3-D digital surface recovery of the optic nerve head from stereo fundus images

Abstract

A novel algorithm for 3D digital mapping of curved surfaces from a 2D stereo image pair is developed. This approach for visualization of curved surface topography involves fusion of a stereo depth map with a linearly stretched intensity image of the curved surface. Prior to fusion of the depth map with the intensity image, a cubic B-spline interpolation technique is applied to smooth the sparse depth map obtained from the computed stereo disparity map. The quantitative representation of the optic nerve head surface topography following this algorithm provides a technique for a possibly more reproducible parametric evaluation of the optic nerve head than just qualitative stereoscopic viewing of the fundus.<>