| Abstract |
In this paper, we propose a new method for
describing 3D-shape in order to perform similarity search for
polygonal mesh models. The approach is based on
characterization of spatial properties of 3D-objects by suitable
feature vectors, i.e., the goal is to define 3D-shape descriptors
in such a way that similar objects are represented by “close”
points in the feature vector space. We present a descriptor
which is invariant with respect to translation, rotation,
scaling, and reflection and robust with respect to level-ofdetail.
A coarse voxelization of a 3D-model is used as the input
for the 3D Discrete Fourier Transform (3D DFT), while the
absolute values of obtained (complex) coefficients are
considered as components of the feature vector. Multiple
levels of abstraction of the feature are embedded by the
applied transform. The performance of the proposed method
is compared to some previous approaches by means of
precision/recall tests. Generally, results show that the new
approach introduces improvements in the 3D-model retrieval
process. |