N A t xem a 
um) 
tween sen- 
| Of 
using itera- 
bsorbing 
ctions as a 
easured tem- 
rence between 
nospheric 
higher ra- 
ted by 
(2) 
JETTNER 
e e — ES EUR TR ir A 
  
and KERN 1965). Computation of atmospheric radiance was performed with 
H (3) 
= 9T : 
G I L, (T, (h)) Sn sec (8)dh with © = scan angle 
an data from McCLATCHEY et al. 1972 and vertical radiosonde profiles of Hannover 
3.76, 
The hemispheral radiance is estimated with 
7/252 | (4) 
G=2r J/ J G, cos z sin z dA dz with z = zenith angle. 
0 À 
1 
T of equation (2) is found again iteratively. 
Fig.3 shows the correction as a function of the apparent surface temperature. 
  
   
  
  
  
  
  
  
  
  
  
  
5 AT 
9 
3 
8 
34 
& 
us = 
Y v 
z E € - 0,954 
Ze 2 o s 
£g 1 $ «= 0,972 
Q^ & 
o V- 
5 €= 0,990 
8 5 
3 ? 
5 15 16 17 18 19 20 21 22 23 24 25 26 
8 | ] Scheinbare Oberflüchentemperatur in 9Celsius 
o T YT Y, T , YT T T 1, Y 
10.00 12.00 4.00 16.00 18.00 20.00 
T GEMESSEN 
Fig. 2: Temperature differences as Fig. 3: Correction of reflection 
function of scan angle 
Because of the temperature dependance of the correction algorithm the image data 
have to be filtered first in order to avoid wrong corrections in the noisy parts. 
A student test of the image data shows that 71 % of all pixel do not fulfill the 
following condition: 
X - À 68.3 “i SX s X + À 68.3 *g 
with X = mean value of 3x3 neighbourhood 
Oo = standard deviation 
XM - center pixel of a 3x3 neighbourhood 
> 
I 
fractions of the student distribution at probability p. 
"FILTERING OF SCAN-LINE-NOISE 
Spectral Analysis 
Filtering of scan-line-noise with usual techniques like 'moving average' is not 
sufficient (EHLERS and LOHMANN 1982). Using spectral analysis a filter function 
shall be designed for suppressing scan-line-noise. Therefore the image is trans- 
formed by a Fast-Fourier-Transform (FFT) into the frequency domain. The Fourier 
coefficients for an image with size NxM can be written as 
137