Fitting smooth surfaces to dense polygon meshes

摘要: Recent progress in acquiring shape from range data permits the acquisition of seamless million-polygon meshes physical models. While dense polygon are an adequate representation for some applications, many users prefer smooth surface representations reasons compactness, control, manufacturability, or appearance. In this thesis, we present algorithms and end-to-end software system converting irregular arbitrary topology into tensor product B-spline patches with accompanying displacement maps. This choice yields a coarse but efficient model suitable interactive modification animation fine more expensive rendering. The first step our process consists interactively painting patch boundaries onto polygonal surface. placement is considered part creative not amenable to automation. We techniques representing, creating editing curves on surfaces. The second finding gridded resampling each bounded section mesh. Our algorithm lays grid springs across mesh, then iterates between relaxing subdividing it. provides parameterization mesh section, which initially unparameterized. automatic, efficient, robust, even complex surfaces. Prior have lacked one these properties, making them unusable meshes. strategy also user flexible method design parameterizations--an ability that previous literature approximation does address. The third final fitting hybrid surfaces maps re-sampling. The map image error fitted spring grid. Since just images facilitates use processing operators manipulating geometric detail object. Our steps fast enough million under 10 minutes--important system.