIE-1931 chart
• CIE xy chromaticity diagrams with 2 degree observer
◦ Still foundation of most color models
7
Colour as a Volume
Black
Black
White White
Sony
8
4
3/20/2017
Pointer’s Gamut
• 1980 research by Michael R. Pointer
• Set of real-world colours that can be reproduced using subtractive color mixing
◦ Only light that is reflected by an object, not emitted by the source
◦ Diffuse not specular reflection
www.tftcentral.co.uk
20 MARCH 2017
9
ITU 601-7 & 709-5
ITU-R BT 601-7
601-7 525
CIE x
CIE y
Red
0.630
0.340
Green
0.310
0.595
Blue
0.155
0.070
White
0.3127
0.3290
709-5/601 625
CIE x
CIE y
Red
0.640
0.330
Green
0.300
0.600
Blue
0.150
0.060
White
0.3127
0.3290
ITU-R BT 709-5
10
5
3/20/2017
DCI P3
SMPTE
RP 431-2: 2011
XYZ P3
CIE 1931 x
CIE 1931 y
XYZ P3
CIE 1976 u
CIE 1976 v
Red
0.680
0.320
Red
0.496
0.526
Green
0.265
0.690
Green
0.099
0.578
Blue
0.150
0.060
Blue
0.175
0.158
12
ITU BT.2020-2
ftcentral.co.uk
2020-2
CIE 1931 x
CIE 1931 y
2020-2
CIE 1976 u
CIE 1976 v
Red
0.708
0.292
Red
0.557
0.517
0.056
0.587
Green
0.170
0.797
Green
Blue
0.131
0.046
Blue
0.160
0.126
White
0.3127
0.3290
White
0.3127
0.3290
13
6
3/20/2017
ACES
SMPTE
ST 2065-4:2013
ACES
CIE 1931 x
CIE 1931 y
XYZ P3
CIE 1976 u
CIE 1976 v
Red
0.7347
0.2653
Red
0.6234
0.5065
0.0000
0.6000
0.0002
-0.3339
Green
0.0000
1.0000
Green
Blue
0.0001
-0.0770
Blue
14
Colour Difference Equations
SD 525/59.94/2:1 & 625/50/2:1
Y = 0.299 R +0.587 G +0.114 B
Pb = 0.564 (B-Y)
Pr = 0.713 (R-Y)
HD 1920x1080 & 1280x720 (ITU-R BT. 709)
Y = 0.2126 R +0.7152 G +0.0722 B
Pb = [0.5/(1-0.0722)] (B-Y)
Pr = [0.5/(1-0.2126)] (R-Y)
UHD1/2 3840x2160 & 7160x4320 (ITU-R BT. 2020)
Y = 0.2627 R +0.6780 G +0.0593 B
Pb = (B-Y) /1.8814
Pr = (R-Y) /14746
15
7
3/20/2017
Code Values for Colours
• 4K can use Rec 709, DCI P3, or Rec 2020 with 10 or 12 bits
• SDI digital code values for RGB primaries are the same for all spaces
◦ Translation between SDR and HDR will scale the colours
Dynamic Range Background
20 MARCH 2017
17
8
3/20/2017
Visual Dynamic Range
20 NITS = 0 STOPS = 18% REFLECTANCE
10^8
(5000 nits)8
Bright
adaption
(20nits) 0
-2
Adapted Eye
~7-stops
-4
Dark
adaption
10^-6
Nitere (Latin) = to shine, glitter, or look bright
(100nits)2.5
2
Push
Starlight
Sony, ARRI, Canon
~16-stops
starlight
10^-4
Dark Adaption
10^-2
Mesopic Scotopic
moonlight
Moonlight
4
7-stops
10^0
Adapted
Eye
indoor
lighting
(18%) 20
Bright
Adaption
10^2
24-stops with some adaption
Indoor
Lighting
Photopic
Sun light
outdoor
10^4
6
Pull
10^6
Sunlight
outdoor
-6
(.08 nits) -8
Nits (cd/m^2)
Stops
18
OETF, EOTF and OOTF
OETF
SceneReferred
Image
Data
Camera
Encoding
EOTF
Display
Decoding
Opto-Optical Transfer Function
BBC R&D
• Opto-Electrical Transfer Function
◦ transforms scene luminance to
digital code values
• Electro-Optical Transfer Function
◦ transforms code values back to
displayed luminance
Display curve is NOT inverse of capture curve
9
3/20/2017
Video OETF/EOTF for SDR
camera
tristimulus
OETF
(BT.709)
gamma
precorrect
Y’, Pb’, Pr’
encode
0.45
[M]
gamma
of Display
Y’, Pb’, Pr’
decode
delay
[NPM]^-1
tristimulus
of Display
EOTF
(BT.1886)
Filter &
quantize
LPF
Filter &
interpolate
Y’, Cb, Cr
274M 4:2:2
delay
2.4
[NPM]
Integral to CRT’s
grid drive
R,G,B
[M]^-1
G’,B’,R’
LPF
Y’,Pb’,Pr’
• BT.709 says camera output, V, is linear to 1.8% and then proportional to (light)0.45 above that
◦ Lower gain (straight line) in the blacks mitigates camera noise (not needed with new cameras)
• ITU BT.1886 says TV monitor displayed Light = (V + offset)2.4 over entire range
• Note that 1/0.45 = 2.22 and not 2.4… OOTF system gamma is not unity!
