HDR10+ Technology & Content Tools

SAMSUNG – July 2018

1
 HDR Supported by HDR10+
 HDR systems deployed since 2015.
 Studio masters delivered in “PQ” encoding. “Perceptual Quantizer”. PQ Enables content creators
to exactly specify the image color and brightness as viewed on a reference monitor in the grading
suite. – it’s up to displays to play this directly or adapt to the consumer’s viewing environment.
 Premium Certification of HDR content, distribution and reproduction was announced by the
Ultra-HD Alliance in December 2015.

 HDR10+ metadata delivers scene

specific image information enabling
TVs to better handle HDR10
material on a wide range of displays
and viewing conditions.
What are HDR10+ and ST 2094-40?
PQ
Perceptual Quantization
SMPTE ST 2084
“HDR Essence” Same for all HDR

HDR10 (for distribution)

10-bits
(wide color) BT.2020
(HDR) Static Metadata
SMPTE ST 2086
MaxFALL & MaxCLL

Open HDR format (2015)

Dynamic Metadata
SMPTE ST 2094-40
Statistical Information
Basis OOTF for Guided Tone Mapping
HDR10+LLC
(certified 2094-40 profile; see whitepaper)
Open Technology to extend reach of HDR
into mid & lower tier televisions
 Dynamic Tone Mapping
 Static Tone Mapping(HDR10): Tone mapping using static metadata(one set of metadata for entire
content) such as Mastering display information(ST 2086), MaxCLL, MaxFALL

 Dynamic Tone Mapping(e.g. HDR10+): Tone mapping using dynamic metadata (per scene)

500nit

Display peak: 500 nit

1,000nit

Scene Peak: Tone mapped

from 100nit to 1,000nit Static Tone Mapping @ 500 nit

Use different
curve per scene
By Pass
Scene Peak: 100nit Display peak: 500 nit

Tone mapping
Scene Peak: 1,000nit
@ 500nits
Dynamic Tone Mapping
6 months - December 2017 Amazon

• Amazon status: HDR10+ on all HDR content.

Certification Program Opened June 20th 2018

 Technical Spec
Test Spec
Logo Usage
Different adopter classes
 Content
 Source
 Display
 SoC
 Tools
 Information: “[Link]”
 HDR10+ LLC
HDR10+LLC is a profile, logo and certification program formed by Samsung, Panasonic and 20th
Century Fox Studios.
• HDR10+ LLC Profile follows 2094-40 supporting statistics and guided tone curves. Two statistical values are
interpreted as new statistical image parameters.
• HDR10+ does not specify implementations

HDR10+ LLC Specifies tests used to

Content Distribution Consumer
certify picture quality and awards
logo to devices that pass
• Panel Performance (e.g. gamut,
accuracy)
• Processed image quality
TV &
SoC • ATC(s) perform the testing
STB

Tools implementing Adopt HDR10+ and Adopt HDR10+ and implement tone mapping
spec used to create incorporate the with HDR10+ metadata.
deliverable metadata metadata in their • Utilize IPR outside of HDR10+ to process
HDR10 transmissions • Follow HDR10+ HDMI VSIF for v2.0a/b
 Benefits of HDR10+ Ecosystem
 Solves problem of gap between TV performance and content mastering char
acteristics
o Volume zone displays 300-500 nits
o Premium zone displays ~1000 nits
 Provides better qualified metadata than simply SMPTE 2086
 Wide reach of HDR content across consumer UHD displays
 Quality bar based on technology and certification
 Adoption based on open technology
 HDR10+
Metadata Steps

JSON encoding

 Image analysis:  JSON processing:  ‘SEI’ processing:

 Identify scenes  Post Production workflows, clip  Binary representation of
 Compute statistics across scene management etc… metadata used by
 Offline encoder support encoders
 Compute Bezier from statistics
 Represents metadata close to SEI  ‘VSIF’ processing (optional)
 Output JSON
format yet directly readable  Binary representation
used by HDMI
 Derived from SEI

(Image courtesy Joe Kane Productions)

 HDR10+
Post Production based Metadata Workflow

Offline encoder for New product ready for

Blu-ray or OTT distribution
1.) Generate JSON from existing
source asset (e.g. ProRes)

2.) (optional) Preview default

performance at target
deliverable (e.g. 400 nits)
3.) (optional) Adjust any scene
if desired (e.g. combine/split)

(Image courtesy Amazon Studios & ColorFront)

 HDR10+
Straightforward Metadata Workflow & Upgrade

Insert metadata with

command line tool

1.) Generate JSON from existing

source asset (e.g. ProRes)

2.) (optional) Preview default New HEVC track ready

performance at target for distribution
deliverable (e.g. 400 nits)
3.) (optional) Adjust any scene
if desired (e.g. combine/split)

