Full text of DOI:10.1080/014311600750019895

This article was downloaded by: [Columbia University]
On: 09 December 2014, At: 09:47
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK
International Journal of
Remote Sensing
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/tres20
Directional morphological
image transforms for
lineament extraction from
remotely sensed images
N. K. Tripathi a , K. V. G. K. Gokhale a & M. U.
Siddiqui b
a Department of Civil Engineering , I.I.T. Kanpur ,
Kanpur, 208 016, India
b Department of Electrical Engineering , I.I.T.
Kanpur , Kanpur, 208 016, India
Published online: 25 Nov 2010.
To cite this article: N. K. Tripathi , K. V. G. K. Gokhale & M. U. Siddiqui (2000)
Directional morphological image transforms for lineament extraction from remotely
sensed images, International Journal of Remote Sensing, 21:17, 3281-3292, DOI:
10.1080/014311600750019895
To link to this article: http://dx.doi.org/10.1080/014311600750019895
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all
the information (the “Content”) contained in the publications on our
platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy,
completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of
the authors, and are not the views of or endorsed by Taylor & Francis.
The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and

Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in
relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-and-
conditions

int. j. remote sensing, 2000, vol. 21, no. 17, 3281–3292
Directional morphological image transforms for lineament extraction
from remotely sensed images
N. K. TRIPATHI, K. V. G. K. GOKHALE
Department of Civil Engineering, I.I.T. Kanpur, Kanpur-208 016, India;
e-mail: nitin@iitk.ernet.in
and M. U. SIDDIQUI
Department of Electrical Engineering, I.I.T. Kanpur, Kanpur-208 016, India
Abstract. Morphological image transforms Ž nd their basis in the notions of
mathematical morphology. An attempt is made to develop morphological trans-
forms for enhancement of directional edges and lineaments from satellite imagery.
Geostructural features are generally oriented in preferred directions. This informa-
tion has been utilized in designing structuring elements for morphological trans-
forms. Several shapes and sizes of structuring elements have been designed and
applied to delineate lineaments in diŒerent litho-environments of the same study
area. Top hat transform has been applied using the directional structuring elements
ß
ß
ß
ß
ß
ß
for edge image at 0 , 30 , 60 , 90 , 120 and 150 . Geological lineaments such as
faults could be easily identiŽ ed on an edge image obtained using top hat trans-
formation followed by an image superimposition technique. The lineament map
was developed using directionally enhanced edge images and the results have
been veriŽ ed.
1. Introduction
The general trend in digital image processing in remote sensing has been to adopt
conventional edge operators or enhancement techniques. Qari (1991) used classical
image processing such as principal component analysis, decorrelation stretching and
highpass Ž ltering for lineament enhancement adopting Landsat TM data. The study
revealed the directional nature of geologic lineaments and emphasized the need for
directional Ž ltering. Mah et al. (1995) also adopted an algorithm for enhancing the
lineaments in particular directions to obtain maximum mapping e ciency. However,
minor lineaments oriented in directions other than that of the main fault could not
be enhanced. The complex mathematical and statistical nature of conventional Ž lters
distort the basic shape and size of inherent features in remotely sensed images.
Although directional edge operators are available, their adaptability in the case of
natural surface features has not been established. Image processing based on mathe-
matical morphology is explored in this paper for enhancement of directional lineaments.
In this case an image is understood to be a set of natural or real numbers which is
constituted by a number of subsets comprising numerous structuring elements. These
Paper presented at ICORG ’97 International Conference on Remote Sensing and
/
GIS GPS, Hyderabad, India, 18–21 June 1997.
International Journal of Remote Sensing
ISSN 0143-1161 print/ISSN 1366-5901 online © 2000 Taylor & Francis Ltd
http://www.tandf.co.uk/journals

