Purpose. To develop computer algorithms for reconstructing 24-bit colo
r, wide-angle composite retinal fundus images from a set of adjacent 4
5 degrees fundus slides, The authors present the description, technica
l details, and results of the image reconstruction technique. Methods.
Patients with retinal degeneration underwent fundus photography with
a 45 degrees field-of-view fundus camera. Individual photographic slid
es were digitized for creating fundus montages. Background variations
in individual 45 degrees images were modeled to first- or second-order
two-dimensional polynomial functions to generate a background image.
The background image was subtracted from the original image to obtain
background corrected image. Background corrected images were registere
d and spatially transformed using a first- or second-order two-dimensi
onal polynomial warp model to reconstruct a composite retinal fundus m
ontage. Results. The authors successfully reconstructed 24-bit color,
100 degrees field-of-view, composite retinal fundus images. The comput
er-reconstructed montages are an improvement over manually generated m
ontages because computer analysis can be performed on the computer-bas
ed montages. In addition, background variations and discontinuities be
tween individual photographs observed in manually generated montages a
re reduced greatly in computer-generated montages. Most important, the
computer-generated montages are better aligned than the manually gene
rated photomontages. Conclusions. This method of reconstructing a wide
-angle composite retinal fundus image from a set of adjacent small- an
d wide-angle fundus slides is a new tool for creating montages as larg
e as 100 degrees field of view. The computer-generated montages may be
used for documenting and quantifying retinal findings, This can great
ly assist studies of retinal manifestations of diseases, such as gyrat
e atrophy, retinitis pigmentosa, sickle cell disease, and acquired imm
une deficiency syndrome.