Publication

An effective method for generating blending meshes by leveraging geodesic curves on triangular meshes. Blending regions are automatically initialized using either minimum-distance points or intersection curves depending on whether the input meshes intersect. Each region is parameterized via geodesic linear interpolation and a reparameterization strategy establishes optimal curve correspondences, ensuring smooth, twist-free connections.

A framework for generating 3D text directly on triangular meshes while faithfully preserving local surface geometry. TrueType Font outlines are mapped onto the mesh along a user-specified placement curve and reconstructed using geodesic Bézier curves. Two mapping strategies based on local tangent frames and straightest geodesics ensure natural font alignment, enabling embossing, engraving, or independent 3D text meshes without self-intersections.

A deep learning method for automatic anthropometric landmark extraction from 3D human scans using a coarse-to-fine strategy. A novel geodesic distance-based heatmap effectively captures point distributions on 3D shapes even with pose variations. The method achieves the lowest average detection error on the SHREC'14 dataset with a maximum error reduction of 76.14%, and requires no predefined templates or handcrafted features.

A novel approach for freely slicing triangle meshes using a freeform curve without discretizing it into line segments. The user draws a stroke defining the cutting trajectory, which is extended into a ruled surface. Intersections are computed efficiently by decomposing the Line–Surface Intersection problem into Plane–Curve and Line–Line subproblems, enabling arbitrary straight and curved slicing while preserving efficiency and accuracy.

An effective neurofeedback system for treating neuropsychiatric disorders in children through EEG signal processing and auto-thresholding. The system determines optimal threshold values using numerical optimization to achieve a target success rate, generating success/failure feedback for reward and inhibit EEG frequencies. Training content is implemented as serious video games, making neurofeedback training engaging and clinically effective without relying solely on therapist experience.

A simple and effective method for interactively segmenting feature regions in a triangular mesh with a single mouse click. A concavity-aware harmonic field is computed on a geodesic disc candidate region, from which an initial isoline is selected. Feature points on the isoline drive anisotropic geodesics that form the final smooth, locally shortest segmentation boundary—consistent regardless of where within the ROI the user clicks.

A simple and effective method for transferring shape details between mesh regions using a Poisson equation. The mean curvature field of a source region, computed via the Laplace–Beltrami operator, is defined as its geometric detail. Source and target regions are co-parameterized on a 2D domain, the curvature field is interpolated by correspondence, and the Poisson equation is solved to replace the target's curvature field with that of the source.

An effective method for constructing C¹ piecewise cubic curves on a triangular mesh by interpolating given mesh points. Hermite interpolation is extended to meshes by defining geodesic vectors as straightest geodesics and devising geodesic operations for points and vectors. The resulting geodesic Hermite spline lies on the mesh surface and supports applications including mesh cutting and segmentation, with control parameters for curve sharpness.

An effective technique for filling holes in triangular meshes by blending two patches based on shape difference. An initial patch is generated by iterative refinement, then enhanced into a target patch by minimizing principal curvature variation. A Laplacian-smoothed source patch is generated, and per-vertex blending weights are determined from the shape difference in terms of Euclidean distance and curvature variation, enabling faithful recovery of salient geometric features.

An algorithm for computing the two-sided Hausdorff distance between a triangle mesh and a quad mesh, guaranteed within a given error bound down to machine precision. The algorithm extends a prior one-sided method by completing the opposite direction (quad-to-triangle) via splitting each quad into two triangles and introducing new matching methods. Experimental results demonstrate interactive-speed computation even for near-zero Hausdorff distance cases.

A robust and efficient Minkowski sum algorithm for planar freeform geometric models. Input boundary curves are approximated by G¹-biarc splines, enabling exact convolution arc computation. Interior disk culling trims redundant convolution arcs using a grid-based structure of interior disks, dramatically reducing the number of candidate arcs and improving overall efficiency compared to prior methods.

A method for constructing a parametric blending surface that smoothly connects two triangular meshes. The user selects a subregion on each mesh; local parameterizations are found via geodesic polar coordinates and base surfaces are constructed from two boundary curves on the parametric domain. The two base surfaces are then smoothly blended, with shape controllable via parameters or direct surface-point manipulation.

An algorithm that computes the one-sided Hausdorff distance from a triangle mesh to a quad mesh with high robustness and low memory consumption. The algorithm avoids storing combinatorial pairs by using point projection via uniform grid, while achieving much higher precision than previous methods. Experimental results show that even near-zero Hausdorff distance cases are handled at interactive speed with manageable memory usage.

We propose a new tangible visualization table for intuitive and effective visualization of terrain data transferred from a remote server in real time. The shape display approximating the height field of remote terrain data is generated by linear actuators, and the corresponding texture image is projected onto the shape display. To minimize projection distortions, we present a sophisticated technique for projection mapping. Gesture-based user interfaces facilitate intuitive manipulations of visualization results. We demonstrate the effectiveness of our system by displaying and manipulating various terrain data using gesture-based interfaces.

We present a system for creating realistic 3D face models by blending multi-scale facial details. Our approach allows artists to intuitively design face models by parameterizing the problem in a way that is understandable and controllable. The system synthesizes faces by combining coarse shape with fine-scale details such as wrinkles and pores, leveraging multiscale displacement representations. This makes it possible to transfer and blend detailed facial features across different face models, enabling flexible and artist-friendly face design workflows for games and animated films.