◦ Blacks are s t r e t c h e d to increase the full dynamic range
Gamma () Q&A
◦ It is caused by the voltage to current grid-drive of the CRT (not the phosphor)
▪ CRT response follows a black-level sensitive power-law
▫ Closely matches human vision brightness sensitivity
▫ Luminance = (V + black-level)^gamma
▫ Room light or black-level adjustment dramatically changes gamma
▫ Black-levels can track room lighting with auto-brightness since early ‘70s
◦ CRT gamma is fairly constant ~2.4 to 2.5, but what about flat panel displays?
▪ BT.1886 says all displays should be calibrated to 2.4
◦ Why keep using gamma power-law?
▪ Cameras do not need to be set for BT.709 gamma
▪ Human Visual Response? Not needed if a display matches the scene
▪ For SDR displays, even with WCG colorimetry, BT.1886 still applies
10
3/20/2017
Image Reproduction
IDEAL IF DISPLAY LIGHT PATTERN MATCHES SCENE
Linear
Display
Current driven CRT
or linear flat-panel
display
Need to match light output
But do not look the same!
D65 light source
Perceptually
uniform steps
NTSC & MPEG
(8 bits crushes blacks)
Perception is
1/3 power-law
“cube-root”
~100 steps (but
need15 bits to
show the small
black level steps)
Linear Camera
(natural response)
RGB
If camera response were linear,
>15-bits would be needed
and MPEG & NTSC S/N fidelity
would need to be better than ~90 dB
Lightness perception
only important for S/N
considerations
Image Reproduction
DISPLAYED IMAGE MATCHES SCENE WITH ONLY 7 BITS!
Need to match light output.
Look the same with only 7-bits
CRT response
(BT.1886)
Quantization and
noise appear
uniform due to
inverse transform
D65 light source
Perceptually
uniform steps
~100 steps (7-bits)
1 JND
1 JND
1 JND
Perception is
1/3 power-law
“cube-root”
8-bits
Camera
gamma
R’G’B’
Thanks to the CRT gamma, we
compress the signal to roughly
match human perception and only
7 bits is enough!
11
3/20/2017
System Gamma
Extended BT.709 Gamma Curve
CRT Gamma & System Curve
1
Camera precorrection is
inverse gamma
(0.45)
1
CRT V
k
gammaX V
k
Lstar V
k
Linear display
0.5
100
gammaX V
k 0.5
BT.1886 display
gammaT V
k
0
CRT gamma (2.4)
compared to total
system gamma
(1.2)
0
0
0.5
1
0
V
k
0.5
1
V
k
• Legacy system gamma is about 1.2 to compensate for dark surround viewing conditions
• Amazing Coincidence!
◦ CRT gamma curve (grid-drive) nearly matches human lightness response, so the precorrected camera output is close to being ‘perceptually coded’
◦ If CRT TVs had been designed with a linear response, we would have needed to invent
gamma correction anyway!
Why High Dynamic Range?
HDR RETAINS BRIGHT SPECULAR HIGHLIGHTS AS WELL AS DETAIL IN BLACKS
COLOURS CAN APPEAR MORE SATURATED
Sky Light: >500K nits
Direct sunlight >
1 billion nits
(don’t look at it)
Laptop or TV: 100 to 200 nits
Hard to see in bright daylight
Shadows: .1 to 10 nits
With the day-adapted eye, shadows can be 10 nits;
in a living room, less than 0.1 nits
12
3/20/2017
HDR for Camera Acquisition
Camera Acquisition
“We’d like to see what’s outside the window as well as what’s inside the room.”
