Page 289 - Digital Analysis of Remotely Sensed Imagery

Spectral Image Analysis 251

Photo Description/Math
Element Graphic Sample Expression Utility
Tone/color DN in Sole clue in
multispectral parametric
domain classifiers
Size Number of Not useful in
spatially the spectral
contiguous pixels domain

Shape Hard to Area/ Not useful
perimeter
ratio Size/
perimeterww
Shadow DN of a small Not useful
value, mostly
noise

Texture Standard Useful for
deviation of identifying
DN or spatial certain covers
autocorrelation of a unique
texture
Pattern Could be Useful in
measured as pattern
spatial adjacency recognition
Location/ Logical Restricted use
association expressions in in identifying
the form of “IF … certain
THEN …” features
TABLE 7.1 Photo Elements, Their Mathematical Expression, and Their Utilities in
Spectral Image Analysis.
as digital number (DN) in the digital context. According to the con-
vention, a pixel of a small value in a spectral band has a dark tone
(e.g., shadow). A large pixel value has a bright tone (e.g., cloud). The
potential range of all digital numbers is governed by the quantization
level of the sensing system (refer to Sec. 1.4.2). If a sensing system uses
8
8 bits to record Earth resources satellite data, there are 2 (256) possible
gray levels ranging from 0 to 255. All pixels in spaceborne multispec-
tral images have multiple DNs, each corresponding to a spectral band
in the multispectral domain. Since each multispectral band has a
unique wavelength range, the value of this pixel likely varies from
band to band. The magnitude of variation in DN from a band to the
one immediately adjacent to it is determined by the radiometric resolu-
tion of the remote sensing system. The specific value of a pixel’s DN is