High-dynamic range image generation
from single low-dynamic range image
ISSN 1751-9659
Received on 5th October 2014
Revised on 11th July 2015
Accepted on 21st August 2015
doi: 10.1049/iet-ipr.2014.0782
www.ietdl.org
Yongqing Huo
1
✉
, Fan Yang
2
1
School of Communication and Information Engineering, University of Electronic Science and Technology of China, No. 2006, Xiyuan
Avenue, West Hi-Tech Zone, Chengdu 611731, People’s Republic of China
2
LE2I-CNRS 6306 Laboratory, University of Burgundy, Aile de I’Ingenieur, Dijon 21078, France
✉ E-mail: hyq980132@uestc.edu.cn
Abstract: Due to the growing popularity of high-dyn amic range (HDR) image an d t he high comp lexity to capture HDR
image, researchers focus on converting low-dynamic range (LDR) content to HDR, which gives rise to a number of
dynamic range expansion methods. Most of the existing methods try t heir best to tackle highlight areas du ring the
expanding, however, in some cases, they cannot achieve ap proving results . In this study, a n ovel LDR image
expansion technique is presented. The t echnique first detects the highlight areas in image; then preprocesses them
and reconstructs the information of these regions; finally, expands the LDR image to HDR. Unlike the existing schemes,
the proposed approach escapes the complicated treatment to highlight areas in the process of expansion, which makes
the expansion straightforward; at the same time, it facilitates the expansion scheme and minimises the formation of
the artefacts. The experimental results show that the proposed method performs well; the tone mapped versions of the
produced HDR i mages are popular. The results of the image quality metric also illustrate that the novel approach can
recover more image details with minimised contrast loss and reversal, compared with the existing schemes considered
in the comparison.
1 Introduction
Along with the emergency of high-dynamic range (HDR) image
sensors and display devices, researchers have focused on HDR
image processing, including capturing, coding, compression,
displaying and so on [1]. HDR display is capable of providing a
rich visual experience, video or image displayed on them looks
more natural and preferential. The development in HDR hardware
and display technique indicates that HDR display devices will
become commonplace in the near future in most fields, from
entertainment to scientific visualisation.
However, capturing HDR images or videos directly is not as easy
as capturing low-dynamic range (LDR) originals [2], it requires
specialised equipment to automate. This has led to research on
providing HDR content from LDR content, which not only obtains
HDR images by the existing consumer cameras, but also makes it
possible to re-use the large amount of already existing legacy
contents on HDR monitor. This need to expand the range of LDR
image or video to create HDR depiction which matches real-world
luminance values as faithfully as possible.
In general, at image or video capturing moment, professionals
(photographers, movie/video markers) try to avoid excessive
highlight because it distracts and obscures surface details.
However, it is unavoidable that there are highlights in image for
some applications in practice. Most of the existing image
expansion methods make a general assumption that highly
saturated pixels need to be expanded much more than the rest. As
a result, bright image areas representing features such as
highlights, or the sun in the sky, are largely boosted, thus
sometimes results in contouring artefacts for bright objects [3]. On
the other hand, in some applications, such as photography, movie,
video, object segmentation and recognition, highlight makes image
unnatural and leads to false results. So, for pleasant entertainment
or better processing, it is necessary to explore a dynamic range
expansion method that can have the image without highlights.
However, it is very complex to get rid of the highlight in the
process of expansion, moreover, it will induce artefacts. Thus, in
this paper, we propose an expansion method that not only obtain
images without highlights, but also avoids settling highlights in
expanding procedure. Its most important novelty is the fashion
used for expanding the highlight areas, which is different from the
existing algorithms. In the proposed scheme, the highlight areas
processing is separated from the dynamic range expanding step.
This novelty escapes special treatment to highlight areas in the
process of expansion and facilitates the dynamic range expansion
step. Moreover, the novelty not only minimises the formation of
the artefacts introduced by the special treatment to highlight areas
used in other existing algorithms, but also brings effective
dynamic range expansion and high quality HDR images. The
proposed approach first decomposes the image into highlight areas
and non-highlight areas; then preprocesses the highlight areas to
make the image natural and proper; finally, based on the
preprocessing result, an efficient expansion step is conducted on
the LDR image to generate the HDR image.
This paper is organised as follows: The related researches are
introduced in Section. 2. In Section. 3, the mathematical model of
the proposed approach is illustrated. Selected representative results
of a comprehensive experimental evaluation are given in Section. 4.
Conclusions and further discussions are presented in the last section.
2 Related work
Daly and Feng [4, 5] firstly addressed the problem of dynamic range
expansion by bit-depth extension techniques and de-contouring
methods. However, their schemes expand image into 10-bits per
colour channel per pixel, which is lower than that of HDR image
can carry. Afterwards, researchers have presented several dynamic
range expansion algorithms; reader can read the review paper [6].
These algorithms can be regrouped in four categories: global
processing, classification and interactive method, expand-map
approach and technique similar to HDR image capturing.
Landis [7]proposedthefirst global expansion method, which
applied an exponential function for pixel values above certain
IET Image Processing
Research Article
IET Image Process., 2016, Vol. 10, Iss. 3, pp. 198–205
198
&
The Institution of Engineering and Technology 2016