(Image courtesy Amazon Studios & ColorFront)

 LiveHDR10+
Workflow

HDR10 Encoder

HEVC
Realtime Encode
To live
produced broadcasting
content SEI
Insert

HDR10+ Metadata
Generation Block

Generation algorithm included in HEVC

encoder
SEI directly produced (skips JSON step)
 ColorFront – NAB 2018
3C HDR10+ support: Demonstrated to studios
Main Themes:
• Metadata generation is
easy.
• Metadata can be easily
validated for HDR10 and
HDR10+ at same time
• This is similar QC process

(Image courtesy Amazon Studios & ColorFront)

 PR: Black Magic Design – NAB 2018
Launched HDR10+ support in Resolve 15
 PR Quote: Support for native HDR10+™ controls in Da Vinci Resolve Studio
 Scenarist & ATEME Blu-ray HDR10+
HDR10+ support: Demonstrated Deluxe authored HDR10+ UHD Blu-ray disc
playback across HDMI at NAB

(Image courtesy Scenarist)

 Summary
 HDR10+ Tools ready
o Mastering
o Post Production
o Encoding
o Authoring
 Upgrade path for HDR10 content
 Important HDR10+ Technology Features
 Scene based statistics
 Option for preview of result (professional uses)
 Samsung implementation provides for superior content to TV adaptation
 Our solution covers tone mapping and TV makers performs panel
adaptation to finalize the processing.

HDR10+
 Improved scene statistics over HDR10
 MaxCLL is fine for use as a static “target”, but is too sensitive for dynamic use.
 HDR10+LLC fixes to 7 percentiles:
{1%, 25%, 50%, 75%, 90%, 95%, 99.98%}.
 This solution is robust and tolerant of spurious pixels. (random extremely bright pixel/s)
 Two new parameters are stored in the percentile vector to follow 2094-40 syntax:
 {1%, llc, llc, 25%, 50%, 75%, 90%, 95%, 99.98%}

Example 11% 250 nits:

• Results in 90%, 95%, 99.98% as “250”
 Guided Tone mapping with Bezier curve technology
 The Bezier curve allows smooth, yet nuanced, mapping
 Critical tones can be preserved
 “In-between” areas can be compressed
 Provided in metadata (can be creatively controlled)
compressed

Target Luminance

compressed
Sky detail preserved

compressed

Face details preserved

Knee Point

Linear range Shadows preserved

Source Luminance
Tone Mapping OOTF
 Bezier anchor points can be provided
 Calculated at same time as statistics
 Can be creatively controlled or created.
 Used in all Samsung TV implementations
 Available from Samsung royalty free to SoC and TV makers

Example 11% 250 nits:

• Results in straight line guide curve (aka
“Bypass”)
 Bezier Curve

 Example : Bezier Curve

Creative
Intent Line
 Why Bezier Basis Guided OOTF
 Continuity of slope : Bezier Curves

Creative
Intent Line

We can make the slopes

continuous and preserve
the creative intent in
shadow area.

Knee Point
 Device Implementation: Guided OOTF
 Transmit the basis curve and mix with product!

OOTF(1000,700)?

700 1

Product Display
0.8
(P=700 nit)
0.6

0.4
M
700nit
nit
700 nit
OOTF(1000,300) 0.2

0
0 0.2 0.4 0.6 0.8 1

1000

Basis OOTF(1000,300) is Basis OOTF Metadata:

transmitted to device (𝒌 = 𝑘𝑥 , 𝑘𝑦 , 𝐩 = (p1 , 𝑝2, … , 𝑝𝑁−1 ))

Metadata
Parser
 Bezier vs Gamma
 The Bezier curve allows avoidance of slope discontinuities
 Slope discontinuities can cause sharp contour artifacts
 Tone mapping in display should not be done using gamma as it is very coarse

Target Luminance

No slope
discontinuity
(Bezier)

Source Luminance

Tone Mapping OOTF

 Gamma based adaptation is difficult

Creative
Intent Line

No Creative Intent
Preservation No performance maximization

This slope discontinuity

can introduce banding
artifacts after the tone
mapping

Knee Point
 Summary
 Maximizes creative intent across various displays
 Enables content-specific mapping, avoids over aggressive static tone curves
commonly found in displays
 Shadows fully preserved
 Ensures a low-APL baseline (e.g. will track PQ on low APL images)
 Minimizes artifacts
 Hard slope changes can be avoided
 Improves robustness of statistical measurements over static
 Tolerant of spurious pixels
 Workflow compatible
 Manual or Automatic metadata generation for offline or live use cases
 Deployment is being established across HDR ecosystem
 Tools, Software, SoC, and Display