Detecting linear structures within the ASTER satellite

image by effective denoising and contrast

enhancement in the device independent color space
Sukumar M

Nelson Kennedy Babu C

Department of Computer Science & Engineering

St.Peters Institute of Higher Education & Research,
Avadi, Chennai, Tamilnadu, India
msukumar.btech@gmail.com

Department of Computer Science & Engineering

Dhanalakshmi Srinivasan College of Engineering
Coimbatore, Tamilnadu, India
cnkbabu63@gmail.com

Abstract Segmentation of linear structures (lineaments) is of

significance in the field of remote sensing but has some technical
difficulties since the size of the image is too large to process and
the color space used to acquire the image. While preprocessing
this kind of high resolution multispectral satellite images, the
device independent color space L*a*b* is preferred now-a-days
and also the preprocessing techniques is expected to preserve
brightness / contrast of the image. In this paper, ASTER image
dataset is used. Non-parametric Modified Histogram
Equalization for contrast enhancement (NMHE) and Brightness
Preserving Dynamic Fuzzy Histogram Equalization (BPDFHE)
method is selected and applied to the input image for
preprocessing. Isotropic Undecimated Wavelet transform is used
to segment the texture (lineaments) features from the
preprocessed image. For effective denoising and contrast
enhancement, the combination of above said Brightness
preserving contrast enhancement method and the undecimated
wavelet transform gives better results for the ASTER dataset
images.
KeywordsHistogram Equalization (HE), IUWT, Contrast
enhancement, ASTER (key words)

I.

INTRODUCTION

Multi-disciplinary approach to mineral exploration

comprise large scale and detailed mapping aided by
interpretative analysis of remotely sensed and aero
geophysical data, ground geophysical survey, geochemical
prospecting and subsurface exploration through pitting,
trenching and followed by drilling.
Geologic lineament mapping is considered as a very
important issue for problem solving in engineering in site
selection for mineral exploration, hydro geological research.
Major linear features may be used to find mineral deposits.
Linear features are topographic features such as ridges and
canyons that follow a straight line and are probably the surface
expression of a fault. Satellite imagery and high altitude aerial
photography are useful for this purpose. Mineral deposits tend
to be aligned along linear features. The intersection of linear
features is an excellent place to prospect. Lineaments may
represent deep fractures which could provide access to ore
fluids. Major goal of this research is to extract the linear

structures / textures and extract the stock works (a zone of

intersecting faults) from the multispectral ASTER image.
Recently more and more researchers have proposed
different approaches to detect or segment linear features from
the satellite image.
Image enhancement is the basic step in most of the image
processing applications. One of the effective ways to enhance
the image is by equalizing the histogram values of the image.
Initially, the histogram equalization methods enhance the
image fully i.e. it doesnt consider the contrast and brightness
(intensity) values present in the image. It creates undesirable
effect while post processing the image [1]. To overcome these
kinds of problems, many researchers proposed various
algorithms like Bi-Histogram Equalization (BBHE) [2]. In this
method, the image is enhanced by finding the mean value of
the histogram as a part of histogram partitioning. Minimum
Mean Brightness
Error
Bi-Histogram
Equalization
(MMBEBHE) which is same as BBHE, it splits the histogram
based on the intensity of the image and the least mean
difference is used to equalize the image [3]. Dynamic
Histogram Equalization (DHE) first smooth the image using
1D smoothing filters and splits the histogram based on the
local minimum [4]. Brightness Preserving Dynamic Histogram
Equalization (BPDHE) is an extension to HE which produces
the output image with the same mean intensity level of the
input image which refers that the mean brightness of the
image is maintained [5]. Non parametric Modified Histogram
Equalization (NMHE) can be applied in both grey level and
color images and videos too. This method preserves the
overall content of the image and also enhances the contrast
[6]. Brightness Preserving Dynamic Fuzzy Histogram
Equalization (BPDFHE) manipulates the image histogram by
redistributing the grey level values present in the valley
portion between two consecutive peaks [7]. Brightness
preserving Fuzzy Dynamic Histogram Equalization
(BPFDHE) can solve the problems like contouring effect and
the information loss in the potential information region. This
in turn improves the crispness of the interval and the number
of pixels in the interval [8]. In the study of mechanical
properties of materials, "isotropic" means having identical
values of a property in all directions. This definition is also

used in geology and mineralogy [9]. Stationary Wavelet

Transform otherwise called as undecimated wavelet
transforms. This is one of the powerful approach to denoise
the image and also in the field of pattern recognition. The
Isotropic Undecimated Wavelet Transform, IUWT, algorithm
is well suited for the astronomical data where the subjects of
matter are more or less isotropic in most cases [10] and [11].
Isotropic Undecimated Wavelet Transform (IUWT) is a simple
method for denoising and segmentation [12].
The rest of the paper is organized as follows. The Second
section explains about the materials and methods. Third
section explains the experiments & results and the final
section states the conclusion and future work.
II.

