Full text of DOI:10.1080/01431160120769

This article was downloaded by: [Boston University]
On: 26 October 2013, At: 15:04
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
Texture analysis of IKONOS
panchromatic data for
Douglas-fir forest age
class separability in British
Columbia
S. E. Franklin , M. A. Wulder & G. R. Gerylo
Published online: 25 Nov 2010.
To cite this article: S. E. Franklin , M. A. Wulder & G. R. Gerylo (2001) Texture
analysis of IKONOS panchromatic data for Douglas-fir forest age class separability
in British Columbia, International Journal of Remote Sensing, 22:13, 2627-2632
To link to this article: http://dx.doi.org/10.1080/01431160120769
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, 2001, vol. 22, no. 13, 2627–2632
Texture analysis of IKONOS panchromatic data for Douglas-Ž r
forest age class separability in British Columbia
S. E. FRANKLIN*, M. A. WULDER† and G. R. GERYLO*
*Department of Geography, University of Calgary, Calgary, Alberta, Canada,
T2N 1N4
†Canadian Forest Service, PaciŽ c Forestry Centre, Natural Resources Canada,
Victoria, British Columbia, Canada, V8Z 1M5
(Received 20 October 2000; in Ž nal form 2 February 2001)
Abstract. This Letter presents the results of textural separability tests obtained
by Ž rst- and second-order texture methods on diVerent aged Douglas-Ž r stands
in IKONOS panchromatic imagery acquired 3 June 2000 over the Sooke River
watershed in British Columbia. The eVects of diVerent measures and window sizes
on the textural separability are discussed. Small windows sizes were not as eVective
in separating stands as larger windows sizes. Second-order (spatial co-occurrence
homogeneity) texture values were the most eVective in distinguishing between age
classes. A Ž rst-order (variance) texture measure, though still useful, provided less
separability.
1. Introduction
Texture is a key visual criteria when interpreting information on the spatial
distribution of forest vegetation from aerial photographs (Graham and Read 1986,
Lillesand and Kiefer 1994). This use of texture  ows naturally from the powerful
innate ability that humans have in recognizing textural diVerences, although the
complex neural and psychological processes by which this is accomplished have so
far evaded detailed scientiŽ c explanation (Hay et al. 1996). With the increased
availability of new high spatial detail digital imagery, from airborne platforms (Anger
1999) and satellite platforms (Barnsley 1999), it is apparent that, increasingly, digital
texture should be considered a potentially important information source for forestry
purposes (Green 2000). Insight into how texture might be analyzed by computer has
focused on the structural and statistical properties of textures (Haralick 1986).
Texture is a complex multiscale phenomena (Ahearn 1988), but there are few guide-
lines available to help forestry users conduct a texture analysis with high spatial
detail digital imagery.
Forest stands have been separated by diVerences in digital texture related to
species composition (Franklin et al. 2000), age-class and structure (Cohen et al.
1990), and crown closure (St-Onge and Cavayas 1995, 1997). Our intention in this
Letter is to examine IKONOS panchromatic imagery, with an original image acquisi-
tion spatial resolution of 0.82 m, for textural diVerences in stands of Douglas-Ž r, to
International Journal of Remote Sensing
ISSN 0143-1161 print/ISSN 1366-5901 online © 2001 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/01431160110053590

2628
S. E. Franklin et al.
determine whether commercially-available Ž rst- and second-order texture measures
are useful in identifying diVerences related to age classes of trees. Knowledge of
diVerences in forest age class is currently considered a necessary component in
sustainable forest management planning and practices (e.g. Spies 1997).
2. Study area and methods
The study area is located in the Sooke region in the southern portion of Vancouver
Island, British Columbia, Canada (centered near 48.58 N, 123.72 W). This area is
°
°
located within the PaciŽ c Maritime ecozone. Within the forested land, the dominant
tree species include Douglas-Ž r, western red cedar, and hemlock (Wulder et al. 2000 ).
A cloud-free IKONOS panchromatic image was acquired on 3 June 2000 (Ž gure 1).
The solar elevation was 60.7 and the solar azimuth was 146.5 . The image was
°
°
georeferenced to a vector photointerpreted forest inventory coverage stored within
a geographic information system (GIS) database. Data in the GIS species composi-
tion and age class categories, as deŽ ned by the provincial forest inventory system
(Gillis and Leckie 1993), were used to stratify the database into pure Douglas-Ž r
stands for several age groups (table 1).
Figure 1. A sub-area of the IKONOS image showing locations of three diVerent age classes
sampled for texture; A indicates forest age class 2 (21–40 years), B indicates forest age
+
class 5 (81–100 years), and C indicates age class 9 ( 251 years)

