Nilpotent Approximations of Sub-Riemannian Distances for Fast Perceptual Grouping of Blood Vessels in 2D and 3D

Abstract

We propose an efficient approach for the grouping of local orientations (points on vessels) via nilpotent approximations of sub-Riemannian distances in the 2D and 3D roto-translation groups SE(2) and SE(3). In our distance approximations we consider homogeneous norms on nilpotent groups that locally approximate SE(n), and which are obtained via the exponential and logarithmic map on SE(n). In a qualitative validation we show that the norms provide accurate approximations of the true sub-Riemannian distances, and we discuss their relations to the fundamental solution of the sub-Laplacian on SE(n). The quantitative experiments further confirm the accuracy of the approximations. Quantitative results are obtained by evaluating perceptual grouping performance of retinal blood vessels in 2D images and curves in challenging 3D synthetic volumes. The results show that (1) sub-Riemannian geometry is essential in achieving top performance and (2) grouping via the fast analytic approximations performs almost equally, or better, than data-adaptive fast marching approaches on R n and SE(n).

Hamiltonian Fast Marching: A Numerical Solver for Anisotropic and Non-Holonomic Eikonal PDEs

“…The weights α i = α i (p) ≥ 0 and offsetsė i =ė i (p) ∈ Z d , where 1 ≤ i ≤ I, depend on the current point p ∈ X. Our objective is numerically solve the eikonal equation (3). For that purpose, we rely on the fast marching algorithm to compute U :…”

“…In each case, we provide the generic expression of the primal metric, as well as the dual metric which is defined by (4) and appears in the eikonal equation (3) and the geodesic backtracking ODE (5).…”

“…Assume now that several endpoints (x * i , θ * i ) i∈I are present in the image, instead of a single one. Then the visual system infers a curve joining the source point (x * , θ * ) to the closest endpoint, again with respect to the distance defined by the Reeds-Shepp model, a process called perceptual grouping [3]. The images created by Poggendorff put this physiological process in evidence by displaying a partially occluded straight line or circle.…”

“…Tubular structure segmentation in medical images is one of the main applications intended for the HFM library. Early versions, and related software of the first author, are used for that purpose in [12,52,15,22,14,33,3]. Vessel extraction in images of the retina is illustrated in Figure 24, using two popular models involving respectively radius and orientation lifting.…”

Anisotropic tubular minimal path model with fast marching front freezing scheme

“…Based on the Eikonal equation framework, the curvature-penalized minimal path models [23,24,25] are established over an orientation-lifted space such that the orientation dimension can be used to represent the curve tangents.…”