Hierarchical feature subspace for structure-preserving deformation

This paper aims to propose a new framework for structure-preserving deformation, which is interactive, stable, and easy to use. The deformation is characterized by a nonlinear optimization problem that retains features and structures while allowing user-input external forces. The proposed framework consists of four major steps: feature analysis, ghost construction, energy optimization, and reconstruction. We employ a local structure-tensor-based feature analysis to acquire prior knowledge of the features and structures, which can be properly enforced throughout the deformation process. A ghost refers to a hierarchical feature subspace of the shape. It is constructed to control the original shape deformation in a user-transparent fashion, and speed up our algorithm while best accommodating the deformation. A feature-aware reconstruction is devised to rapidly map the deformation in the subspace back to the original space. Our user interaction is natural and friendly; far fewer point constraints and click-and-drag operations are necessary to achieve the flexible shape deformation goal. Various experiments are conducted to demonstrate the ease of manipulation and high performance of our method.