Computer Graphics. Shape-Similarity Search of 3D Models by Using Moment Envelopes.

Effective storage and retrieval of three-dimensional (3D) models has become an important issue. In this paper, we proposes a method for shape-similarity search of 3D polygonal-mesh models. As the feature vector, we employed a set of envelopes of statistical distributions of mass concentrated on the surface. The envelopes are parameterized along principal axes of model inertia. As a distance measure among, we experimented with theEuclidean norm as well as a distance measure based on elastic matching by using the dynamic programming technique. Experiments showed that our approach could identify models having similar shapes fainy well. As a shot-coming feature vectors, the query results did not necessarily reflect subjective measures used by human beings.