IEEE Consumer Electronics Magazine - May 2018 - 40

The color space used in the conducted experiments is
NCL Y'CbCr. Its definition is provided in [13], and its gamut
mapping equations are presented in [11] and [17]. When considering HDR and WCG, the limitations of Y'CbCr representation can be listed as follows [14]:
▼ quantization distortions (bit depth limitations)
▼ chroma subsampling distortions due to a perceptually
uneven distribution of code words
▼ color volume mapping distortions due to incorrectly predicted hue and luminance
▼ error propagation from the chroma to luma channel.

ICtCp Color SpaCe
Recently, Dolby Laboratories has proposed an ICtCp color
space [13] addressing the limitations of Y'CbCr. ICtCp
extends known IPT color space [6] by exploring a higher
dynamic range (up to 10,000 cd/m2) and larger color gamuts
(e.g., BT. 2020) [20].
The process of ICtCp mimics the early stages of human
color vision. The human eye is composed of three photo receptors (cones). Each cone is particularly sensitive to long (L),
medium (M), or short (S) wavelengths. The lighting adaptation
process of the eye consists of a nonlinear signal response to
reduce dynamic range. This nonlinear output goes through a
color differencing process to extract important information and
to separate the signal into three distinct components. This
explains the following ICtCp conversion steps: 1) compute the
LMS response, 2) apply a nonlinear encoding perceptual quantizer (PQ) transfer function (TF), and 3) apply a color differencing equation (3 × 3 matrix). The color space conversion
from LMS wavelengths to ICtCp is described in [5] and illustrated in Figure 2, where intensity (I) corresponds to nonlinear
brightness of pixels, and yellow-blue Tritan (Ct) and red-green
Protan (Cp) are the color channels. The characteristics of the
ICtCp color space are exposed and demonstrated in [9], [14],
and [20] and summarized below.
▼ Achromatic channel I and isoluminance: Mutual information of chroma and luma channels will result in severe discrepancies after compression. The decorrelation between
luminance and chrominance prevents having such artifacts.
In addition, the intensity channel I corresponds closely to
the PQ luminance Y. Therefore, ICtCp is isoluminant.
▼ Hue linearity: A color space is hue linear when the hue
remains constant while the saturation or intensity are

3 × 3 Matrix

PQ EOTF-1

3 × 3 Matrix

+
+

-
- +
-

FIgure 2. The ICtCp color representation [5].
40 IEEE Consumer Electronics Magazine

may 2018

Intensity
C Tritan
C Protan

changed. ICtCp has straighter constant hues lines
than Y'CbCr.
▼ Perceptually uniform colors: The MacAdam ellipses [22]
are more circular in the ICtCp color space when compared
to Y'CbCr. Therefore, ICtCp is perceptually uniform, leading to efficient color subsampling.
▼ Quantization to limited bit depth: Experiments demonstrate that 10-bit ICtCp provides an approximately 1.5-bit
color difference improvement over 10-bit Y'CbCr and, at
the same time, presents less of a color quantization error.
Moreover, ICtCp and Y'CbCr have a similar complexity of
conversion because they require the same sequence of operations. Overall, these characteristics favor the exploitation of the
ICtCp color space to represent HDR and WCG content.

TFs
The consistency of multimedia content playback, regardless of
the exploited display or compression scheme, is very important
for service providers. TFs define the transformations needed
from the camera to the display to insure the this consistency.
This section aims at defining main types of TFs and describes
in particular the PQ and the HLG TFs.
▼ An opto-electro TF (OETF) converts linear scene light into
the video signal, typically within a camera.
▼ An electro-optical TF (EOTF) converts the video signal
into the linear light output of the display.
▼ An opto-optical TF (OOTF) has the role of applying the rendering intent. OOTFs are usually a concatenation of OETF,
artistic adjustments, and EOTF.
▼ An electro-electro TF converts linear light output of the
reference display into the linear light of the used display.
The report in [14] provides a comprehensive description
of their use and functions. Figure 3 illustrates how EOTF,
OETF, artistic adjustments, and OOTF are combined in the
end-to-end processing, from scene to display lights. Figure 3(a)
presents the usual OOTF, whereas Figure 3(b) and (c) shows
differences between the concepts of having the OOTF in the
camera or in the display, respectively.
A nonlinearity in the basic video signal was required to
improve the visible signal-to-noise ratio in analog systems,
and the same nonlinearity helps to prevent quantization artifacts in digital systems. Typical gamma curves are natural
characteristics of a cathode ray tube. However, they are inefficient when on the extended range of luminance and color.
This explains the development of new TFs that are dealing
with larger color volumes. The two most established TFs for
representing HDR and WCG contents are the PQ and HLG
TFs, defined and described in [13] and [27].

pQ
The PQ was designed based on the model shown in Figure 3(b),
where the OOTF is considered to be embedded in the camera
(or imposed in the production process). This system is then
defined by its EOTF. The PQ EOTF has been split into two
steps: a linearization equation followed by a level calibration
equation. The linearization was constructed to align with

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - May 2018