3282
N. K. T ripathi et al.
subsets have a relationship which can be described using the notions of set theory
(Matheron 1975, Serra 1982, 1986, Giardina and Dougherty 1988). Morphological
image processing has been utilized in many areas and has gained prominence in
recent years. Conrad (1972) employed principles of mathematical morphology for
studying the permeability characteristics of petroleum rocks using the simulation of
a set of geological faults (Serra 1982). Photographs of dust particles have been used
to understand their size distribution and shape characteristics (Klein and Serra 1973).
In the recent past, scientists in the area of remote sensing have initiated studies
involving morphological image processing for feature identiŽ cation in satellite
imagery (Ansoult et al. 1990, Vincent and Soille 1991). Morphological transforms
have been generally applied for the binary image case (Ghosh 1988, Haralick et al.
1989, Maragos and ZiΠ1990). Remotely sensed digital images are multilevel, i.e. in
greyscale. Sternberg (1986) presented a generalization of the morphological opera-
tions of erosion, dilation, opening and closing for the greyscale image case. The same
concept is being adopted and explored here for remotely sensed greyscale images
obtained from the Indian Remote Sensing Satellite (IRS-1B). In the present work
the possibility of employing the concept of morphological image processing for
remotely sensed satellite data for enhancing structural features is primarily explored.
2. Study area and data used
The north-eastern part of the Singrauli coal basin, lying in central India, was
chosen as the study area. This area is marked with several faults and a dense network
of geological lineaments. The sequence within the coalŽ eld in this region is essentially
of Lower Gondwana Formations. A major fault traversing in an east–west direction
is located in the northern part of the basin. This limits the northern boundaries of
coal seams and also marks the boundary of metamorphic and Barakar sandstone
zones. Govind Ballabh reservoir is situated in the south-east of the study area. There
are nine coal-mining blocks: Jhingurdah, Kakri, Bina, Marrak, Dudhichua, Jayant,
Nigahi, Khadiya and Amolori. Open-cast mining is adopted in this area because
coal is available at shallow depth. The general location and geological settings are
shown in Ž gure 1.
It has been suggested by earlier workers that the red band (0.6–0.7 mm) oŒers
better diŒerentiation of soil and geological boundaries (Jensen 1986). Data from
the Indian Remote Sensing Satellite (IRS -1B) of path 23 and row 51of Band 3
(0.62–0.68 mm) has been used in the present work. A Band 3 image of the study area
is shown in Ž gure 2. The Bijul river is seen in the upper left corner. This river is
structurally controlled by the Vindhyan ridges. The major boundary fault can be
seen distinctly traversing across the centre of the image from east to west. The coal-
mining blocks are arranged in garland fashion and situated to the south of this fault.
As detailed in Ž gure 1, the tone and texture variation for metamorphic and Barakar
zones is observed in Ž gure 2. This area is full of major and minor lineaments in
addition to other geomorphic and geostructural details.
The toposheet at 1:250 000 from the Survey of India and the geological map at
the same scale from the Geological Survey of India were used. In addition, several
maps generated by Coal India Limited after Ž eld mapping were also utilized for the
present study.
3. Morphological transformations
Satellite imagery is used for obtaining information on inherent objects and eŒorts
have been directed towards maximizing information retrieval. Human vision fails to

