Part‐type Segmentation of Articulated Voxel‐Shapes using the Junction Rule

Abstract

We present a part-type segmentation method for articulated voxel-shapes based on curve skeletons. Shapes are considered to consist of several simpler, intersecting shapes. Our method is based on the junction rule: the observation that two intersecting shapes generate an additional junction in their joined curve-skeleton near the place of intersection. For each curve-skeleton point, we construct a piecewise-geodesic loop on the shape surface. Starting from the junctions, we search along the curve skeleton for points whose associated loops make for suitable part cuts. The segmentations are robust to noise and discretization artifacts, because the curve skeletonization incorporates a single user-parameter to filter spurious curve-skeleton branches. Furthermore, segment borders are smooth and minimally twisting by construction. We demonstrate our method on several real-world examples and compare it to existing part-type segmentation methods.

Improved Part-Based Segmentation of Voxel Shapes by Skeleton Cut Spaces

“…We get very similar results, but nd more ne-grained segments than [38] -see nger and ear details of the animal models, pig tail, dragon spikes, and microscope lens. Segment borders are smooth and locally orthogonal to the shape's symmetry axis, i.e., similar to how a human would cut the shape at the respective places (Sec.…”

“…We rst compare our results with [38], the best voxel-based PBS method that we are aware of which also uses medial descriptors for segmentation. We get very similar results, but nd more ne-grained segments than [38] -see nger and ear details of the animal models, pig tail, dragon spikes, and microscope lens.…”

“…The considered methods are [4, 24-27, 37, 38, 50]. Here, Reniers et al (1) denote [37], and Reniers et al (2) denote [38]. These methods span from voxel-based to mesh-based, and use various segmentation heuristics (skeleton, curvature, salience, and topology-based).…”

“…We argue that our method creates equally or, in some cases, more plausible PBSs. Since both our method and [38] use medial descriptors, computed by the same underlying method [40], a relevant question is how the two methods di er. We use (a) medial surfaces, while [38] uses curve skeletons; and (b) we nd segment borders by analyzing all possible cuts, while [38] places such borders around the curve-skeleton branch junctions.…”

“…However, curve skeletons can contain many spurious junctions which change widely when the shape is slightly perturbed. Reniers et al alleviate this by heuristics that shift cut-points along the curve skeleton to optimize for cut stability and planarity [38]. Yet, this method cannot segment shapes of large geometric, but little topological, variability, like a pawn chess piece: its curve skeleton has no junction points, so [38] cannot separate the pawn's head, body, and base, although these have di erent thicknesses.…”