Cowen et al. (1995) report that there needs to be a minimum of
four spatial units (e.g., 4 pixels) within an urban object area to
effectively be identified using a remotely sensed image. In other
words, the sensor spatial resolution needs to be at least one-half the
diameter of the smallest object of interest. For example, if we need to
identify a mobile home (an urban object) that is 4 m wide, the
minimum spatial resolution of high quality imagery without haze or
other atmospheric problems would be 2.0 by 2.0 m (Fig. 1). This
implies that the required spatial resolution of remotely sensed data to
prepare an urban land use and land cover map needs to be at least one
half the size of the smallest object to be identified in the image. This
operational concept or real world situation is not precisely in line with
the theoretical definition of the spatial resolution as the smallest
linear separation between objects that can be separated in an image
(Jensen, 2005; Lillesand et al., 2008). In a real world situation, a 4 m
wide object to be identified in an image most likely does not locate
perfectly over 4 pixels in a 2 mresolution image as illustrated in Fig. 1.
Hence, we need a pixel size that is remarkably smaller than an object
to identify that object in an image. Since urban objects are notably
smaller in comparison to natural features, it is apparent that we need
a significantly small pixel size for urban applications.
The geometric elements of image interpretation (e.g., pattern,
shape, size, and orientation) are important when using highresolution
image data for urban applications. However, the question
of whether we should evaluate the usefulness of a given type of
imagery (e.g., Landsat Thematic Mapper and IKONOS) for extracting
specific types of information (e.g., swimming pools) based solely on