VideoQ, Inc. Presentation www.videoq.com HDR Test Patterns VideoQ, Inc. Training Presentation May 2017 www.videoq.com 2 www.videoq.com www.videoq.com Part 1 – General HDR Info The following 8 slides contain the background information, helpful for the optimal selection and usage of VideoQ HDR Test Patterns
26
Embed
HDR Test Patterns - VideoQvideoq.com/Downloads/VideoQ HDR Test Patterns Training Presentation...VideoQ, Inc. Presentation HDR Test Patterns VideoQ, Inc. Training Presentation May 2017
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
VideoQ, Inc. Presentation
www.videoq.com
HDR Test PatternsVideoQ, Inc. Training Presentation
May 2017www.videoq.com
2www.videoq.comwww.videoq.com
Part 1 – General HDR Info
The following 8 slides contain the background information,
helpful for the optimal selection and usage of VideoQ HDR Test Patterns
SDR & HDR: Long-term Peaceful Coexistence ?• Currently there are 4 types of video data formats in use: the ubiquitous SDR format, widespread camera
RAW LOG format (de-facto it is the HDR format) and at least two new HDR formats partially defined in ITU BT.2100, SMPTE ST 2084/2086/2094 and other standards:
• HDR-LOG (Camera RAW Video), commonly “anchored” to 18% Gray Light LevelAbsolute Light Levels in nits can be retrieved via Exposure Index (ASA) metadata
• HDR-PQ (Perceptual Quantizer), covering a wide Light Levels Range from 0.001 nits to 10,000 nitsSome experts further split it into Dolby Vision, HDR10 and HDR10plus, though some others see it as one standard
• HDR-HLG (Hybrid Log Gamma), based on the SDR concept of Relative Light LevelsThe HDR-HLG format is promoted as backward compatible to SDR production and distribution, but this is debatable
• Rendition of HDR-PQ and HDR-HLG images by HDR displays depends on a number of factors; production of HDR content in a single format suitable for any HDR display is scarcely possible.
• The critically important Reference White (aka “Diffuse White”) level definition is still in discussion.
• The key tool for setup, calibration, and benchmarking of display devices, video processors, encoders and video production software is the VideoQ HDR Test Patterns Suite.
4www.videoq.comwww.videoq.com
Dynamic Range Conversion: Essential for HDR Progress• Mixed SDR/HDR environment requires SW and HW engines for auto-enhancement, up-, down- and
cross- conversion within and/or between all four above-mentioned formats.
• All of these processes are related to the optimal choice of mezzanine DR format and architecture of vision mixers/switchers operating in such SDR/HDR environment.
Narrow (aka “Broadcast”) Video Data Range, e.g. 10 bit values 64 … 940
0
0 1
1
Relative
Light
Level
Ref. Black
Full Video Data Range, e.g. 10 bit values 0 … 1023
Relative Signal Level: 0 …1
Similar to SDR TV, in HDR-HLG format
the relative Y, R, G or B Signal Levels (SL)
convert to the relative Light Levels via the
Electro-Optical Transfer Function (OETF).
The HDR-HLG OETF formula is a combination
of the SDR curve reduced to 75% of the signal
range and a “soft knee logarithmic function”,
occupying upper 25% of the signal range.
Full HDR-HLG EOTF formula can be found in
the ITU-R BT.2100 standard (Table 5).
For the HDR-HLG format, the Max Relative Light
Level (LL = 1) is 3.8 times higher than the
Reference White Light Level. In other words Ref.
White (“Diffuse White”) LL is defined as 26.3 %
of the max possible display light output.
Note that, unlike HDR-PQ, HDR-HLG signal can
be rendered by any display, even by any old-
fashioned SDR display, though, for best results
some contrast adjustment may be needed.
Y
0 64 940 1023
RGB This sub-range is used
mainly for specular
highlights – on condition
that they should not take a
significant share of the
screen area!