Directional morphological image transforms
3283

3284
N. K. T ripathi et al.
accomplish this task e ciently due to its spectral and spatial constraints (Gonzalez
and Wintz 1977). Digital image processing can alleviate this limitation to a larger
extent by enhancing the quality of the image. Morphological transformations attempt
to establish a spatial relationship between image subsets. These are referred to as
image objects. In morphological image processing the basic set selected to identify
any larger image subset of the principal image Z is very important. This basic set is
termed as a structuring element.
In an image space Z for each of its points i, a separate structuring element h(i)
can be associated. An image can be modiŽ ed to various forms by changing the
structuring element or the transformation types.
The principles of mathematical morphology have also been extended to the
greyscale image case. Greyscale images are described as a grey level function f (x, y)
on the points of Euclidean 2-space. In Euclidean 3-space the grey level function is
[
]
considered as a set of points x, y, f (x, y) imagined as a thin, undulating and not
necessarily a continuous sheet. A greyscale image is described in terms of mathe-
matical morphology by an umbra U(f ) in Euclidean 3-space, where a point p with
its coordinates (x, y, z) belongs to the umbra if and only if z f (x, y). In image
<
processing the umbrae are very signiŽ cant as they remain umbrae even after morpho-
logical operations such as union, intersection, dilation or erosion are performed.
Sternberg (1986) extended the fundamental morphological operations of dilation,
erosion, opening and closing from the binary image case to the greyscale images as
explained in following subsections.
3.1. Dilation
The dilation of a set X by structuring element Y can be expressed as:
XC Y
X
(1)
5 p
Y
×
y Y
The dilation of U a by B is the union of translations of U a by points of B. At a
[ ]
[ ]
given location (x, y) the grey levels d(x, y) of the dilation are determined by the
[ ] [ ]
maxima of the translated umbra U a . Although U a is translated by all the points
Figure 2. IRS 1B LISS II Band 3 image of the Singrauli coalŽ eld.

Directional morphological image transforms
3285
of B, only those points lying on the surface of B can create maxima. The dilated
greyscale image is computed then, as the maximum of the sum of grey levels of
b(x, y), the function describing the surface of B, with each of the points of a(x, y).
[
]
Õ
Õ
d(x, y) max a(x i, y j) b(i, j)
(2)
5
1
i, j
3.2. Erosion
The erosion of a set X by structuring element Y can be expressed as
XHY
X
(3)
5 o
×
y Y
y
[ ]
The erosion of U a by B is likewise determined as a diŒerence of the grey levels of
b(x, y) and points of a(x, y), the grey levels of the eroded image e(x, y) determined
by the minimum over all the points of b(x, y).
[
]
Õ
Õ
Õ
Õ
Õ
e(x, y) min a(x i, y j) b( i, j)
(4)
5
i, j
The expressions for greyscale dilation and erosion bear a marked similarity to the
convolution integral encountered frequently in digital image processing, with sums
and diŒerences replacing multiplication and minimum and maximum replacing sum-
mation. Image processing through iterative morphological transformation is a pro-
cess of selective information removal where irrelevant image content is irrecoverably
destroyed, enhancing the contrast of the essential image features.
3.3. Opening
The opening of X by Y is deŽ ned in terms of an erosion followed by a dilation.
Mathematically it can be represented as:
X
[XHY ]C Y
(5)
5
Y
The opening of X by Y is the union of translation of Y completely contained in X.
3.4. Closing
The closing of X by Y is a dilation followed by an erosion. Mathematically it
can be represented as:
X
[XC Y ]HY
(6)
Y 5
The closing of X by Y is the complement of the translations of Y completely
contained in X .
c
Greyscale images opened and closed by three-dimensional structuring elements
are transformed in predictable and useful ways. Rotationally symmetric structuring
elements are often employed for opening transformations because the results do
not depend on the orientation of the image. These transforms utilize sequences of
operations involving dilation, erosion and diŒerence for detecting edges and are
non-reversible in nature.
4. Morphological edge detection
Most of the transforms available for edge detection are based on the basic
operations of dilation and erosion. Techniques adopted for edge detection in satellite
imagery are discussed as follows.

3286
N. K. T ripathi et al.
4.1. Edge detection based on dilation
The digital satellite image f (x, y) is dilated using structuring element h(i, j). The
original image (Ž gure 2) is subtracted from this dilated image. The resultant trans-
formed image g(x, y) reveals a large number of edges. It is noticed by visual inspection
that outer boundaries of the objects are emphasized. Mathematically the process is
as follows:
[
DILATE f (x, y)] f (x, y)C h(i, j)
(7)
(8)
5
[
Õ
g(x, y) DILATE f (x, y)] f (x, y)
5
Therefore, this process seems to be relevant for extracting the geological boundaries
of various formations and other geomorphic units marked by outer peripheral
boundaries.
4.2. Edge detection based on erosion
In another approach image f (x, y) is eroded using structuring element h(i, j). The
eroded image is subtracted from the original image and the transformed image g
revealed a large number of edges.
¾
[
ERODE f (x, y)] f (x, y)Hh(i, j)
(9)
(10)
5
[
Õ
g (x, y) f (x, y) ERODE f (x, y)]
¾
5
The edge output using this method brings out mainly inner boundaries and numerous
noisy edge contours. The abundance of inner edges of surface features hinders visual
interpretation of lineaments and geostructural features like faults. Several structuring
elements of size 3 3 and 5 5 having solid, ring, plus and diagonal shapes at 45ß
Ö
Ö
and 135ß (Ž gure 3) have been applied and it is found that in the case of the present
image, the 3 3 plus shape structuring element oŒers edge image with less noise and
Ö
more interpretability.
Figure 3. Structuring elements used in edge detection.

