generated.
These result are on the spatial
image. So that transformation from the
spatial image to fundus is necessary. This
transformation is done by using shift value of
refractive disorder calculated on the process.
The relation between shift value and
transformation factor (ß) is measured in
advance. The relation between spatial and
fundus on position and height are as below.
(position on fundus) = (position on spatial) X B
( height on fundus ) = ( height on spatial ) X g
6) Analytical processing for diagnosis
The results of stereo matching can
be presented on the monitor as a natural
colored 3D image. Thus, it offers oculists
realizable condition image of the cup and disc
in the area. In addition, it is also possible to
display the wire frame, the contour lines,
cross-sectional drawings, and so on.
One side, arranged data obtained
from the stereo matching can be utilized for
diagnosis in routine work or as diagnosis soft
ware. Other side, by using the interactive
analysis function, many measured data can
be utilized for various clinical applications.
For the glaucoma diagnosis, after
the drawing a disc outline by manual on the
displayed ortho photo in the system,
analytical values and lines as below are
automatically calculate.
3 Cup outline
3t Horizontal cross section
3 Vertical cross section
3 Nerve fiber analysis
# Max. disc diameter ( vertical )
# Max. disc diameter (horizontal )
# Cup / Disc ratio ( vertical )
# Cup / Disc ratio ( horizontal )
# Disc area
# Cup area
# Rim area
# Cup / Disc area ratio
# Rim / Disc area ratio
# Max. cup depth
# Cup volume
Also these analytical processing as
below are incorporated in the svstem.
# Contour lines
# Multi cross sections
# GUI functions
# List up of diagnosis history for a fundus
5. Examination
562
Examination for the system is done
using each images by film scanner and CCD
camera.
Table - 1 Examination image
Pixel X line | resolution
film scanner | 3883 X 2592 94 um
CCD camera | 768 X 480 47.5 u .m
From the result for matching used
these. image, remarkable feature points are
plotted on the image as figure - 13. These
images show matching is done exactly.
Also bird eye view images are made
from DEM and image taken CCD camera and
two enlarged images by interpolation from
same images are shown as figure - 14. This
proves the interpolation is effective.
The system configuration using
these examination is table - 2 and figure - 12.
And process time for one image is as table - 3.
Table-2 Hard ware
Item Name
fundus camera Nidek 3-Dx
CCD 1/2 inch 3CCD ( SONY DXC-
930)
Still Video Recorder |FUJIX DF-10M
Computer PC
CPU i486 DX2 66MHz
Memory 24MB
Table - 3 Result for the processing time
pixels resolution |DEM analysis
[4 m] time time
[second] |[second]
CCD 768 X 480 475 90 120
scanner | 1024X612 35.6 180 340
5. Accuracy
Dummy model of eyes are made to
confirm matching accuracy. One of the
dummy model of eye is two laminated small
discs shaped doughnut figure which
thickness is 0.5mm each (fig - 11). The inner
diameter of each doughnut is different 1 mm
and 2 mm. And random burns on doughnut
laminates are made by laser beam. Using
this dummy model of eye, matching without
interpolation is done. And profiles are
measured on four directions. The average of
profile value is refer to nominal value of
dummy model of eye. So that horizontal
direction is not good because of occlusion by
cliffs of doughnut, however vertical direction
in around 50 u m.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