LL = 0.263
HDR-HLG
SDR
720
Ref. White
Ref. White
HDR-HLG Format Limit: LL = 1 (100%)
10www.videoq.comwww.videoq.com
Light Level Models and Units of Measurement1. Classical CIE standard definitions of Luminance and related TV screen pixel Brightness parameter (in cd/m2)
should not be confused with de-facto parameter of SDR & HDR Light Level, which regretfully is often expressed in the same units, such as cd/m2.
2. Light Level values are useful only within the assumed mathematical model, created for a particular purpose:• By default, the purpose of the SDR/HDR model is to simulate a human observer estimation of color image intensity.
For this purpose the EOTF(max(R,G,B)) operator is applied to all pixels of the video frame, e.g. to get widely used Frame-Average Light Level (FALL) value in nits.In accordance with this model the LL value (perceived intensity) for the Color Bars Test Pattern Blue Bar is equal to the White Bar, whilst CIE Brightness value
for the Blue Bar is more than 13 time lower than for the White Bar.For RGB color spaces that use the ITU-R BT.709 primaries CIE Brightness function is 0.2126xEOTF(R) + 0.7152xEOTF(G) + 0.0722xEOTF(B).The formula reflects the luminosity function: green light contributes the most to the intensity perceived by humans, and blue light the least. But in accordance with the abovementioned max(R,G,B) alternative model they produce equal contributions!
• If the model’s purpose is the estimation of OLED display power consumption for various video content samples, then for each video clip the (EOTF(R) + EOTF(G) +EOTF(B))/3 operator is applied to all pixels of all frames. Obviously, in this case the LL numerical values (in nits) may differ significantly from the case above.
3. Thus, it is highly desirable to state the model used and state the unit – e.g. “DR model nits (mnt)”.If it is clear from the context that it is about LL values, then ubiquitous nt unit name may be used, but cd/m2 units should be avoided
whenever possible, because they are reserved by international standard for CIE Brightness.
VideoQ HDR Test Patterns ApplicationsPicture quality control and calibration tools for general public, video installers, hardware and software developers, video development labs, production, post-production and content distribution facilities in the fields of:
� HDR-PQ or HDR-HLG metadata embedded – as appropriate
� Frame size: 3840x2160 (UHD)
� Frame rate:
� Raw YUV data: any frame rate,
� MP4: 23.976 fps = default, other frame rates available on request
� Audio data: N/A
14www.videoq.comwww.videoq.com
VideoQ HDR Test Patterns Suite – Entry Level Set
Static Multi-Purpose Chart.A sophisticated test pattern for display setup, image quality visual assessment and processing chain performance check
VQMPC
N/ASuper PLUGE (Conical Grayscale) test pattern for the HDR-PQ displays performance check for very low light levels.The Light Level Range is 0.001 nt … 2 nt
VQSP
N/AAverage Picture Level dynamic sequence for testing display auto-brightness control performance. Featuring a static photo on a calibrated variable light level background. FALL range is 75 nt … 1000 nt
VQAPL
Static Display Setup test pattern for general light levels range and tone-mapping quality assessment. With a large “Butterfly” image inserted, suitable for demos and exhibitions
VQDS
Static Levels Alignment test pattern for metadata handling, displayed light levels range and tone-mapping performance check
4 ... 5 knt and 5 … 6 knt Two-tones Patches: 5 knt Circle within 4 knt Square is visible,6 knt Circle within 5 knt Square is not visible,, i.e. Light Levels are clipped at about 5 knt
0.1 ...10 knt White SPLUGE: shows clipping boundary at 5 knt
20www.videoq.comwww.videoq.com
Rendered VQLA-PQ Example: TDMB = 1,000 nt1 ... 2 knt Two-tones Patch: 2 knt Circle within 1 knt Square is not visible, i.e. clipped at 1knt level
0 ... 1 knt Grayscale: shows no clipped levels,i.e. the specified TDMB range is properly rendered
0.1 ...10 knt White SPLUGE: shows clipping boundary at 1 knt
There are two groups of bursts with frequencies proportional to luma pixels rate FY: full length horizontal bursts band and full height vertical bursts band.Maximum luminance frequency burst of exactly 0.5 FY is in the middle of each band. Two slightly oblique bands of 0.4 FY surrounds the middle burst.