13
3/20/2017
Camera RAW and Log
• Shooting RAW
◦ Camera photosite sensor data before any processing
▪ No white balance, ISO or colour adjustments
▪ 12 to 16 bit depth
▪ Not viewable directly on a monitor – must convert to video
De-Bayer
▫ De-Bayering process combines brightness + colour → RGB
R
G
B
• Shooting Log
◦ Maximizes captured sensor data using a logarithmic gamma curve
▪ Includes processing information
▪ Video formats specific to camera manufacturers
▪ Looks washed out on a monitor
▫ Use a Look Up Table (LUT) to transform for viewing
28
Log Gamma
• Log gamma on modern cameras allows you to shoot a ‘flat’ image,
capturing more details in both highlights and shadows
• While a flat image may not look pleasing while on set, it provides
more freedom for colour grading in post-production
709
Log
29
14
3/20/2017
Capturing Camera Footage
• Setup your test chart within
the scene
Specular
Highlights
18% Grey
• Adjust the lighting to evenly
illuminate the chart
• Adjust the camera controls to
set the levels
◦ ISO/Gain, Iris, Shutter, White
Balance
90% Reflectance
White
Super
Black
Reflective
Black
Camera Log on 709 Waveform Monitor
mV on left
% on right
15
3/20/2017
Camera Log – Code Values, Nits, %
Gamma
0% Black
10-bit CodeValue
%
18% Grey
(20 nits illumination)
10-bit Code-Value
3
394
90% Reflectance
White
10-bit Code-Value
%
37.7
636
65
%
S-Log
90
S-Log2
90
3
347
32.3
582
59
S-Log3
95
3.5
420
40.6
598
61
Log C Arri
134
3.5
400
38.4
569
58
C-Log Canon
128
7.3
351
32.8
614
63
V-Log Panasonic
128
7.3
433
42
602
61
Red Log
95
4
468
46
671
69
BMD Film
95
4
400
38
743
78
ACES (proxy)
ND
ND
426
41.3
524
55
BT.709
64
0
423
41.0
940
100
HDR Camera Monitoring – LUT vs F-Stops
4525 nits
+8
HDR (16-stops)
Studio Control Router
Camera Control
Reference Pix Monitor
Portable Recorder
Etc.
Live Camera Interface
(SDI / IP)
HDR
(90%)100 nits
2.3
(18%) 20 nits
0
BT.709
SLog1
SLog2
SLog3
Canon Clog
ARRI LogC
Custom
SDR/HDR
BT709 View LUT
Stop waveform
Independent of
Camera Format
Tektronix WFM (aka Scope)
0.07 nits
-8
FORMAT
S-Log3
F-stop Waveform (stop/Nit scale)
mV Waveform SLog3
800
700
Log2 scale
700
600
Vmv Luma_Slog3_8b
500
L, n
Fstop Iyi
L, n
400
Vmv Iy i
L, n
HDR Live Scene - up to 16 stops
(white clipped 8-bit BMP ~10 stops)
Depends on
Camera Format
300
200
100
0
0
100
0
200
400
n
The only common
reference for distinguishing
HDR from SDR is Light
S-Log3 to BT.709
LUT
600
800
8
7
6
5
4
3
2
1
0
1
2
3
4
5
6
7
8
(SLog3 to Stop LUT)
2.5
0
0
200
400
600
800
n
Raw S-Log3
33
16
3/20/2017
Simple LUT To Get Stops Waveform
CONVERSION LUTS FOR CAMERAS COMPARED TO BT.709
Stop values (0 = 20 nits, 18% reflection)
Camera to f-stop (light), 0=18% refl
8
420
7
940
6
5
4
3
LUT_Slog1_fstop
ia
2.5
2
LUT_Slog2_fstop
ia
1
LUT_Slog3_fstop
ia
0
0
LUT_BT709_fstop
ia
1
2
3
4
5
6
7
8
0
200
400
600
800
1000
10-bit CodeiaValues
34
Waveform Graticule in Code Values and Stops
Digital Code
Values on left
F-Stops
on right
17
3/20/2017
HDR for Delivery
36
HDR – Viewer Preferences
Dolby
18
3/20/2017
HDR EOTF for No Visible Banding – 10 or 12 bits?