Remote Sensing L etters
2629
Table 1. British Columbia forest inventory age classes.
Age class
Age limits (years)
1
2
3
4
5
6
7
8
9
1–20
21–40
41–60
61–80
81–100
101–120
121–140
141–250
+
251
Age class absent in study area.
Within three or four stands of each age class, 20 pixel locations were selected
near the stand centre. At these locations, one Ž rst-order (variance), and one second-
order (homogeneity) texture measures were derived:
(x x)
- 2
æ
variance
(1)
(2)
=
n
1
-
P(i, j)
(i j)
homogeneity
=æ æ
1
+ - 2
i j
where P(i, j) is the spatial co-occurrence matrix. Three diVerent moving window
sizes were used (5 5, 15 15 and 25 25) based on an initial analysis of Queen’s
×
×
×
case semivariograms (eight cardinal directions) to determine the approximate area
over which pixel variations occurred (Franklin et al. 1996). The high spatial resolution
of the imagery resulted in a semivariance spatial domain relating individual tree
rather than stand level characteristics. The principle behind using Ž rst-order (i.e.
variance-based) and second-order texture (i.e. derivative-based) is simple: ‘texture
information is contained in the overall, or ‘‘average’’, spatial relationship which gray
tones in the image have to one another’ (Haralick et al. 1973).
A one-way analysis of variance (ANOVA) was performed to determine if a
signiŽ cant diVerence (at p 0.05) existed between mean texture values for each age
<
class. The magnitude and signiŽ cance level associated with each ANOVA F value
were used to determine the most eVective texture measures, and window sizes, in
separating the various age classes. A larger F value indicates greater separation.
Since ANOVA results do not distinguish which age classes diVer from one another,
the Bonferroni multiple mean test was also applied to identify where signiŽ cant
diVerences occurred between class pairs. The Bonferroni multiple mean test adjusts
the overall signiŽ cance level based on the number of pairs being compared (Neter
et al. 1990).
3. Results
The ANOVA tests demonstrated that there was a signiŽ cant diVerence (at p 0.05)
<
in the mean texture values of the age classes for each texture measure, and window

2630
S. E. Franklin et al.
size combination, except the variance texture measure using a 5 pixel 5 pixel
×
window (table 2). Greater separation between age classes was achieved using the 25
pixel 25 pixel windows than the smaller pixel windows, as indicated by the larger
×
F values of the former. Further, greater separation between age classes was achieved
using the homogeneity measure (F 39.6, p 0.0001) than the variance measure (F
=
=
<
17.4, p 0.0001).
<
Tabular results provided from Bonferonni analyses, showing separability between
age classes based on the mean texture values obtained from each class are lengthy,
and consequently a detailed presentation of all results was not practical here. Instead,
a summary of the trends observed is shown graphically in Ž gure 2 in which the grey
regions within each triangle indicate a signiŽ cant diVerence (at p 0.05) existed
<
between the mean texture values for the speciŽ ed age class pair. No signiŽ cant
diVerence existed between mean texture values obtained from each age class with
the 5 5 variance texture measure. However, as the variance window size was
×
increased, the number of signiŽ cantly diVerent age class pairs increased. For example,
age class 9 was found to be signiŽ cantly diVerent from all other age classes with the
variance texture measure in a 15 15 window. As the window size was increased to
×
25 pixels 25 pixels, the younger age classes (2 and 3) were distinguished from the
×
youngest age class (1), and a signiŽ cant diVerence was detected between age class 1
and the two oldest age classes (8 and 9). Again, class 9 was statistically diVerent
from all other classes. While these results were relatively poor in respect to separation
of many classes (i.e. fewer classes were separable), they followed the trend of increasing
F value and pair separation with an increased window size.
The trend was demonstrated further with the homogeneity texture measure.
Distinction between mean homogeneity values from each age class pair increased as
the window size increased from 5 5 to 15 pixels 15 pixels. At the smallest window
×
×
size, only young and old age classes could be separated (i.e. age classes 1 to 3 paired
with age classes 6 to 9). However, as window size increased to 15 pixels 15 pixels,
×
the middle age classes could also be separated from the younger and older age
classes. While the magnitude of the ANOVA F value was highest with the 25
pixel 25 pixel homogeneity measure, no additional distinction between remaining
×
age class pairs was provided; the Bonferroni test did not show increased separation
between any other age class pairs as a result of the more signiŽ cant distinction
between age class pairs resulting from a larger diVerence in mean texture values.
4. Conclusions
Texture features in high (Ž ne) spatial resolution airborne and satellite sensor
imagery contain forest information, and may be employed to increase the utility of
high spatial resolution imagery as a useful source of information. Digital texture is
Table 2. ANOVA results.
Variance
F
Homogeneity
F
Window size
×
5
15 15
5
×
1.7
10.1*
17.4*
5.6*
30.0*
39.6*
×
25 25
<
Marked (*) F values signiŽ cant at p 0.05.

