A visual 3D retinal model could facilitate the process of
disease diagnosis. We have studied 3D retinal surface reconstruction by
using the affine camera due to the narrow field-of-view and the small
depth variation in each field. We have proposed the affine bundle
adjustment algorithm, which refines visual reconstruction and camera
parameters simultaneously. Moreover, we have developed a new 3D
reconstruction algorithm that incorporates the retina’s geometric
constraint and the radial distortion removal into the optimization
process of bundle adjustment to further improve the quality of 3D
retinal curvature estimation.
Examples of 3D retinal curvature estimation.
T. Chanwimaluang and G. Fan, “Constrained Affine
Bundle Adjustment for 3D Retinal Surface Reconstruction”, in the
IEEE International Workshop on Beyond Multiview Geometry: Robust
Estimation and Organization of Shapes from Multiple Cues (in
conjunction with CVPR07), Minneapolis, Minnesota, June 2007.
T. Chanwimaluang and G. Fan, “Affine Camera for 3D
Retinal Surface Reconstruction”, in Proc. International Symposium on
Visual Computing, Lake Tahoe, NV, Nov. 11-13, 2006, also in LNCS, Vol.
4292, Editors: G. Bebis, Springer, 2006.