Estimating principal deformable freeform features by curvature analysis

A suitable estimation of the initial positions, orientations and dimensions of a template referring to a freeform feature in an existing geometry is a prerequisite for template fitting and template based shape reconstruction and manipulation. In this paper, a curvature analysis based method is proposed to estimate the principal freeform feature in a specified region of interest for template fitting. Based on the minimal, maximal, mean or Gaussian curvature computing, the geometric information is transferred to the curvature domain. Using a variant of Laplacian smoothing methods, the high frequency noises and interferences in the curvature domain are suppressed and the principal feature is addressed. By feeding back the extracted feature information to the geometric shape, the geometry of the template is estimated based on the feature analysis. With several case studies, the effectiveness of the proposed method is verified. Different effects caused by different curvature computing methods are discussed as well