underdetermined, 
roduced. Here, we 
ace model, (dz/dx)? 
ich is minimized at 
he quantities in this 
n particular, z is the 
nverged adequately 
ition algorithm, an 
unavoidable error 
s generated at each 
uncation error can, 
lution based on the 
residuals were less 
of the iterations, a 
et al., 1996) with 2 
sed to estimate the 
metric function that 
ss, to be used in the 
ed with photometric 
(lunar) and Lambert 
AcEwen, 1991), and 
nage was adopted to 
rat image. This was 
d scattering model 
ng albedo across a 
These results lend 
operties have been 
two stereo pairs of 
ns (termed I, II, and 
etric terrain models 
1age correlation was 
of the study area. 
> texture and large 
in the lower part of 
| fail, and therefore 
, and 22 pixels on 
lyzed: While large 
n smoothing effects 
edium-sized craters 
sy at a patch size of 
size of 14 pixels for 
dels 
(Fig. 2a, 3a, and 4a) 
nometry reflect the 
er differently. The 
bed but clearly show 
logrammetry models 
y large and medium 
opography seems to 
I and II. 
son between the two 
specific image lines 
ht differences with 
neters occur (cf. Fig. 
ires such as the large 
This crater is more 
1996 
than 50% deeper in the photogrammetry model than in 
photoclinometry. Small-scale craters in the range of 10 pixels 
are fully recovered by "shape-from-shading" but are not seen in 
the photogrammetry models. Very small features covered by 
less than 5 pixels are not even resolved by photoclinometry. 0 
He 
N Jm 
MN 
| 
   
  
    
I 
I 
37 m/pixel 
region III 
region II 
59 m/pixel 
  
€ 1450 
= 
> 
= 
1300 
region I 125— 
line I 
38.5 m/pixel line II 
141000 
13.8 
12.6 
E 13.4- 
5 n 
X 
120 
Fig.1: This figure shows the images involved in the matching 12.8 ] 
procedure. “The images at the top and bottom were taken as a ] 
reference, the search image is shown in the center of the figure. 12.6 
Only regions I, II, and III were considered for matching. 
14 x1000 4 
  
    
    
profile I 
  
    
Fig.2a. Region I :terrain models derived from 
photogrammetry (top) and 2D photoclinometry (bottom). 
+ PHOTOGRAMMETRY 
—— 2D PHOTOCLINOMETRY 
  
j 
4 
   
    
  
PHOTOGRAMMETRY 
—— 2D PHOTOCLINOMETRY 
  
  
0 
100 
sample [pixels] 
Fig.2b. Region I: height profiles along image lines 
247 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996