IEEE Consumer Electronics Magazine - October 2015 - 83

interpret them. This is usually achieved by minimizing the difference between the original content color and a color belonging to the display color gamut. The quality of the result will
greatly depend on the color space where the color difference is
computed. CIE has defined several color spaces for manipulating color information, among which the CIE 1976 L*,a*,b*
color space (commonly referred to as CIELAB) and the CIE
1976 L*,u*,v* color space (commonly referred to as
CIELUV) are of particular interest because of their perceptual
uniformity. Perceptual uniformity is an interesting feature
because it allows for direct measurement of the perceptual difference between two different colors. Note that performing
color mapping on the fly can be quite computationally expensive. Thus, 3-D look-up tables (3-D LUTs) are usually created
to map one color gamut to another. These LUTs are derived in
such a way that each existing tristimulus value of a color
gamut has only one correspondence in a smaller color gamut.
In cases where the color gamut of a video content is smaller
than the display gamut (for instance, BT.709 content for
BT.2020 display), no gamut expansion is performed. This is
because the HVS is very sensitive to color deviation on common objects/surfaces (skin, nature, etc.), and thus, blind gamut
expansion may result in an unnatural appearance of the content.
However, when two versions of video content are available
before transmission, as illustrated in Figure 17(a), metadata can
be derived to guide the gamut expansion at the display side.
If the standard representation of the content is larger than
that of the display but its color values stay within the gamut
of the display, color mapping is not required (for instance
video content that is represented in the BT.2020 color gamut
whose color values are within BT.709 color gamut). However,

(low-frequency component) [43]. Finally, HVS operators
simulate different behavior of the HVS [44]. The selection of
a TMO usually depends on the targeted application since
TMOs have different strengths and drawbacks. Because of
the lack of high-quality HDR video content, the temporal
coherency aspect of tone mapping has been dismissed for a
long time. Several studies point out that naively applying a
TMO to each frame of an HDR video sequence leads to different types of temporal artifacts [45], [46]. Recent works
[47], [48] have addressed some of the problems inherent to
video tone mapping, focusing either on their computational
cost or their suitability for real-time applications.
When the display's dynamic range is larger than that of the
content, an inverse TMO (iTMO) is applied to expand the
dynamic range of the content. iTMOs estimate luminance
information that has not been recorded during the acquisition.
To recover lost information in saturated areas, iTMOs rely on
luminance expansion maps to represent details in areas of high
luminance [49]. Both tone mapping and inverse tone mapping
can greatly benefit from metadata to increase the quality of the
reconstructed content. This is the case in the distribution pipeline illustrated in Figure 17. Note that the loss of information
when performing a TMO-iTMO process is only due to the
quantization step. Without this step and with enough metadata,
any TMO is fully invertible without any loss.
Color gamut mapping operators are used to project color
values from a larger color gamut to a smaller one. Figure 20
illustrates a content represented in the BT.2020 color gamut
that should be displayed on a monitor with the BT.709 color
gamut. Hence, all the values outside the yellow triangle should
be projected inside the triangle for the display to be able to

(a)

(b)

(d)

(c)

(e)

FIGURE 19. (a), (b), and (c) The different exposures of the same scene, (d) a false color representation of the generated HDR image, and
(e) a tone-mapped HDR image. [Photos courtesy of Zicong (Jack) Mai.]

october 2015

IEEE Consumer Electronics Magazine

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - October 2015