this area is domin ated by a beech-m aple climax fores t ( Fagus grandi folia Ehr h–Acer
saccharum Mars h.) sit uated on uncult ivated moraine isl ets, with cereal crops grow n in
the rich low land marine clay deposi ts of the Champl ain Sea (Mei lleur et al., 1994). In
Figure 8.3 a, two thin dark-t oned roads diagon ally divide the scene into three main
sections, whi le numer ous h edge row s (narr ow brigh t text ured d iagonal linear features)
separate smoo th-toned agricultu ral fields and textured fores ted areas. At the top left of
the ima ge, a (wide , smoothl y textured) porti on of an elect rical utilit y corridor is
visible, while near the image ce nter, a brigh t-toned corn field appears illuminate d
when compa red to its neighb oring darker toned alfalfa fields.
8.2.2 Linear Scale Spa ce
Sc ale Space (SS) repr esents an uncommitted framework for ear ly visual operations
that was developed by the computer vi sion community to automatically analyze
real-world structures at multiple scales ( t ), speci fi ca lly w hen there is no a priori
information about these structures or t he appropriate s cale(s) f or thei r analysis
(Lindeberg, 1994). The term uncom mitted f ramework ref e rs to obs er vati ons made
by a f ront-end vi sion system (i.e., an initial-stage measuring device) such as the
retina or a camera that involves “ no know ledge” and “ no preference” for anything.
In this text , we emphasize linear SS, though we note that a signi fi cant body of work
exists on nonlinear SS. Interested readers are referred to a general defi ni ti on o n
Wi kipedi a
9 as well as to more recent, related l iterature ( Tzot sos et al., 2011; Doxani
et al., 2012).
8.2.2.1 Gaussian Operators When scale information is unknown within a scene,
the only reasonable approach for an uncommitted vision system is to represent the
FIGURE 8.3 IKONOS subimage and study site map: (a) 500 ´ 500-pixel 4.0-m panchromatic image; (b) map location of image.
9 http://en.wikipedia.org/wiki/Scale_space, last accessed April 21, 2013.
METHODS
149
saccharum Mars h.) sit uated on uncult ivated moraine isl ets, with cereal crops grow n in
the rich low land marine clay deposi ts of the Champl ain Sea (Mei lleur et al., 1994). In
Figure 8.3 a, two thin dark-t oned roads diagon ally divide the scene into three main
sections, whi le numer ous h edge row s (narr ow brigh t text ured d iagonal linear features)
separate smoo th-toned agricultu ral fields and textured fores ted areas. At the top left of
the ima ge, a (wide , smoothl y textured) porti on of an elect rical utilit y corridor is
visible, while near the image ce nter, a brigh t-toned corn field appears illuminate d
when compa red to its neighb oring darker toned alfalfa fields.
8.2.2 Linear Scale Spa ce
Sc ale Space (SS) repr esents an uncommitted framework for ear ly visual operations
that was developed by the computer vi sion community to automatically analyze
real-world structures at multiple scales ( t ), speci fi ca lly w hen there is no a priori
information about these structures or t he appropriate s cale(s) f or thei r analysis
(Lindeberg, 1994). The term uncom mitted f ramework ref e rs to obs er vati ons made
by a f ront-end vi sion system (i.e., an initial-stage measuring device) such as the
retina or a camera that involves “ no know ledge” and “ no preference” for anything.
In this text , we emphasize linear SS, though we note that a signi fi cant body of work
exists on nonlinear SS. Interested readers are referred to a general defi ni ti on o n
Wi kipedi a
9 as well as to more recent, related l iterature ( Tzot sos et al., 2011; Doxani
et al., 2012).
8.2.2.1 Gaussian Operators When scale information is unknown within a scene,
the only reasonable approach for an uncommitted vision system is to represent the
FIGURE 8.3 IKONOS subimage and study site map: (a) 500 ´ 500-pixel 4.0-m panchromatic image; (b) map location of image.
9 http://en.wikipedia.org/wiki/Scale_space, last accessed April 21, 2013.
METHODS
149