Two central 0.5 FY sub-bands are especially sensitive to any errors in pixel clock, mapping or scaling. Four other sub-bands allow differentiation between horizontal and vertical distortions thru the whole picture area – from left picture edge to the right picture edge and from top to bottom.
Within the burst vertical and almost vertical lines test horizontal frequencies, whilst horizontal and almost horizontal lines test vertical frequencies.
0.4*FY Near Horizontal Burst
0.5*FY Vertical Burst
0.4*FY Near Vertical Burst0.4*FY Near Horizontal Burst 0.5*FY Horizontal Burst 0.4*FY Near Vertical Burst
32www.videoq.comwww.videoq.com
VQMPC: Tri-band Combination Burst Pattern Usage
Example of correct settings (no scaling):There are no visible beat waves on both horizontal and vertical Tri-band Patterns
Example of scaling artifacts:Scaling causes beat waves on both horizontal and vertical Tri-band Patterns
Dedicated areas indicate different sub-sampling issues:
Sampling Conversion Test Result:4:2:0 sub-sampling mode detected
Test appearance after different codecs
4:2:2 or 4:2:0 (UV H sub-sampling)?Yes
Sampling Conversion Test Result:4:2:2 sub-sampling mode detected
Pristine 4:4:4?No
4:2:0 (UV V sub-sampling)? No
Case #1 Case #2
36www.videoq.comwww.videoq.com
VQMPC: OETF Test Details and Usage ExampleFSh: Original Horizontal Sampling RateFSv: Original Vertical Sampling Rate
Checkerboard: FSh/2, FSv/2
For an average human observer the high frequency checkerboard textures are visible only from a very small viewing distance. From a normal UHD viewing distance the central squares look like solid gray.If the display OETF complies with the standard curve, then the average gray levels should match the background squares gray levels, i.e. the contrast of the central squares vs. the background is very low.
Case #1:The display EOTF at 400 nt level is close to the PQ standard (correct function curve)
Case # 2:The display EOTF at 400 nt level significantly differs from the standard curve
VQMPC: Chroma Gain Test Details and Usage ExampleFSh: Original Horizontal Sampling RateFSv: Original Vertical Sampling Rate
Checkerboard: FSh/6, FSv/6
For an average human observer the high frequency checkerboard Green-Magenta texture is visible only from very small viewing distance. From normal viewing distance the central square looks like solid gray.If the display chroma gain (saturation) is close to the nominal value (nominal gain = 1.0), then the central square average gray color should match the background gray color, i.e. the contrast of the central square vs. the background is very low.Note that the actual shades of gray perceived by the observer are affected by the display EOTF, thus the Chroma Gain Test depends on the success of the EOTF Test shown on the previous slide.