MATERIALS

& METHODS

A. ASTER Image Dataset

In order to segment the linear features from the High
resolution Multispectral image (e.g.) ASTER satellite image is
used. ASTER is an Advanced Spaceborne Thermal Emission
and Reflection Radiometer; a multispectral imager which
covers a wide spectral region of the electromagnetic spectrum
from the Visible Near Infra Red (VNIR) to the Thermal Infra
Red (TIR). ASTER Image dataset is the best tool for the
minerals exploration application because the image acquisition
cost is low. ASTER image covers large area. The availability
of ASTER data is also easy. It can accurately map lithologic
and mineralogical units on the surface. VNIR data at 15m
resolution is currently the best resolution multispectral satellite
data available commercially.
B. Non parametric Modified Histogram Equalization
Non-parametric
Modified
Histogram
Equalization
(NMHE) [6] holds an independent parameter setting for
dynamic range of images. In addition, it removes spikes and
also it doesnt need any additional parameters to be given
manually to every image. This method is able to process only
the gray scale images. The procedure for NMHE is given as
follows:
1. Remove spikes from the histogram
a) Compute the modified histogram by comparing the
dissimilar pixels with its neighbors
b) Normalize the modified histogram
c) Calculate the measure of un-equalization (Mu)
2. Clip the histogram and find the measure of un-equalization
(Mu)
3. Obtain modified probability density function based on the
Mu factor
4. Obtain modified histogram equalized image
C. Brighness Preserving Dynamic Fuzzy Histogram
Equalization
Brightness preserving dynamic fuzzy histogram
equalization [BPDFHE] technique equalize the image
histogram by distributing the gray values present in the valley
portions of the histogram. It clearly shows that no remapping
of the histogram peaks takes place. This method is used in

both grayscale and color images. The BPDFHE technique

consists of following operational stages:
1. Change the input image to the L*a*b color space
2. Computation of fuzzy histogram
a. Produce the smooth histogram

h(i) is the frequency rate of gray levels

i(x,y)i is the triangular fuzzy membership function
i(x,y) is the grey values as a fuzzy number
[a,b] is the triangular membership function
3. Partition the histogram based on the local maxima value.

where h(i) is the first order derivative of fuzzy histogram

h(i) corresponds to the ith intensity level.
To reduce the approximation errors, second order
derivative is calculated from the fuzzy histogram

4. Equalize every partitioned histograms dynamically

Partitioning the histograms based on {[Imin,m0],[m0+1,m1],
[mn+1,imax] parameters used to dynamically
equalize the histogram by
spani=highi-lowi
Highest and lowest intensity values contained in the
partitioned histogram is
factor=spani x log10Mi
Mi is the total number of pixels present in the partitioned
histogram

[start1, stop1]=[0, range1]

[startn+1, stopn+1]=[

, l-1]

Global Histogram Equalization method is used to equalize

the partitioned histograms. The remapped values are
obtained for the ith partitioned histogram is as

where y(j) is the new intensity level, h(k) is the value of the
histogram,
the partitioned fuzzy histogram.

is the total population count in

5.

Normalizing the brightness of the image

segmentation from the preprocessed enhancement images is

D. Isotropic Undecimated Wavelet Transform

Isotropic undecimated wavelet transform is suitable for
astronomical imaging. It decomposes the image into different
scales. IUWT introduces a multi resolution algorithm for
detecting bright spots. The feature detection is the process of
extracting and combining multilevel elements of response,
with each element coming from successive resolution level. To
keep the significant response of the filter to the desired
feature, the denoising technique uses hard thresholding value.
Finally, the newly selected coefficient allows us to combine
multi scale information to detect the spots. But, its
performance is slightly poor in case of low quality images, at
that time, soft thresholding is used; instead of hard
thresholding [13].
1. Initialize i to 0, starting with the original image M0(x,y)
2. Increment the value of I, the data Mi(x,y) is convolved with
rows and then by columns along with the kernel h. and the
result is Mi+1(x,y). The kernel h is [

quite satisfactory in lineament detection.

] and is

modified in terms of scale i by inserting (2 i-1-1) zeros

between two taps.
3. Calculate DWT
4. Return to step 2 till scale i equals to the number k which is
the deepest resolution level.
III.

EXPERIMENTS

& RESULTS

In this work, ASTER satellite image is used as input to the

system and apply the above said algorithms and measure the
Absolute Mean Brightness Error (AMBE) and PSNR values.
AMBE is the absolute difference between the mean of input
and output images.

