IMPROVEMENT OF THE HEIGHT MEASUREMENT ACCURACY 
FOR THE STEROSCOPIC IMAGES INCLUDING RANDOM NOISE 
Kazutaka Adachi, Riichi Nagura 
Department of Computer Engineering 
Faculty of Engineering 
Himeji Institute of Technology 
2167, Shosha, Himeji, 671-2201, Japan 
E-mail: naguraQcomp.eng.himeji-tech.ac.jp 
Commission VI, Working group No.4 
KEYWORDS: Height measurement, Correlation analysis, Signal to Noise ratio, 
Adaptive filter, Imaging angle 
ABSTRACT 
This paper reports the improvement of the height measurement accuracy for the stereoscopic images 
from satellite including random noise. In the high resolution observing system, the signal input level 
from the earth’s surface to the satellite decreases with high resolution, and the random noise tend to 
relatively increase. Generally, filtering methods are effective to reduce random noise, but the filtering 
causes the deterioration of an edge characteristic on the image signal. 
In this paper, we propose an adaptive filtering method that automatically selects different 
characteristic filters fitting to geographical features. Namely, the wide band filter is applied to the case 
of a steep topography such as a top of mountain, and for the relatively flat region, narrow band filter is 
applied. 
We simulate the proposed method with numerical height pattern, and show that the proposed method 
  
is effective to get high accurate result from the stereoscopic images including random noise. 
1.INTRODUCTION 
The observation from satellite can obtain the 
wide range of the earth's surface data and is 
hopeful to use for making a global terrain map or 
other fields. But the signal input level from the 
earth's surface to the satellite decrease with high 
resolution. Consequently, random noise tends to 
relatively increase, when large optical system is 
difficult to be loaded into satellite. In high 
resolution observing system, the investigation of 
Signal to Noise Ratio (S/N) is one of the most 
important issues. That is to say, to get more 
accrate 3-D height information, we are hoping to 
methods which can reduce the influence by 
random noise effectivly. 
Generally, it is effective to put a low pass filter 
(LPF) for noisy image to get more accurate 3-D 
information. Filtering methods reduce random 
noise's influence, but the filtering causes the 
deterioration of an edge characteristic on the 
image signal in a steep region such as a top of the 
mountain or a ridge. Thus, we make use of specific 
character that is not affected by random noise in a 
steep region and is affected by them in a flat. 
In this paper, we propose an adaptive filtering 
method that selects different characteristic filters 
automatically fitting to geographical features. The 
proposed method can obtain more accurate 3-D 
information without the deterioration of an edge 
characteristic on the image signal. 
In the correlation analysis methods, it is 
important to decide to the imaging angle of 
satellite system. In this paper, we also investigate 
the imaging angle for correlation analysis and 
report the result. 
2.SIMULATION 
2-1.CORRELATION ANALYSIS 
The target object's height information can be 
obtained by using stereoscopic images. The 
principle of correlation analysis method is shown 
in Fig.1. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 309