Topologically-robust 3D shape matching based on diffusion geometry and seed growing

Abstract

Abstract3D Shape matching is an important problem in computer vision. One of the major difficulties in finding dense correspondences between 3D shapes is related to the topological discrepancies that often arise due to complex kinematic motions. In this paper we propose a shape matching method that is robust to such changes in topology. The algorithm starts from a sparse set of seed matches and outputs dense matching. We propose to use a shape descriptor based on properties of the heat-kernel and which provides an intrinsic scale-space representation. This descriptor incorporates (i) heat-flow from already matched points and (ii) self diffusion. At small scales the descriptor behaves locally and hence it is robust to global changes in topology. Therefore, it can be used to build a vertex-to-vertex matching score conditioned by an initial correspondence set. This score is then used to iteratively add new correspondences based on a novel seed-growing method that iteratively propagates the seed correspondences to nearby vertices. The matching is farther densified via an EM-like method that explores the congruency between the two shape embeddings. Our method is compared with two recently proposed algorithms and we show that we can deal with substantial topological differences between the two shapes.

Finding Matches in a Haystack: A Max-Pooling Strategy for Graph Matching in the Presence of Outliers

“…Graph matching is widely used in computer vision problems such as finding feature correspondences [7,10,25], shape matching [22,30], object recognition [2,11], and video analysis [4]. Since graph matching is mathematically expressed as a quadratic assignment problem [8], which is NP-hard, most research has long focused on developing accurate and efficient approximate algorithms [14,27].…”

A Generic Deformation Model for Dense Non-rigid Surface Registration: A Higher-Order MRF-Based Approach

“…Recent works proposed searching among multiple parametrizations and/or combining multiple matching cues (e.g., texture or Gaussian curvature) to improve matching accuracy [10,11,25,29,31,32]. Another popular approach is to embed the surface into an Euclidean space such that the Euclidean distance approximates the intrinsic properties of the surface [4,12,14,19]. Nevertheless, when it comes to dense and anisometric deformations, the accuracy of the above methods would unavoidably deteriorate due to the isometric assumption.…”

Dense Image Matching With Two Steps of Expansion

“…In this paper, we introduce a method based on two steps of expansion and inspired by the seed-and-grow method (Adams and Bischof, 1994;Sharma et al, 2011;Wang et al, 2013;Yin et al, 2014). It determines the corresponding information on all the feature points of an image through two steps of expansion over a small number of seed and feature points.…”