Fig.3.2. Change in Color space (RGB to CIELAB)

Fig.3.3. Enhanced Image using NMHE

Fig.3.1. Loading ASTER Satellite Image

Lower the AMBE depicts the better brightness
preservation in the image and Higher the PSNR gives the good
contrast enhancement. From the experiments and the values of
AMBE and PSNR, BPDFHE technique is better when
compared with the NMHE method. And the resultant

Fig.3.4. Image enhancement using BPDFHE

Fig.3.5. Edge detection using IUWT
Fig.3.6. Mapping of lineaments in the input image

IV.
C
I

CON
LUS
ON
&

[8]

[9]
[10]

[11]
[12]

[13]

FUTURE WORK
In this paper, linear structures are detected within the
ASTER satellite image by using the effective denoising and
contrast enhancement methods. Isotropic Undecimated
Wavelet Transform is mainly used in the field of medical
image processing to segment the vessels. IUWT along with
the BPDFHE technique enhances the bright spots present in
the satellite image. In Minerals targeting system, geologic
lineaments need to be extracted. But the complexity in
detecting those lineaments is: One side of the lineament looks
brighter and the other side is not. In this work, the image is
effectively denoised and contrast is enhanced and some of the
linear structures are detected. In the future work, sensitive
shape optimization algorithms planned to be adopted for better
lineament detection.

REFERENCES
[1]

[2]

[3]

[4]

[5]

[6]

Chen, S.-D., Rahman Ramli,A.: Preserving brightness in histogram

equalization based contrast enhancement techniques, Digital Signal
Process., 2004, 14, pp.413-428
Yeong-Taeg Kim, Contrast enhancement using brightness preserving
bi-histogram equalization, IEEE Trans. Consumer Electronics, vol.43,
no.1, pp. 1-8, Feb. 1997
Soong-Der Chen and Abd. Rahman Ramli, Minimum mean brightness
error bi-histogram equalization in contrast enhancement, IEEE Trans.
Consumer Electron., vol.49, no.4, pp.1310-1319, Nov. 2003.
Abdullah-al-wadud, M.,Kabir, M.H.,Dewan, M.A.A., Oksam, Chae:, A
dynamic histogram equalization for image contrast enhancement, IEEE
Trans. Consumer Electron., 2007, 53, pp. 593 - 600
Haidi Ibrahim,N.S.Pik Kong, Brightness preserving dynamic histogram
equalization for image enhancement, IEEE Trans. Consumer Electron,
Vol.53, No.4, Nov 2007
S.Poddar et al., Non-parametric modified histogram equalization for
contrast enhancement, The Institution of Engineering and Technology,
Vol.7, Iss.7, pp. 641-652, 2013.
[7] MPS Kuber et
al., Improving
brightness using
dynamic fuzzy
histogram
equalization,
Intl. Journal of
signal

processing, image processing and pattern recognition, Vol.8, No.2,

pp.303-312, 2015
Abd. Sarrafzadeh et al., Brightness preserving fuzzy dynamic
histogram equalization, Proceedings of the Intl. multi conference of
engineers and computer scientists, vol.1, Mar. 2013.
https://en.wikipedia.org/wiki/Isotropy
Koteswararao and Dr.Prasad, Decimated and Undecimated Wavelet
Transform based estimation of Images, Intl. Journal of Innovative
Research & Sci. Engg & Technology, Vol.3, Issue:10, pp. 16981-16988,
2014
J.L.Starck et al., The Undecimated Wavelet Decomposition and its
reconstruction, DRAFT, 2006
[12] Kui Jiang et al., Isotropic undecimated wavelet transform fuzzy
algorithm for retinal blood vessel segmentation, Journal of Medical
Imaging and Health Informatics, vol.5, No.7, Nov. 2015.
De-Shuang Huang et al., Intelligent Computing Theories and
Methodologies, Springer, Aug. 2015

M.Sukumar received his B.Tech. degree in

Information Technology from Anna University,
Chennai, India in the year 2007 and M.Tech degree
in Computer and Information Technology from the
Center for Information Technology and Engineering
of
Manonmaniam
Sundaranar
University,
Tirunelveli, India in the year 2011. Currently, he is
an Assistant Professor in the Department of
Information Technology, Sri Vidya College of
Engineering & Technology, Virudhunagar, India and
also pursuing Ph.D in St.Peters University, Chennai,
India. His research interests include Image Processing, Remote Sensing and
He is the Student member of UACEE.
C.Nelson Kennedy Babu received his M.Sc
Degree from Madurai Kamaraj University,
Madurai, India and M.Tech degree in Computer
and Information Technology from Center for
Information Technology and Engineering of
Manonmaniam Sundaranar University, Tirunelveli,
India in the year 2004 and Ph.D degree in
Computer Science from Madurai Kamaraj
University, Madurai, India in the year 2009. He
has more than two decades of service in the field
of computer science and engineering. Currently, he
is the Professor in the department of Computer Science and Engineering of
Dhanalakshmi Srinivasan College of Engineering, Coimbatore, India. His
research interests include Signal and Image Processing, Remote Sensing,
Visual Perception, Mathematical Morphology and Pattern Recognition. He is
the senior member of IEEE.