Directional morphological image transforms
4.3. Edge detection using top hat transformation
3287
Edge detection is achieved using transformations involving operations of closing
and diVerence. This technique is known as top hat transformation. Top hat morpho-
logical transformation extracts topographical peaks or valleys from a digital image.
An explanation of the process is illustrated in Ž gure 4. Top hat transformations are
developed for the binary image case. In the present study an attempt is made to
apply it to the case of the greyscale remotely sensed image.
The top hat transformation based on the close operation is expressed
mathematically as follows:
Õ
Top hat ( f )
f
f
(11)
5
B
where f is an image obtained after performing the close operation on image f.
B
The top hat transformation based on the open operation is expressed mathematic-
ally as follows:
Õ
Top hat ( f )
f
f
(12)
5
B
Where f is an image obtained after performing the open operation on image f.
For extraction of structural features such as faults, folds, fractures and lineaments,
B
a transformation involving closing and diŒerence with the original image is expected
to yield better results than the one involving opening and diŒerence operations. In
geologic imagery, near the structural features such as faults, fractures are marked by
a moist zone containing soil, a granular zone and vegetation, characterized by a
sudden drop in grey level values due to much lower re ectance than rock features.
Therefore, the main object while targeting edges is to Ž lter out topographic valleys.
As illustrated in Ž gure 4, the topographic valleys are Ž ltered using the top hat
transformation based on the close operation.
The top hat transformation should be applied in all possible directions as the
orientation of linear features in a new geological image is unknown. The structuring
element h(i, j) in uences the Ž nal output due to its size and shape. Several such
elements can be designed to obtain an output closer to the desired one. This
transformation is performed using a sequence of closing by three structuring elements
Figure 4. Top hat transformation using closing and diŒerence operations.

3288
N. K. T ripathi et al.
with directions 0ß, 60ß, 120ß and also the complementary structuring elements in the
directions 30ß, 90ß, 150ß followed by the diŒerencing operation with the original
image. The structuring elements used have hexagonal shape as shown in Ž gure 5. In
these, structuring elements, the main active cellular elements are positioned at the
desired orientation.
These edge transforms are applied on IRS-1B LISS II image as shown in Ž gure 2.
The result of transformation using a structuring element at 0ß direction on the input
image appeared very noisy. The other directions in which the transformation was
applied were 60ß, 120ß, 30ß, 90ß and 150ß which displayed distinct edges. The edge
image obtained using top hat 60ß and 120ß are shown in Ž gures 6 and 7.
4.4. Edge detection using top hat transformation and image superimposition
It was found that top hat transformation could bring out edges very prominently
in various directions. These edge outputs may be used for interpreting lineaments.
Lineaments such as faults, fractures, dissected zones and joints are delineated on the
satellite imagery using certain visual keys like isolated linear vegetation trend, drain-
age pattern and moisture in comparatively dry regions (Lillesand and Kiefer 1979).
These cover types resulting from the presence of moisture are associated with
low grey levels. In the present work the original image comprising all natural
Figure 5. Hexagonal structuring elements for directional edge detection in top hat
transformation.
ß
Figure 6. Edge image using top hat transformation at 60 .