• ‘Barten Ramp’
◦ Contrast Sensitivity
versus Luminance
Dolby
• Dolby ‘Perceptual Quantizer’ curve
◦ Efficient use of codewords
HDR EOTF for No Visible Banding – 10 or 12 bits?
Dolby
• Maybe 10 bits are enough?
19
3/20/2017
HDR Standards
SMPTE
• ST 2084 – EOTF for Perceptual Quantization (PQ)
▪ Adopted from Dolby Vision
• ST 2086 – Static metadata for colour and dynamic range of the mastering
display
▪ E.g., DCI-P3 gamut, D65 white point, with max specular reflectance of 10k nits
• ST 2094 draft – Dynamic metadata Master Display Color Volume
▪ Content-dependent metadata to enable compatibility with legacy displays
▪ ~15% extra bandwidth for metadata distribution (in HEVC or AVC)
• SMPTE IMF – ST 2067-21 (App 2 Extended)
▪ Can signal ST 2084 and carry ST 2086 metadata
▪ Next step is a carrier for ST 2094 dynamic metadata
20
3/20/2017
SMPTE
• SMPTE IMF – ST 2067-21 (App 2 Extended)
▪ Can signal ST 2084 and carry ST 2086 metadata
▪ Next step is a carrier for ST 2094 dynamic metadata
• ST 352 SDI VPID –video payload identifier
◦ Progressive/interlace, picture rate, sampling
structure, aspect ratio, bit-depth
◦ Colorimetry, HDR mode
◦ 4k/UHD quad-link 2SI or square-division mapping
CTA
• HDR10 Media Profile
◦ Adopted by the Blu-ray Disc Association (BDA) for 4K Ultra HD
◦ Based on ST 2084 PQ
▪ 10-bit, BT.2020 colour
▪ Mastered over a range of 0.05 – 1000 nits (20,000:1)
▪ Static metadata between source and display
▫ ST 2086 from mastering display
▫ MaxFALL – Max Frame Average Light Level
▫ MaxCLL – Max Content Light Level
21
3/20/2017
Dolby Vision
• Licensed end-to-end solution based on ST-2084 PQ curve
◦ Mastered over a range of 0.0001 – 10,000 nits (1,000,000:1)
▪ 12 bits, Rec 2020 colour
◦ DV-mastered content, played from a DV-capable source, sent to a
DV-capable display
◦ Display includes a chip that identifies its output capabilities (light
output, colour space, etc.), which it passes as metadata back to the
source
▪ Dynamic metadata is possible on a per frame or per scene basis
▪ Base layer of SDR content encoded with a Gamma EOTF decodable by
any streaming video decoder
▪ Enhancement layer metadata tells the DV decoder how to do SDR to HDR
conversion
▫ Adds ~15% to the size of the bitstream
44
Hybrid Log-Gamma (HLG)
• Hybrid Log-Gamma (from BBC/NHK)
◦ Seamless ‘gamma’ power-law processing in blacks
similar to BT.709 below 100 nits
◦ Extends log processing of high brightness peaks
up to 3 stops to mitigate blown-out or clipped
whites
◦ Can be displayed unprocessed on an SDR screen
▪ Does not require mastering metadata
▪ EOTF adjusts system gamma to correct for viewing
environment (10 to 500 nits)
Wikipedia
◦ Good for both SDR & HDR live production
• Standards
• ARIB STD-B67(Japan)
• ITU-R BT.2100 and DPP (UK) approved both HLG and ST-2084 PQ
22
3/20/2017
SDR vs PQ vs HLG
SDR
HLG
PQ
• HLG has similar power-law gamma
to SDR in dark range, then
changes to a log gamma above
~14 nits
• PQ has more gain in dark range
but is less compatible with BT.1886
SDR
SL-HDR
ETSI TS 103 433
• Jointly developed by Technicolor, Philips, STMicro, CableLabs
• Workflow to grade both HDR and SDR
• Backwards compatibility
◦ metadata can reconstruct an HDR video from an SDR stream
▪ added to HEVC or AVC via SEI messages
▪ ‘single layer’ video stream
47
23
3/20/2017
HDR – Evolving Consumer Standards
• UHD Alliance
◦ SMPTE ST-2084 EOTF
◦ Mastering:
▪ 100 percent of P3 colors must be displayed
▪ HDR range >1,000 nits peak brightness and <0.03 nits for black
• UltraHD Premium™ TV branding specs
◦ Either >1000 nits peak brightness and <0.05 nits black
◦ OR >540 nits peak brightness and <0.0005 nits black
▪ More than a million to one ratio? How can you get that black?