Remote Sensing L etters
2631
Figure 2. Graphical results of Bonferroni multiple mean comparison test; shaded areas
indicate signiŽ cant separability between two age classes.
a complex concept and application in remote sensing. Tools for texture analysis in
most public or commercially-available image processing systems are fragmented and
poorly developed, but many such systems provide Ž rst- and second-order texture
functionality. We hypothesized that these relatively simple textural measures, with
diVerent window sizes, would be useful in separating diVerent forest age classes in
IKONOS panchromatic imagery. Second-order (spatial co-occurrence homogeneity)
texture obtained from the larger pixel windows was most eVective for distinguishing
stands in diVerent age classes. A Ž rst-order (variance) measure provided a weaker
separation over the same window sizes. The variance texture measure may be
complementary to the homogeneity measure.
Acknowledgments
This research was funded by the National Sciences and Engineering Research
Council of Canada and the Canadian Forest Service. Helpful comments by the
anonymous reviewers were greatly appreciated.

2632
Remote Sensing L etters
References
Ahearn, S. C., 1988, Combining Laplacian images of diVerent spatial frequencies (scales):
implications for remote sensing analysis. IEEE T ransactions on Geoscience and Remote
Sensing, 26, 826–831.
Anger, C. D., 1999, Airborne hyperspectral remote sensing in the future? In Proceedings,
Fourth International Symposium on Airborne Remote Sensing and 21st Canadian
Symposium on Remote Sensing (Ottawa: Canadian Aeronautics and Space Institute),
pp. 1–15.
Barnsley, M., 1999, Digital remotely-sensed data and their characteristics. In Geographical
Information Systems, 2nd edn, edited by P. A. Longley, M. F. Goodchild, D. J. Maguire
and D. W. Rhind (New York: John Wiley), pp. 451–466.
Cohen, W. B., Spies, T. A., and Bradshaw, G. A., 1990, Semivariograms of digital imagery
for analysis of conifer canopy structure. Remote Sensing of Environment, 29, 669–672.
Franklin, S. E., Wulder, M. A., and Lavigne, M. B., 1996, Automated derivation of
geographic window sizes for remote sensing digital image texture analysis. Computers
& Geosciences, 22, 665–673.
Franklin, S. E., Hall, R. J., Moskal, L. M., Maudie, A. J., and Lavigne, M. B., 2000,
Incorporating texture into classiŽ cation of forest species composition from airborne
multispectral images. International Journal of Remote Sensing, 21, 61–79.
Gillis, M., and Leckie, D., 1993, Forest inventory mapping procedures across Canada.
Petawawa National Forestry Institute, Information Report PI-X-114. Canadian Forest
Service: Ottawa, Ontario.
Graham, R., and Read, R. E., 1986, Manual of Aerial Photography (London: Focal Press).
Green, K., 2000, Selecting and interpreting high-resolution images. Journal of Forestry, 98,
37–39.
Haralick, R. M., 1986, Statistical image texture analysis. In Handbook of Pattern Recognition
and Image Processing, edited by T. Y. Young and K. S. Fu (New York: Academic
Press), pp. 247–279.
Haralick, R. M., Shanmugam, K., and Dinstein, I., 1973, Texture features for image
classiŽ cation. IEEE T ransactions on Systems, Man, and Cybernetics, 3, 610–621.
Hay, G. J., Niemann, K. O., and McLean, G. F., 1996, An object-speciŽ c image texture
analysis of H-resolution forest imagery. Remote Sensing of Environment, 55, 108–122.
Lillesand, T. M., and Kiefer, R. W., 1994, Remote Sensing and Image Interpretation, 3rd edn
(New York: Wiley).
Neter, J., Wasserman, W., and Kutner, M. H., 1990, Applied L inear Statistical Models, 3rd
edn (Boston: Irwin).
Spies, T. A., 1997, Forest stand structure, composition, and function. In Creating a Forestry
for the 21st Century, the Science of Ecosystem Management, edited by K. A. Kohm
and J. F. Franklin (Washington: Island Press), pp. 11–30.
St-Onge, B. A., and Cavayas, F., 1995, Estimating forest stand structure from high resolution
imagery using the directional variogram. International Journal of Remote Sensing, 16,
1999–2021.
St-Onge, B. A., and Cavayas, F., 1997, Automated forest structure from high resolution
imagery based on directional semivariogram estimates. Remote Sensing of Environment,
61, 82–95.
Wulder, M., Niemann, K. O., and Goodenough, D., 2000, Local maximum Ž ltering for the
extraction of tree locations and basal area from high spatial resolution imagery.
Remote Sensing of Environment, 73, 103–114.