H OJ oo_=t- 
526 
R = (x/127) (R 
- R • ) 
mm 
+ R 
mm, 
(1) 
where x is expressed in terms of the decomposed data value 
x', the absolute offset a, and the absolute gain b of the 
detector 
x = (x' - a)/b. 
(2) 
Numerical values of parameters such as R ,R . a, and b 
are provided by several authors(cf. Ahern and m Aufphy 1978; 
Robinove et al. 1981). Furthermore, the net flux F of solar 
radiation per unit area normal to the direction of propaga 
tion in each band is listed as below (cf. Otterman and Fra 
ser 1976) 
Table 1. Solar spectral flux outside 
Earth's atmosphere in each band 
2 
Wave length(pm) 
0.5 - 0.6 
0.6 - 0.7 
0.7 - 0.8 
0.8 - 1.1 
F(mw/cm ,sr,ym) 
17.70 
15.15 
12.37 
24.91 
Geometric Correction 
When a digital image of an area on terrestrial surface is 
generated by a spaceborne sensor, the image will involve geo 
metric distorsions due to the sensor platform, the sensor 
characteristics, and scene effects. The digital image data 
should be modified to remove these geometric distorsions su 
ch that a point in the input image can be related to its true 
position on Earth's surface to a prescribed accuracy, gene 
rally one-half pixel. Geometric correction is used to cor 
rect the geometric distorsions. 
In what follows we deal with digital correction algorithms 
that were developed to correct Earth observation data. Ground 
Control Points (GCP) are used to get external reference in 
formation, because of the physical feature identified in a 
scene such as the location and elevation precisely known. 
Determination of such parameters as the spacecraft roll, pi 
tch, yaw data and the ephemeris data will be computed from 
knowledge of the GCP locations. 
The (u,v) coordinates of each input pixel are the column and 
row indices of its grey level stored in the input data array. 
On the other hand, the geometrically correct output image 
is also defined as an uniform, two-dimensional array (x,y) 
of pixels. On making use of the bivariate polynomials,the 
(u,v) input image coordinates of the pixel are determind 
at coordinates(x,y) in the output image. In our case we re 
ferred to the affine linear transformation 
u = a^xy + a^x + a 3 y + , (3) 
v = b^xy + b 2 x + b 3 y + b 4 - (4) 
With the aid of the Gaussian least-squares algorithm, such as 
a^ and b^' (i=i,2,3, and 4) were numerically computed, al