Fast freeform hybrid reconstruction with manual mesh segmentation

In this paper, we presented a method for fitting large B-spline topological surfaces on freeform polygon mesh generated from cloud data of objects. The mesh has been manually segmented and large surfaces are fitted on segments in a hybrid approach, i.e., combination of geometric subdivision and non-uniform rational B-spline (NURBS) interpolation which is an emerging research space. An interpolation method has been proposed to parameterize dense cloud data of any complexity level with capability of handling occluded regions. All junctions are treated with trimming of NURBS surfaces with C ⁰ and C ¹ continuities between adjacent patches. This scheme amalgamated process knowledge of reconstruction on segmented and subdivided point cloud data, various NURBS geometry options and junction treatment resulting in faster high-quality reconstruction. Apart from manual segmentation, almost the entire process is automatic which generated superior quality surface models. Pawn, Stanford Bunny, and human head clouds with occluded surface zones are used for tryout and resulting shapes are recorded in initial graphics exchange specification (IGES) files.