- مبلغ: ۸۶,۰۰۰ تومان
- مبلغ: ۹۱,۰۰۰ تومان
In an accurate free-form deformation of a polygonal object, only the linear geometry, e.g., triangles or planar polygons, is deformed as triangular Bézier patches or trimmed tensor product Bézier patches; the related normal field is not considered. Thus, the geometry appearance and shading of the deformed object are typically not smooth. In this paper, both the linear geometry and normal of a polygonal object are simultaneously considered in the framework of accurate free-form deformation. First, each triangle and its normal field are deformed as two cubic triangular Bézier patches. Then, the curved geometry corresponding to the deformed triangles is locally adjusted to tone the smoothness of the geometry appearance according to the deformed normal field. The deformed normal field is adjusted accordingly. As a result, a smooth free-form deformation with visually plausible smooth geometry and shading is obtained. Furthermore, the sharp features in the polygonal object can be preserved. Because the curved geometry and normal field adjustments are local operations, all of the above computations can be performed in parallel on a GPU. The experimental results show that the method can deform a complex polygonal object as a smooth object in real time while preserving sharp features.
7. Conclusion and future work
In this paper, we proposed a GPU-based smooth FFD with sharp features awareness that addresses the unsmoothness of the normal field and the geometry artifact problems in the framework of accurate FFD. The algorithm can produce a high-quality deformation result. It is a highly parallelizable GPU algorithm and is able to deform a relatively large-scale model in real time. The algorithm is intuitive and can be implemented easily. It can handle relatively coarse meshes and generate smooth deformation results. The approach can still be improved in several aspects. First, the uniform tessellation of the cubic triangular Bézier patches will generate many unnecessary small triangles. An efficient adaptive tessellation algorithm via GPGPU is an alternative to our method. Second, the approximation error of the smooth FFD for polygonal object is worth to be analyzed in theory. A feasible error bound is useful to guide the discretization of a smooth object, which is essential for generating high-quality deformation result.