Image Processing Reference
have been utilized as an additional input in the classification process (Kontoes et al.
2000 ; Orun 2004 ; Shaban and Dikshit 2001 ; Stefanov et al. 2001). Stefanov
et al. ( 2001 ) showed that texture measures of variance, generated from ASTER
imagery, could be used to study levels of urbanization in cities (i.e., decentralized,
centralized). Spatial pattern analysis is another method for analyzing image struc-
ture by quantifying the spatial arrangement of image brightness components in an
image. This can be accomplished using geo-statistical techniques such as semi-
variogram analysis as discussed by Brivio and Zilioli ( 2001 ).
Studies of urban change over both space and time have become increasingly
important as the world's urban population continues to grow. The anatomy of spa-
tial growth and change in Bangkok, Thailand was studied using Landsat 5-TM
imagery (Madhaven et al. 2001 ). The researchers utilized two classification meth-
ods, a traditional maximum likelihood supervised classification and a biophysical,
vegetation-impervious-soil (VIS) approach based on the model developed by Ridd
( 1995 ), which was reviewed in detail in Chapter 6. Wilson et al. (2003) developed
a geo-spatial model to quantify and map urban growth. The model, which utilized
Landsat TM imagery, provided researchers with a powerful visual and quantitative
analysis of both the type and extent of urban sprawl.
Other methods have been developed with the aim of improving change analysis
studies. Westmoreland and Stow ( 1992 ) incorporated ancillary data with fused
SPOT panchromatic and Landsat TM multispectral images in an integrated image
processing and GIS platform to aid in the visual interpretation of changed land use.
A framework was developed to aid in the interactive identification of land use cat-
egories and land use updates through the simultaneous display of digital image data
with the vector land-use data to be updated. In a study of urban expansion and flood
risk assessment in Nouakchott, Mauritania, Wu et al. ( 2003 ) utilized image fusion
techniques. SPOT panchromatic imagery were fused with multispectral imagery for
two study dates, 1995 and 1999. The fused imagery were then incorporated with
ancillary demographic data in a GIS to map areas of expansion and flood risk.
Urban Population Studies
Humans continuously modify their living environment. Remotely sensed imagery,
both day and night-time, can be used to study urban populations, their size, density
and distribution, through the characteristics of their environments. Urban populations
have been studied using image-derived variables from color infrared photography,
Landsat MSS and TM, and SPOT imagery, all captured under daylight conditions.
Ogrosky ( 1975 ) suggested the usefulness of satellite
imagery for estimating urban populations. He showed
how imagery derived variables such as urban area
boundaries, transportation links, urban area of nearest
largest neighborhood (NLN), and highway distance to
the NLN were strongly associated with population
count data taken from the census. High-altitude color
have been developed
since the mid-1970s
to estimate urban