Page 266 - Digital Analysis of Remotely Sensed Imagery

228 Cha pte r S i x

be registered to the same coordinate system and resampled to the
same spatial resolution using the methods introduced in the previous
chapter before they can be manipulated.
Multiple-image manipulation may be performed on individual
pixels aspatially. The two input images can be manipulated using a
wide range of arithmetic operations, such as addition, subtraction,
multiplication, division, or their combination. The DN of every pixel
in one image is added to, subtracted from, or multiplied/divided by
the DN of the corresponding pixel in another image. Subtraction is
commonly used to detect edges (see Sec. 6.4.1). It also finds applica-
tions in change detection, which will be discussed in detail in
Chap. 13. By subtracting one image from another, it is possible to
detect variations in the scene. Division is an operation of comparing
multitemporal images. It is better than image subtraction in detecting
changes from images that are recorded at different seasons and at dif-
ferent times of the day. Pseudoimages caused by the change in
shadow length as a result of these differences are able to be elimi-
nated partially through ratioing. Furthermore, it is possible to com-
bine image subtraction with image division to achieve even better
results.

6.5.1 Band Ratioing
Band ratioing refers to division of one spectral band by another
from the same sensor, preferably obtained at the same time. Prior to
division, the two bands must be coregistered precisely if they come
from separate sensors or cover a different ground area. After precise
coregistration, a pixel in one image corresponds to its counterpart in
another image. The ratioing of one image by another means the
pixel value at the same location is divided by one another. After
division, all pixel values that are expressed as a ratio between 0 and
1 may have to be rescaled to 0 to 255. Band ratioing is able to achieve
several purposes, dependent on the nature of the input bands. If the
two bands are obtained at different times, band ratioing is effective
at detecting changes that have taken place during the interval
(change detection will be covered in detail in Chap. 13). If the two
bands are from the same sensor, then this process is effective at
eliminating radiometric variations caused by topography (Fig. 6.19).
The sunlit slopes have a brighter tone than the shadows in the same
bands. However, after the two bands are ratioed, the same feature has
the same or nearly the same values in the resultant image while the
spectral disparity between different features is enlarged. In addi-
tion, band ratioing is also effective at partially eliminating the
impact of atmospheric radiance. For instance, if the atmospheric
effect causes pixel DN values to be 3 higher across all pixels, then a
division of (95 − 3)/(102 − 3) yields a ratio of 0.929. This value is
extremely similar to the ratio of 0.93 derived from the division of
raw DN containing the atmospheric effect.