Directional morphological image transforms
3289
ß
Figure 7. Edge image using top hat transformation at 120 .
features was superimposed on the edge output of the top hat transformation. The
methodology adopted here was as follows:
Step 1. CLOSE operation on the image f ;
Step 2. Subtract the original image from the output image of step 1;
Step 3. Compare each pixel of top hat ( f ) and original image ( f ) and replace the
pixel of the top hat image if there exists a minima in the original image.
[
EDGE ( f ) MIN {Top hat ( f )}, f ]
(13)
5
This method was attempted involving several structuring elements and the best
results were obtained in the case of 45ß, 135ß and plus shape structuring elements of
size three (Ž gure 3). In the case of other structuring elements of larger size, fewer
edges were obtained and other conŽ gurations of structuring elements rendered noisy
edge image, which is unsuitable for lineament mapping. The edge image with best
visual appearance emphasizing the edges and lineaments was rendered by the 3
plus shape structuring element (Ž gure 8).
3
Ö
5. Results and discussions
Four transformations for edge detection were employed in the present study. The
Ž rst approach involving image diVerencing and dilation helped in extracting exterior
boundaries, whereas the second approach of diŒerence of the original image and
eroded image yielded interior edges. The Ž rst transformation involving dilation is
more useful in the case of geologic images for delineation of formation boundaries
and detecting main boundary faults. Several structuring elements were used but the
acceptable results were found with the plus shape element 3 3. The edge image
Ö
produced by the second approach appeared di cult to interpret as a large number
of inner edges were produced.
The third approach for edge detection was top hat transformation. Hexagonal
structuring elements of size six were employed to enhance the edges in various
directions like 0ß, 60ß, 120ß, 30ß, 90ß and 150ß. The edge output showing grey and
white shades oŒers distinct impressions of edges (Ž gures 6 and 7). The discontinuities

3290
N. K. T ripathi et al.
Figure 8. Edge image using top hat transformation and image superimposition using the
plus shape 3Ö3 structuring element.
and lineaments are better perceived in these edge images. A combined analysis of
edge images in various directions as mentioned above oŒer better visual delineation
of lineaments.
The fourth approach of edge detection proposed included image superimposition
as a modiŽ cation over top hat transformation. Three structuring elements (45ß, 135ß
and plus shape 3 3) were utilized which yielded encouraging response by emphasiz-
Ö
ing the moist zones near the geostructural lineaments. If the size of structuring
element is increased, a blurring eŒect is caused in the edge output. Edge output from
the plus shape 3 3 transform is shown in Ž gure 8. The output obtained using this
Ö
process oŒered the most encouraging edge image on which faults and lineaments
could be easily identiŽ ed. This method provides an e cient tool to distinguish
between physical objects (roads, railway lines, built-up and other man-made linear
features) and geo-structural features. A geostructural map (Ž gure 9) has been pre-
pared based on edge images from top hat 30ß, 60ß, 90ß, 120ß, and 150ß and image
superimposition techniques at 45ß, 135ß, and plus shape structuring elements. In this
map, major boundary faults along with other orthogonal faults are illustrated. In
Jingurdah, Kakri, Bina, Marrak and Khadia, a large number of faults are located.
Mines in this area are situated in a garland shape and the structural features tend
to limit their extent. Maps from Coal India Limited and the Geological Society of
India contain these faults. Formation boundaries of metamorphics, Barakar and
Talchir have also been mapped from distinct tonal and textural variations. The
periphery of Sardar Ballabh reservoir is also mapped in the bottom right corner of
the image.
6. Conclusions
Morphological transformation involving top hat transformation employing a
hexagonal structuring element for multi-direction 0ß, 60ß, 90ß, 120ß and 150ß can be
used for enhancement of directional lineaments. In this process, man-made linear
features like roads and canals could be avoided by using top hat transformation and
image superimposition techniques. Geostructural details (Ž gure 9) mapped using this

