Abstract
Extracting the reconstructed decoration parts of a Buddha statue contributes to the analysis in archaeology and culture. A segmented decoration part with good precision is required for reconstructing the solid model of a decoration. This paper presents a novel segmentation algorithm of a point cloud for decoration points with a simple calculation based on divided rectangular surfaces. In our method, the decoration feature of the standard form according to a template is used to detect an exact decoration region. A differentiation graph indicating a variation of tangent vectors is introduced to extract the feature of a decoration region in a two-dimensional domain. A decoration region is detected according to higher correlation values with the decoration feature of the standard form. The proposed method is tested on the arm parts of a Mongolian Buddha statue and the effectiveness of our algorithm is confirmed and evaluated. The decoration points are segmented precisely, and the decoration and background parts are separated. Moreover, holes in the separated background are filled using a B-spline surface fitting technique for generating a complete surface model of the decoration parts and background.
