26-02-2013, 10:06 AM
Chameleon: A Color-Adaptive Web Browser for Mobile OLED Displays
Chameleon.pdf (Size: 1.35 MB / Downloads: 24)
Abstract
Displays based on organic light-emitting diode (OLED) technology are appearing on many mobile devices. Unlike liquid
crystal displays (LCD), OLED displays consume dramatically different power for showing different colors. In particular, OLED displays
are inefficient for showing bright colors. This has made them undesirable for mobile devices because much of the web content is of
bright colors. To tackle this problem, we present the motivational studies, design, and realization of Chameleon, a color adaptive web
browser that renders webpages with power-optimized color schemes under user-supplied constraints. Driven by the findings from our
motivational studies, Chameleon provides end users with important options, offloads tasks that are not absolutely needed in real time,
and accomplishes real-time tasks by carefully enhancing the codebase of a browser engine. According to measurements with OLED
smartphones, Chameleon is able to reduce average system power consumption for web browsing by 41 percent and is able to reduce
display power consumption by 64 percent without introducing any noticeable delay.
INTRODUCTION
DISPLAYS are known to be among the largest powerconsuming
components on a modern mobile device [1],
[2], [3]. OLED displays are appearing on an increasing
number of mobile devices, e.g., Google Nexus One, Nokia
N85, and Samsung Galaxy S. Unlike LCDs where the
backlight dominates the power consumption, an OLED
display does not require backlight because its pixels are
emissive. Each pixel consists of several OLEDs of different
colors (commonly red, green, and blue), which have very
different luminance efficiencies. As a result, the color of an
OLED pixel directly impacts its power consumption. While
OLED displays consume close to zero power when presenting
a black screen, they are much less efficient than LCDs in
presenting certain colors, in particular white. For example,
when displaying a white screen of the same luminance, the
OLED display on Nexus One consumes more than twice of
that by the LCD on iPhone 3GS [4]. Because the display
content usually has a white or bright background, OLED
displays are considered less efficient than LCDs overall. For
example, Samsung reportedly dropped OLED for its Galaxy
Tablet due to the same concern [5].
OLED Display and Color Transformation
Organic light-emitting diode or OLED [16] technology
promises much better dynamic color, contrast, and a much
thinner, lighter panel than conventional LCDs. Unlike
LCDs, an OLED display does not require external lighting
because its pixels are emissive. Each pixel of an OLED
display consists of several colorful OLEDs, usually red,
green, and blue, respectively. Because the red, green, and
blue components of a pixel have different luminance
efficacies, the color of a pixel directly impacts its power
consumption. In contrast, illumination of backlight, not
color, determines the power consumption by an LCD.
BACKGROUND AND RELATED WORK
We next provide background and discuss related work.
Color Spaces
A color sensation by human can be described with three
parameters because the human retina has three types of
cone cells that are most sensitive to light of short, middle,
and long wavelengths, respectively. A color space is a
method for describing color with three parameters. Most
used color spaces include linear RGB, standard RGB (sRGB)
and CIELAB.
The linear RGB and sRGB spaces are designed to
represent physical measures of light. In the linear RGB
color space, a color is specified by (R, G, B), the intensity
levels of the primary colors: red, green, and blue, which will
create the corresponding color sensation when combined. In
the sRGB color space, a color is also specified by (R, G, B),
but the intensity levels are transformed by a power-law
compression, or gamma correction with ¼ 2:2, to compensate
the nonlinearity introduced by conventional CRT
displays. Although CRT displays are no longer common
nowadays, the sRGB color space is still widely used in
electronic devices and computer displays.
Color Transformation
The display darkening described above can be considered
as a special case of color transformation [9]. Color
transformation considers both the lightness and chromaticity
of a color and transforms colors one by one, instead of
uniformly. Our early results presented in [9], [10] show that
color transformation can significantly reduce the display
power consumption without sacrificing user satisfaction.
The objective of color transformation is to find a color
power consumption is minimized, while meeting a perceptual
constraint. Generally speaking, there are two types of
perceptual constraints, constraint in fidelity and constraint
in usability.
Related Work
Chameleon is motivated by prior work in display power
management and leverages its algorithmic solutions. HP
Labs pioneered energy reduction for OLED-based mobile
displays [11] by darkening the display regions that is
outside the focal area. User studies [19] showed that user
acceptance of this technique is high for reading notifications
and menus but low for tasks like reading messages and
books because it is hard to determine the user’s focal area in
the latter case. Web browsing is, unfortunately, similar to
the latter. In contrast, Chameleon does not need to know the
focal area and is more effective in conserving energy thanks
to color transformation.
Web Usage by Smartphone Users
By studying web browser traces from 25 iPhone 3GS users
over three months [6] (called LiveLab traces below), we
make the following observations as related to the design
of Chameleon.
Browsing Behavior
First, mobile users still visit webpages that are not
optimized for mobile devices. While a website can
potentially provide an OLED-friendly version, e.g., with
dark background, our data show that approximately 50
percent of webpages visited by mobile users are not
optimized for mobile devices at all [12]. Therefore, one
cannot count on every website to provide an OLED-friendly
version to reduce power. This directly motivates the
necessity of client-based color transformation.
Second, a small number of websites account for most
web usage. We find that the 20 most often visited websites
of each user contribute to 80-100 percent (90 percent on
average) of the web usage by the same user, as shown in
Fig. 4 (top). Therefore, it is reasonable to maintain a color
transformation scheme for each of the 20 websites and to
use a universal transformation scheme for, or simply not
transform, the other websites. This is the key rationale
behind our design decision to maintain color consistency
per website in Chameleon (Section 4.1).
Color Contribution Collection
Chameleon generates the color contribution vector, of GUI
objects, background images, and possibly logo images
(depending on the user choice). Recall that an element of D
is determined by how many pixels have the ith color and for
how long. Therefore, whenever the display changes, contribution
collection must determine how long the previous
screen has remain unchanged, or time counting, and how
manypixels in that screen are of the ith color, or pixel counting.
To process a large number of colors (>103) and pixels (>105)
in real time, Chameleon employs a suite of techniques to
improve the efficiency of contribution collection.
Why must contribution collection be done in real time? It
would be much easier to use the browser history to record the
URLs of visited webpages with a time stamp and examine
the pages offline. The key problem with this offline method is
that it does not capture what actually appears on the display.
Because often a small portion of a webpage can be shown on
the display and the user must scroll the page, zoom in, and
zoom out during browsing, the problem will lead to
significant inaccuracy in the color contribution vector, D.