Use of the "gain and bias" programme module influences the image data in direc-
tion of brightness and contrast. Technically, a roller ball which can be moved
in x and y direction achieves an electronic weighting of the three image spec-
| tral channels by linear transformation, until on a monitor-screen, the subjec-
tively best image is reached (photo 2).
Histogram-stretching of the image raw
data gives eben better results in color,
contrast, as well as improving visibi-
EN lity of structures. This is reached
by stretching the original data, which
oecupies only a fraction of the po-
tential greytone scale in the parti-
cular spectral band, over the whole
range. Stretching is done for each
spectral band according to the general
formula A'= (A - x):y, where A' and A
are respectively the stretched and the
original histograms, and x and y are fac-
tors selected according to the origi-
nal data (photo 3).
y Histogram stretching proves to be an
effective pretreatment for visüal
image interpretation and all following
methods of image manipulation.
|. area
5. PROCESSING FOR SPATIAL PATTERN Photo 3: Histogram-stretching on the
ANALYSIS basis of photo 1 (color image)
ons The principal component transformation is a method of the multivariate statis-
8. ties for data-reduction. From the original four variables (the four spectral-
channels), four new variables (principal components) are created. The first
principal component contains the greatest variance, i.e. information, while
the second, third and fourth principal components successivley try to maximize
the remaining variance. With LANDSAT images, frequently more than 90 % of the
variance is contained in the first two principal components (DONKER and MULDER,
1976), so in two new images, nearly the whole information of the original four
spectral bands is found.
ctively |
) shows
on
Photo 4: Principal component 1. Photo 5: Principal component 2,
28. 8. T5, (black + white) 20. 6. T5, (black + white)
399
sg m (Mur RR