◦ At least 90% of DCI P3 colour space
HDR and WCG Challenges
49
24
3/20/2017
How Does Workflow Change for HDR & WCG?
• In Production, how do you verify scene lighting,
camera dynamic range (15 to 16 stops), specular
highlights, etc.?
• How do you determine the average light (18%
grey) level in HDR capture (APL too high for
display)?
• In Post, how do you verify that the SDR colour
grade does not have a significantly different ‘look’
than the HDR?
• How do you verify that the HDR delivery coding
(PQ, HLG, Dolby Vision) is correct for the target
devices?
50
SMPTE ST-2084 PQ HDR versus BT.1886 SDR
100X Brighter @ 100 IRE
maxi( x)
if( x< 0 , 0 , x)
1 10
2610
4096. 4
m1 = 0.159301757813
m2
2523.
128
4096
m2 = 78.84375
c1
c3
c1 = 0.8359375
m1
c2
1
c2
c3
2413.
32
4096
c2 = 18.8515625
2392.
32
4096
c3 = 18.6875
230
1 10
Note that ST.2084 has more gain
above code value 230. This
means more quantization error
of bright light. However, the eye
is less sensitive to changes in
bright regions.
Conversely, it has less gain
below 230 so there is less
quantization error in the blacks
where the eye is most sensitive.
Light output (nits or cd/m^2) vs CV
4
940
3
100
Lpq = 10000 nits
10
1
PQ EOTF:
Lpq.
maxi v
m2
c1
Lo
100
Y1886( v )
1
c2
YoPQ
m1
1
Ypq( v )
BT1886 EOTF:
nits

2.4
Lo . ( maxi( v
0
b))

i
1
Yo1886
i
Light output (nits or cd/m^2) vs CV
0.1
0.1
m2
c3. v
0.01
PQ OETF:
Ypq_inv( v )
b
c1
m1
c2. v
1
m1
c3. v
0.01
YoPQ
m2
1 10
1 10
NOTE:
PQ OETF is inverse of EOTF, so
system gamma is unity
3
i
Yo1886 1 10
i
4
1 10
3
4
100 IRE
1 10
5
200
400
600
cv
i
800
1000
1 10
5
70
80
90
100
cv
i
10-bit Code values
51
25
3/20/2017
ST.2084 (PQ) Dynamic Range
HDR MAPPING INTO CAMERA F-STOPS (0 = 20 NITS, 2.5 = 100 NITS)
Light-Stop Histograms (0=18%)
(5000 nits)8
Bright
adaption
HDR Only Zone
8
6
4
6
2.5
Adapted Eye
~7-stops
-2
2
stop1886
BT709 on BT1886 Pix
• HDR coding on BT.1886 display will
look washed out
k
0
stopPQ
k
0
• With an HDR master, DR
compression can improve the SDR
picture on a well calibrated BT.1886
display
2
-4
HDR on BT1886 Pix
4
Dark
adaption
-6
• BT.709 gamma on BT.1886
calibrated display stretches blacks
and actually increases DR
4
(100nits)2.5
2
(20nits) 0
• HDR coding on HDR display is best
match to viewing scene
HDR on HDR Pix
(matches original
with only 7-bits)
• HDR coding does not change APL
since it only provides more light in
the highlights and more light-level
accuracy in the blacks
6
(.08 nits) -8
8
4
4
1 10
2 10
H1886 , Hpq
k
k
0
Stops
3 10
4
52
Scene Referred 709 to PQ LUT Conversion
Camera-Side Conversion
BT.709
toBT.709
PQ to PQ
Camera-side
conversion
100
2084 HDR
709
100nits
HDR
1000nits
0%
2%
18 %
90%
100%
0
9
41
95
100
0
37
58
0
31
51
HDR
5000nits
0
24
42
41
80
SDR BT.709,100,1000
70
SDR BT.709,100,2000
60
SDR BT.709,100,5000
SDR_2_HDR_CS BT709 , 100 , 1000
i
75
76
SDR_2_HDR_CS BT709 , 100 , 2000
i
HDR
2000nits
9
90
SDR_2_HDR_CS BT709 , 100 , 5000
i
68
68
58
59
50
40
30
20
10
0
0
20
40
60
80
100
BT709
BT.709
%i IRE
SDR and HDR displays DO NOT match.
