18 // SMPTE Motion Imaging Journal November/December 2014
TECHNICAL PAPER
Beyond BT.709 By Maciej Pedzisz
The recent high-efficiency video coding (HEVC) standard and current
progress in the specification of HDMI 2.0 has revived discussion about
video colorimetry. For years, BT.709 was successfully used to define
the primary chromaticities of display devices and transformation to the
Y’Cb’Cr’ color space that is used for the compression of high-definition
video. Current advances in display technologies, along with the intro-
duction of ultra-high-definition television, makes the following ques-
tion more relevant than ever before: Can we do better than BT.709?
This paper attempts to answer this question by highlighting methods
of extending the color gamut, showing the difficulties in the transfor-
mation from one color space to another, and pointing out the different
conversion methods used to represent RGB data in the Y’Cb’Cr’ color
space. It emphasizes the importance of constant luminance transform
for color perception and compares BT.709 to BT.2020 from different
viewpoints. Finally, the advantages and disadvantages of switching to
BT.2020 are presented from a broadcast engineering perspective.
INTRODUCTION
Our perception and expectations of image and video quality have
changed over the years. We have witnessed an evolution of pic-
ture resolutions from standard-definition, through high-definition
(HD) 720 and HD1080 formats, up to the recent proposals of 4K
and 8K (ultra-high-definition (UHD)-1 and UHD-2) sizes.
1
Simi-
larly, recent technological progress has enabled the adoption of
3DTV and multi-view systems. Frame rates have also increased,
with the European Broadcasting Union currently investigating the
applicability of video with more than 100 frames/sec.
1,2
Other aspects that have a great impact on picture quality are the
dynamic range (bit depth) and the chroma sampling scheme. A
common choice for consumer applications is 8 bits per color chan-
nel and 4:2:0 chroma sampling. In the professional arena, however,
4:2:2 and 4:4:4 are most commonly used with bit depths as big as
16 bits per color channel.
All of the above mentioned are “obvious” ways of improving qual-
ity, but there is still one, more obscure, element that is very often
overlooked—colorimetry.
Television started in black and white and gradually evolved into
BT.601
3
and BT.709
4
color systems. Both were designed to provide
backward compatibility by approximating perceived luminance by
a mixture of R, G, and B components and then adding two more
color difference signals.
Even though BT.709 is currently prevalent in all HD-based video
systems, its origins date back to the era where TV screens were
based on CRT technology and color primaries aligned with phos-
phorus capabilities. This is no longer the case; the majority of cur-
rent TVs are based on LCD or plasma technologies, not to mention
recent advances like OLED, multiprimary TVs, or quantum-dot.
This recent progress in TV technologies was a driving force to devel-
op new, wide-gamut color spaces, with xvYCC,
5
the most commonly
used (in the video domain), and the recently standardized BT.2020.
6
BT.2020 not only extends the gamut, but also ensures that any
BT.2020 conformant device will operate at UHD resolutions and
above with progressive scan and at least 10 bits per color compo-
nent. This might seem excessive as far as the amount of video data
is concerned; however, the recent advances in the video compres-
sion domain (HEVC
7
and VP9
8
codecs) make such technologies
feasible and desired by a broadcast community that puts consumer
experience as their primary objective.
This paper is organized as follows. After a brief introduction to
colorimetry concepts, it presents BT.2020 in more detail, exploring
important transformations, new concepts, and comparisons with
BT.709. It concludes with a discussion of the advantages and dis-
advantages of BT.2020 from a broadcaster perspective.
BRIEF INTRODUCTION TO COLORIMETRY CONCEPTS
Cone Responses
Humans perceive their surroundings via three types of cones and
one type of rod (for monochromatic vision at low light levels) lo-
cated on the retina of an eye. Assuming appropriate light levels, the
first stages in the human visual system (HVS) can be approximated
by three cone responses corresponding to long, medium, and short
(LMS) wavelengths of light. For many practical applications, these
responses are enough to fully describe the colors of a scene, without
going into the details of the higher processing layers of the HVS.
a
The LMS color space is very useful to describe chromatic adapta-
tion (estimating the appearance of a sample under a different illu-
minant), and normalized response (sensitivity) functions for three
types of cones are shown in Fig. 1 (a).
a. Readers interested in color and image appearance models and their relation to HVS
psychophysics should consult Fairchild.
9
See Martins and Sousa
10
for a more biologi-
cally oriented approach.
@ SMPTE All Rights Reserved 71.183.201.2Downloaded at SMPTE_TEST on 2014-12-09 from IP
