Adaptive extended piecewise histogram
equalisation for dark image enhancement
ISSN 1751-9659
Received on 4th April 2014
Revised on 19th May 2015
Accepted on 29th May 2015
doi: 10.1049/iet-ipr.2014.0580
www.ietdl.org
Zhigang Ling
1
✉
, Yan Liang
2
, Yaonan Wang
1
, He Shen
3
, Xiao Lu
1
1
College of Electrical and Information Engineering, Hunan University, Changsha 410082, People’s Republic of China
2
School of Automation, Northwestern Polytechnical University, Xi’an 720072, People’s Republic of China
3
Department of Mechanical and Aerospace Engineering, University of Central Florida, Orlando 32817, USA
✉ E-mail: zgling_hunan@126.com
Abstract: Histogram equalisation has been widely used for image enhancement because of its simple implementation and
satisfactory performance. However, traditional histogram equalisation uniformly redistributes an entire histogram or
multiple piecewise histograms with the same equalisation strategy, which may produce unnatural artefacts, over-
enhancement or under-enhancement in wide dynamic range dark image enhancement. This study proposes an
adaptive extended piecewise histogram eq ualisation algori thm (AEPHE) for dark image enhancement. First, an original
histogram is divided into a group of extended piecewise histograms. Then, an adaptive histogram equalisation, which
balances intensity preservation and contrast boosting, is fur ther developed and respectively applied to these extended
piecewise histograms. The final histogram for image e nhancement is produced by a weighted fusion of these
equalised histograms. The experimental results indicate that AEPHE is superior to multiple state-of-the-art algorithms.
1 Introduction
Image enhancement aims to improve the visual appearance of input
images and provide better transformed representations through
analysis, detection and recognition for higher level image or video
processing, and it has been actively discussed in the fields of
image processing and computer vision for the past several decades
[1–9]. However, it remains difficult to design an appropriate image
enhancement method which works universally for various
applications, for example, wide dynamic range images present
additional challenges.
As it is easy to implement and produce satisfactory results in many
cases, histogram equalisation (HE) has become one of the most
popular image enhancement methods. HE uniformly redistributes
high dynamic range greyscale to increase the average difference
between any two altered grey levels. For example, global HE
(GHE) [10] based on a cumulative distribution function
redistributes an original histogram for contrast enhancement.
However, significant peaks in original histograms usually trigger
excessive enhancement and unnatural appearance. Some adaptive
HEs have been developed to avoid this problem, including adaptive
local HE (ALHE) [11] and contrast-limited adaptive HE (CLAHE)
[12]. Both ALHE and CLAHE split the input image into numerous
small windows, and then apply HE to each window for contrast
enhancement. However, lack of global histogram information
sometimes leads to over-enhancement or undesirable checkerboard
effects. To adjust contrast enhancement, a histogram modification
framework (HMF) [13] formulates contrast enhancement as an
optimisation problem. Furthermore, a two-dimensional HE (2DHE)
[14] boosts image contrast by increasing grey-level differences
between the neighbouring pixels. On the basis of 2DHE, a layered
difference representation of 2D histograms is present to enhance
images [15]. Adaptively modified histogram equalisation (AMHE)
[16] modifies the probability density function (PDF) of the
greyscale and then applies histogram specification to the modified
PDF. However, over-enhancement and under-enhancement still can
occur and some artefacts appear in some smooth regions because
these methods entirely redistribute the original histogram.
As a result, improved HE methods have been achieved with
multiple piecewise histograms, including brightness preserving
bi-HE (BBHE) [17], equal area dualistic sub-image HE [18],
minimum mean brightness error bi-HE (MMBEBHE) [19],
recursive mean-separate HE (RMSHE) [20] and brightness
preserving dynamic fuzzy HE (BPDFHE) [21]. The BBHE splits
the original histogram into two piecewise histograms and then
independently equalises them in the fixed grey range for improved
brightness preservation. An extension of BBHE, the MMBEBHE
[19] separates the original histogram using some threshold levels
in pursuit of a minimum absolute mean brightness error (AMBE).
The RMSHE [20] recursively divides the original histogram into
several piecewise histograms via local mean values, and
respectively equalises them. The BPDFHE [21] adopts fuzzy
statistics of an input image to improve grey-level brightness and
preserve contrast, and Celik et al.[22] employed the Gaussian
mixture model to partition the original histogram. These
equalisation methods provide better brightness preservation.
However, because of adopting the same strategy for each
piecewise histogram, these equalisation methods fail to avoid
either unexpected over-enhancement with artefacts in bright
regions or under-enhancement in dark regions of images,
particularly in wide dynamic range images. To the best of our
knowledge, this issue is important but still open.
In this paper, we propose an adaptive extended piecewise HE
algorithm (AEPHE) for the enhancement of dark images with a
wide dynamic range. First, we propose two novel measures, one
for intensity preservation measure and one for contrast boosting, to
represent the statistical characteristics of the original histogram. To
balance intensity preservation and contrast adaptively, we further
develop a novel adaptive HE based on these two measures to
avoid unexpected over-enhancement or under-enhancement. In
addition, we present an extended piecewise histogram partition
strategy to improve dark region boosting. Finally, all equalised
piecewise histograms are fused by a weighting function in order to
smoothly merge the effect in the overlapping parts. The
experimental results demonstrate that AEPHE significantly
enhances dark regions without introducing excessive enhancement
or unnatural artefacts.
The paper is organised as follows. Section 2 briefly introduces the
related HMF. Section 3 first develops the intensity preservation
measure and the contrast boosting measure, and then presents the
IET Image Processing
Research Article
IET Image Process., 2015, Vol. 9, Iss. 11, pp. 1012–1019
1012
&
The Institution of Engineering and Technology 2015