As 3D applications ranging from medical imaging to industrial design continue to grow, so does the importance of developing robust 3D shape retrieval systems. A key issue in developing an accurate shape retrieval algorithm is to design an efficient shape descriptor for which an index can be built, and similarity queries can be answered efficiently. While the overwhelming majority of prior work on 3D shape
analysis has concentrated primarily on rigid shape retrieval, many real objects such as articulated motions of humans are nonrigid and hence can exhibit a variety of poses and deformations. In this thesis, we present novel spectral geometric methods for analyzing and distinguishing between deformable 3D shapes. First, we comprehensively review recent shape descriptors based on the spectral decomposition of the Laplace-Beltrami operator, which provides a rich set of eigenbases that are invariant to intrinsic isometries. Then we provide a general and flexible framework for the analysis and design of shape signatures from the spectral graph wavelet perspective. In a bid to capture the global and local geometry, we propose a multiresolution shape signature based on a cubic spline wavelet generating kernel. This signature delivers best-in-class shape retrieval performance. Second, we investigate the ambiguity modeling of codebook for the densely distributed low-level shape descriptors. Inspired
by the ability of spatial cues to improve discrimination between shapes, we also propose to adopt the isocontours of the second eigenfunction of the Laplace-Beltrami operator to perform surface partition, which can significantly ameliorate the retrieval performance of the time-scaled local descriptors. To further enhance the shape retrieval accuracy, we introduce an intrinsic spatial pyramid matching approach. Extensive experiments are carried out on two 3D shape benchmarks to assess the performance of the proposed spectral geometric approaches in comparison with state-of-the-art methods.