A study and development for STEREO FUNDUS CAMERA SYSTEM 
using Stereo Matching Technique. 
TSURU Kosuke, Inoue T., Ogino N., Matuyama M., 
Aero Asahi Corp. 
Saitama Kawagoe Minamidai 3-1-1 Japan 
Nanjo T., Yano. N., Hino T. 
Nidek Co., LTD. 
Aichi Gamagori Hiroishicho 34-14 Japan 
ISPRS XVII Commission V 
Abstract 
Recently, ME equipment are popularized rapidly in field of medical treatment. 
disease can be treated simply at present. But in case of glaucoma diagnosis, the cases of 
eye 
All 
disease is not clearly known and damages badly. Therefore analyzing system is urgency need. So 
we developed STEREO FUNDUS CAMERA SYSTEM using stereo matching technique. 
This system takes stereo photos with a ocular fundus camera ( special stereo camera ), 
create DEM and ortho photo automatically by using stereo matching technique. And the system 
allow to measure and/or analyze symptoms for glaucoma diagnosis using super imposed DEM to 
ocular fundus images. Also stored data is worth because of automatically processing has high 
reproducibility. and the extraction of temporal change of symptoms is easily possible. 
KEY WORDS: Medicine, Automation, Matching, DEM/DTM, Close_Range, Optical 
1. Introduction 
Measurements of the ocular fundus 
(fig 1) by photogrammetry using an 
analytical plotter already is done (Takamoto 
1986). And digital system on 
photogrammetry is developed and tried 
(Nanba 1989). ITowever both method is doing 
manual way or require to the assistance by 
skilled operators. So that the processing time 
to get the result is big. Or using the systems 
are required specialists. 
On the other hand, analysis of 
photo images using image correlation 
technique is being developed as the method 
to generate contours automatically using 
aerial photographs from the end of 1950's 
(Mori 1981, 1982. 1985). As well as there are 
studies using stereo images by liner array 
sensor (Murai 1984, 1987). These purpose are 
automation of stereo matching from stereo 
images. and arca correlation method is used 
for stereo matching technique. 
Using area correlation for map 
automation, the decision for window size that 
makes the correlation is difficult. Also there 
is problems of occlusions like valley and 
building which hide opposite side of project 
center. To solve these problems, image 
filtering and/or step corrclation ( Course-to- 
558 
fine ) methods are adopted. 
Analysis for the fundus of the eye is 
also taken height information from a pair of 
stereo photos using area correlation, but 
there are some remarks and problems. 
At first, making maps are require 
the result to keep high accuracy. However in 
fundus analysis, first purpose is to monitor 
the fundus condition. And the system is to be 
used on site medical examination and 
treatment, so that speedy processing is 
require. Also custom friendly operation is 
require because of oculists whose are 
nonprofessional for the computer system. 
On second, a fundus image is taken 
photograph through the human crystalline 
lens and pupil of a patient. This means the 
crystalline lens assumes one of the camera 
lenses in the system. Further two bundle of 
ray for stereo photographing have to through 
the pupil, this means photo base length is 
very short. 
On third, fundus camera itself is 
developed for monitoring but not for 
measuring or analysis, so that the camera 
lacks high accuracy. Making a dummy model 
of eve to get camera calibration and 
orientation factors are very difficult, because 
of the dummy model of eye itself is very small, 
and there are some problems like no way to 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996