A Skeleton Family Generator via Physics-Based Deformable Models

Abstract

This paper presents a novel approach for object skeleton family extraction. The introduced technique utilizes a 2-D physics-based deformable model that parameterizes the objects shape. Deformation equations are solved exploiting modal analysis, and proportional to model physical characteristics, a different skeleton is produced every time, generating, in this way, a family of skeletons. The theoretical properties and the experiments presented demonstrate that obtained skeletons match to hand-labeled skeletons provided by human subjects, even in the presence of significant noise and shape variations, cuts and tears, and have the same topology as the original skeletons. In particular, the proposed approach produces no spurious branches without the need of any known skeleton pruning method.

A study on skeletonization of complex petroglyph shapes

“…A third category of approaches performs the medial axis transform to polygonal shapes in continuous space [12,20,21]. Additionally, there is a group of more recent skeletonization algorithms that utilize physics-based modeling of the shapes [14,22] for which we suggest a fourth category.…”

A similarity-based approach for shape classification using Aslan skeletons

“…Unfortunately, the skeleton descriptions of two visually similar shapes, when expressed as a graph, may be topologically different which poses an instability to which there are three possible solutions. The first is to post-process the skeletons by pruning problematic branches [3,5,28,47,61]. The second is to handle the problem during the matching step by utilizing error-tolerant matching algorithms [9,23,46,52,63].…”

“…Another appealing idea is presented in [61] where pruning process takes into account global context of a set of skeletons. In addition to these new pruning-based methods, the skeletonization method in [28] employs a physicsbased deformable model where the coarseness of extracted skeletons depends on a single parameter. Lastly, the approach in [35] is also of our interest since it offers a simplified graphical representation of a shape skeleton by analyzing ligature regions.…”

“…For illustration purposes, only one node is drawn here (adapted from [9]). limitation is that shapes with holes [4,26,28] can not be expressed easily in this representation because the construction of the global coordinate frame becomes complicated even though skeleton branches are extracted. Moreover, in the presence of large occlusion and severe missing parts, the shape center might move to a remote location, directly affecting the construction of the coordinate frame and accordingly the skeletal attributes.…”

“…In boundary-based representations, shapes are either expressed by a set of boundary points or by a set of boundary curves (or some features extracted from them) [10,11,22,31,32,55]. Skeleton-based approaches, on the other hand, aims at capturing a structural representation of shape by modeling the shape in terms of a set of axial curves [2,5,12,23,28,51,57,61,63]. As skeleton-based representations can describe the spatial relationships among shape parts, they implicitly introduce insensitivity to articulation and deformation.…”