% or IRE
Blacks are stretched in the BT1886 Display but not in the PQ display (matches scene)
53
26
3/20/2017
Side by Side 709 and PQ
709
HDR PQ
SMPTE 2084 PQ Look Up Tables
Linear Ramp Test Signal
BT.709
LUT
ST-2084 1000 Nits
Ref White 100 nits
LUT
ST-2084 1000 Nits
Ref White 300 nits
55
27
3/20/2017
Normal or Diffuse Reflectance White Point
• Same challenge as with different camera formats
◦ E.g., a diffuse white point of 100 nits is set at 61% for S-Log3, 58% for
Log C, and at 63% for C-Log
• HDR PQ
◦ No agreement on Diffuse White point level
◦ Many are using 100-200 nits as the default; others use 10k nits
• Camera operator or colorist/editor must also know what reference
monitor will be used for grading the content
◦ E.g., If a 1000 nit monitor is used for grading, a diffuse white point of
100 nits is set at 75 % for SMPTE ST 2084 (1K)
◦ If a 2000 nit monitor is used, diffuse white is set at 68 %
Relative Code Value
HDR Aligned at 100% Diffuse = 100 Nits
Nits
57
28
3/20/2017
Code Values - Narrow or Full?
• Standard ITU-R BT.2100
◦ Full Defined
▪ Luma Y 0d –1023d for 10-bit
▪ Chroma Cb/Cr 0d –1023 for 10-bit
▪ 0d – 4092d for 12-bit Luma
▪ Chroma Cb/Cr 0d – 4092d for 12-bit
▪ Note:- SDI codewords excluded and range clipped
Narrow Defined
▪ Luma Y 64d –940d for 10-bit
▪ Chroma Cb/Cr 64d – 960d for 10-bit
▪ 256d – 3760d for 12-bit Luma
▪ Chroma Cb/Cr 256d – 3840d for 12-bit
58
Potential Issues with Bright HDR displays
• Colour shift in the Mesopic-level adaption
• As light moves below Photopic (dominated by cones) and gets closer to Scotopic
(dominated by rods) colour saturation will diminish
• This may occur in dark scenes in low-light home theatres
• Light/Dark Adaption (‘bleaching’ process rather than pupil size)
• Sustained bright images cause the photopigment in the retina to reduce and can result
in the perception of after-images
• Dark adaption can take seconds or even minutes
• Changes from bright to dark scenes may take longer in dark theatre as opposed to
same scene in higher ambient light
• Viewing Distance
• Static adaption is only about 7 to 9 stops
• To take full advantage of HDR (>9 stops) with local adaption, you need to be closer
than 2 screen widths (eye strain risk)
• Large Area Flicker
• Strobing of high peak light levels may cause distress (PSE BT.1702)
• Perceptual flicker frequency may be increased since it is a function of retinal adaption
• Frame rate judder may be more visible
59
29
3/20/2017
Monitoring for HDR and WCG
20 MARCH 2017
60
Monitoring for Wide Colour
• Tools:
◦ Waveform, vector, gamut displays
• Measurements:
◦ RGB level balance checks
◦ Vector targets are slightly different
◦ Ensure within legal gamut limits if content will be
down-converted from Rec 2020 or P3 to 709 or 601
colour
30
3/20/2017
Monitoring for HDR
• Tools:
◦ Test patterns, charts, gadgets for camera alignment
◦ Viewing LUTs?
◦ Waveform monitor
• Measurements:
◦ Depends on HDR flavour
▪ ST 2084/PQ, HLG, Dolby Vision
▪ Mastering reference (540, 1000, 1200… 10,000 Nits?)
◦ Set points will change for normal white, specular
highlights, 18% gray, reflective black and super black
◦ Metadata?
▪ ST 352 VPID; ST 2086 static; ST 2094 dynamic
Monitoring for HDR
• Waveform monitor scales for different HDR formats
◦ EOTF curves (PQ, HLG, S-Log, C Log…)
◦ Units of nits, % reflectance, f-stops, code values
◦ Mastering reference levels for HDR delivery (1k, 1.2k, 4k, 5k, 10k nits)
31
3/20/2017
Monitoring for HDR
• Zebra pattern on picture monitor for specular
highlight areas above diffuse white
• Luminance histogram
40 nits
• Average picture level
Waveform Monitoring for HDR and WCG…
• Makes camera set-up easier
• Makes colour grading faster and more
accurate
• Ensures that you master correctly for all
delivery formats
• Helps you…
Stay ahead of the curve!
32
3/20/2017
33