Case # 1:The display color saturation (chroma gain) is close to the nominal value
Case # 3:The display color saturation (chroma gain) is much higher than the nominal value
Case #2:The display color saturation (chroma gain) is much lower than the nominal value
38www.videoq.comwww.videoq.com
Rendered VQMPC-PQ Examples
Target Device Max Brightness: 1,000 nt
Target Device Max Brightness: 5,000 ntTarget Device Max Brightness: 500 nt
Compatible SDR Image: Max LL = 50% of HLG Range Compatible SDR Image: Max LL = Ref.White (26% of HLG Range)
40www.videoq.comwww.videoq.com
VQMPC: HDR Color Bars Test Details and Usage Example
The special HDR-PQ “shallow” Grayscalefeatures uniform light levels steps up to 100 nt
8 White PLUGE components (nested squares)
Case 2:Clipping at 10 bit value of 940 as required by HDR standard:- It is normal for all HLG displays, - It is also normal for the PQ displays capable of rendering
the full valid range, i.e. if TDMB = 10,000 nit
Case 1:The HDR display renders the full 10 bit video data range without clipping at the level 940; this is not allowed by any HDR standard (PQ or HLG)
White PLUGE 10 bit levels:860
940 = upper limit of valid (“Narrow”) 10 bit video data range
1020 = upper limit of full 10 bit video data range
Case 3:The display is not rendering full valid range; the clipping is at a video level much lower than 940:- It is not normal for any HLG display,- It is normal for the PQ displays with a TDMB < 10,000 nit
Special HDR-PQ 100 nt Color Bars
Special HDR-HLG Ref. White Color Bars
The special HDR-HLG “shallow” Grayscalefeatures uniform light levels steps up to Ref. White
VQSP-PQ – Very Low Light Levels TestRaw YUV Data Image Target Device Max Brightness: 200 nt
If the processing chain preserves the full 10 bit resolution and the display black level cut-off point setup is correct (i.e. the display under test OETF complies with the standard PQ curve), then all the gray levels, even the very low, e.g. 0.001 nt, should be visible.
This test is for viewing in a dark room (i.e. very low ambient light levels), and the observer should be given enough time for the visual system adaptation to these conditions. Rendering of low light levels is relatively easy for not so bright displays, for brighter displays this test would be much more challenging.
42www.videoq.comwww.videoq.com
VQAPL-PQ Dynamic TestFirst frame – lowest FALL value Last Frame – highest FALL value
TimeThis dynamic sequence serves to test the display’s auto-brightness control system.
The HDR display can include eye safety and power consumption protective measures, e.g. automatic reduction of the light output, when the content variable light level (FALL value) goes above some threshold. However, such protection should not be accompanied by a significant distortion of the central insert image, and the recovery time (recovery here means full return to normal mode after FALL value drop-down) should not be too long.
In 50 seconds this sophisticated dynamic source tests more than one billion (10243) colors of the 10 bit YUV color space. Thus it covers all combinations of Y, U and V values – from 0 to 1023, including all “illegal” colors.
VQCSX is equally suitable for SDR, HDR-PQ and HDR-HLG systems, checking processors, codecs and display performance. VQCSX is also suitable for both visual and instrumental tests, the results are visible on regular video monitors, waveform monitors and/or vectorscopes. VQCSX is especially efficient in combination with the VideoQ VQV Viewer-Analyzer tool.
Time
48www.videoq.comwww.videoq.com
VQCSX Usage Examples: HDR-PQ & HDR-HLG
VQCSX is equally suitable for testing HDR-PQ and HDR-HLG. Screenshots above are the examples of VQCSX usage in combination with VideoQ VQV Viewer-Analyzer.The rendered UV color palettes and vertical grayscales on the test pattern edges are very different due to the significant difference between PQ and HLG EOTFs. In the PQ 1000 nit rendering mode the PQ 1000 nit grayscale (on the right edge) shows the full light range without clipping, whilst the HLG grayscale (on the right edge) is clipped. This is reversed on the right screenshot (HLG rendering mode).
• Formed by an Engineering Awards winning team sharing between them decades of global video technology.
• VideoQ is a renown player in calibration and benchmarking of Video Processors, Transcoders and Displays, providing tools and technologies instantly revealing artifacts, problems and deficiencies, thus raising the bar in productivity and video quality experience.
• VideoQ products and services cover all aspects of video processing and quality assurance - from visual picture quality estimation and quality control to fully automated processing, utilizing advanced VideoQ algorithms and robotic video quality analyzers, including latest UHD and HDR developments.
Operations• Headquarters in Sunnyvale, CA, USA
• Software developers in Silicon Valley and worldwide