Directional morphological image transforms
3291
Figure 9. Geostructural map produced based on morphological transformation.
approach have been veriŽ ed from the published maps of the Geological Society of
India (GSI) and Coal India Limited (CIL). Most of the major faults mapped in
Ž gure 9 are conŽ rmed by GSI and CIL maps. In addition, several new faults were
also interpreted. The results of visual delineation of lineaments presented in Ž gure 9
are exploratory in nature. Although, major lineaments have been conŽ rmed, the
reliability of minor faults and lineaments is yet to be established by appropriate
detailed Ž eld exploration.
The present approach of mathematical morphology has yielded satisfactory
results. It seems to be better suited for digital enhancement of lineament based on
its unique characteristics of preserving the basic morphology and homotopy of the
image by a proper choice of structuring elements and not involving complex
mathematical and statistical operations as adopted in conventional image processing
techniques. The success obtained in enhancing the linear features in the present study
indicates the potential for future development of structuring elements and trans-
formations for extraction and delineation of geomorphic features in addition to
geostructural features from satellite imagery.
Acknowledgments
The authors extend their sincere gratitude to engineers and staΠof Coal India
Limited for their help in conducting the Ž eld veriŽ cation and collecting the ground
data. The authors are grateful to the Department of Civil Engineering, Indian
Institute of Technology Kanpur, India and Ministry of Human Resource
Development, India for funding this work at diŒerent stages and providing the
necessary infrastructure. We are obliged to the Coal Mining Planning and Design
Institute (CMPDI), India, for help provided during Ž eld veriŽ cation of lineaments
in the Singrauli coal basin.

3292
Directional morphological image transforms
References
Ansoult, M. M., Soille, P. J., and Loodts, J. A., 1990, Mathematical morphology: A tool
for automated GIS data acquisition from scanned thematic maps. Photogrammetric
Engineering and Remote Sensing, 56, 1263–1271.
Ghosh, P. K., 1988, A mathematical model for shape description using Minkowski operators.
Computer V ision Graphics and Image Processing, 44, 239–269.
Giardina, C. R., and Dougherty, E. R., 1988, Morphological Methods in Image and Signal
Processing (New Jersey: Prentice Hall).
Gonzalez, R. C., and Wintz, P., 1977, Digital Image Processing (Massachusetts: Addison
Wesley).
Haralick, R. M., Xinhua, Z., Lin, C., and Lee, J. S. J., 1989, The digital morphological
sampling theorem. IEEE T ransactions on Acoustics, Speech and Signal Processing, 37,
2067–2089.
Jensen, J. R., 1986, Introductory Digital Image Processing (New Jersey: Prentice Hall).
Klein, J. C., and Serra, J., 1973, The texture analyser. Journal of Micro-optics, 95, 349–356.
Lillesand, T. M., and Kiefer, R. W., 1979, Remote Sensing and Interpretation (New York:
John Wiley).
Mah, A., Taylor, G. R., Lennox, P., and Balia, L., 1995, Lineament analysis of Landsat TM
images, Northern Territory, Australia. Photogrammetric Engineering and Remote
Sensing, 61, 761–773.
Maragos, P., and Ziff, R. D., 1990, Threshold superposition in morphological image analysis
systems. IEEE T ransactions on Pattern Analysis and Machine Intelligence, 12, 498–504.
Matheron, G., 1975, Random Sets and Integral Geometry (New York: John Wiley).
Qari, M. Y. H. T., 1991, Application of Landsat TM data to Geological Studies, Al-Khabt
area, Southern Arabian Shield. Photogrammetric Engineering and Remote Sensing,
57, 421–429.
Serra, J., 1982, Image Analysis and Mathematical Morphology (London: Academic Press).
Serra, J., 1986, Introduction to mathematical morphology. Computer V ision Graphics and
Image Processing, 35, 283–305.
Sternberg, S. R., 1986, Grayscale morphology. Computer V ision Graphics and Image
Processing, 35, 333–355.
Vincent, L., and Soille, P., 1991, Watersheds in digital spaces: an e cient algorithm based
on immersion simulation. IEEE T ransactions on Pattern Analysis and Machine
Intelligence, 13, 583–598.