We present a method for transferring deformation weights of a human character to three-dimensional (3D) scanned clothes. First, clothing vertices are projected onto a character skin. Their deformation weights are determined from the barycentric coordinates of the projection points. For more complicated parts, such as shoulders and armpits, continuously moving planes are constructed and employed as projection reference planes. Clothing vertices on a plane are projected onto the intersection curve of the plane with a character skin to achieve a smooth weight transfer. The proposed method produces an initial deformation for physically based clothing simulations.

We facilitate the editing of hierarchical B-spline curves at multiple resolutions by expressing a displacement function at each level in rotation minimizing frames (RMFs) on the curve at the next lower level. When the curve is edited at a particular level, RMFs at all the higher levels are updated, and the control points of the displacement function at each higher level are obtained from these new RMFs. This transfers details created at the current level to higher levels. Our method presents a hundred-fold faster way to reflect editing results compared to the traditional approach using Frenet frames.

An efficient surface area evaluation method is introduced by using smooth surface reconstruction for three-dimensional scanned human body data. Surface area evaluations for various body parts are compared with the results from the traditional alginate-based method, and quite high similarity between the two results is obtained. We expect that our surface area evaluation method can be an alternative to measuring surface area by the cumbersome alginate method.

We present a simple and effective method for constructing a gallery that consists of weathering effect elements called time-dependent appearance manifolds (TDAMs). Since TDAMs are computed from sample video clips showing dynamic weathering phenomena, they represent very smooth changes in the appearance of weathered pixels over time. Once a gallery with a variety of weathering effects is prepared, users can interactively choose and apply the predefined effects onto the surface of 3D graphic models. This video-based weathering method supports predictable, intuitive, and natural effects, allowing users to produce photorealistic augmented videos.

PN (point-normal) triangles are cubic Bézier triangles which meet at their edges to surface a triangular mesh, but this only achieves G0 continuity. We define blending regions that span the edges shared by adjacent pairs of triangular domains and blend the corresponding Bézier triangles using a univariate blending function formulated in terms of barycentric coordinates. This produces G2 continuity across boundaries while preserving G1 continuity at vertices. The sharpness of the blends can be controlled locally by varying the extent of these blending regions.

We model developable surfaces by wrapping a planar figure around cones and cylinders. Complicated developables can be constructed by successive mappings using cones and cylinders of different sizes and shapes. We also propose an intuitive control mechanism, which allows a user to select an arbitrary point on the planar figure and move it to a new position. Numerical techniques are then used to find a cone or cylinder that produces the required mapping. Several examples demonstrate the effectiveness of our technique.

We present a new method for generating sweep surfaces using bivariate B-spline motion. The sweep surface is defined as the trace of a single point under bivariate B-spline motion. Direct manipulation of the sweep surface is achieved by controlling its motion components while producing various editing effects. We demonstrate the effectiveness of our technique by modeling and deforming various three-dimensional shapes.

We present an efficient framework for the shape approximations of spherical solid objects based on trivariate B-splines. We construct a smooth correspondence between a given object and a unit solid cube by computing their harmonic mapping, and utilize the mapping to approximate the given object with B-splines in a coarse-to-fine manner. Our framework provides user-controllability of shape approximations, based on the control of the boundary condition of the harmonic parameterization and the level of B-spline fitting.

We present an interactive-speed algorithm for computing the Hausdorff Distance (HD) between two freeform geometric models represented with NURBS surfaces. The algorithm is based on an effective technique for matching a surface patch from one model to the corresponding nearby surface patch on the other model, employing a bounding volume hierarchy (BVH) of Coons patches. By comparing local HD upper bounds against a global HD lower bound, we eliminate the majority of redundant surface patches from further consideration.

We present a bounding volume hierarchy (BVH) for freeform NURBS surfaces, using Coons patches and bilinear surfaces as the bounding volumes. The hierarchy provides an efficient approximation of NURBS surfaces within a given error bound. This BVH facilitates a wide range of geometric operations including collision detection, minimum distance computation, and Hausdorff distance computation for complex freeform models at interactive speeds.

We construct surfaces with darts, creases, and corners by blending different types of local geometries. We also render these surfaces efficiently using programmable graphics hardware. Points on the blending surface are evaluated using simplified computation which can easily be performed on a graphics processing unit. Results show an eighteen-fold to twenty-fold increase in rendering speed over a CPU version. We also demonstrate how these surfaces can be trimmed using textures.

We present a real-time algorithm for computing the precise Hausdorff Distance between two planar freeform curves. The algorithm approximates each curve with a sequence of G1 biarcs within an arbitrary error bound. The distance map for the union of arcs is rendered efficiently to the graphics hardware depth buffer. By sampling the distance map along the other curve, we estimate a lower bound for the HD and eliminate many redundant curve segments, reducing computation to considerably smaller subproblems solved with a multivariate equation solver.

We present a layered geometric approach to the skinning of character animation. On top of a global shape deformation, B-spline surface patches are attached to various body parts for local control. The patches are directly manipulated to generate example local shapes at important poses. During animation, B-spline control points interpolate a set of example patches kinematically while also moving dynamically by elastic simulation, generating natural local details including muscular effects at hundreds to thousands of frames per second.

We present a simple and efficient technique for plausible elastic deformation of three-dimensional objects. An elastic sweep surface is constructed by interpolating key cross sections with positions, orientations, and boundary shapes determined by physical simulation of simple mass-spring systems. The deformable parts of an object are approximated by elastic sweep surfaces, and the vertices of the deformable parts are bound to nearby sweep surfaces. As an external force is applied, the corresponding parts change their shapes elastically in real time.

We propose a sweep-based approach to the freeform deformation of three-dimensional objects. Instead of using a volume enclosing the whole object, we approximate only its deformable parts using sweep surfaces. The vertices on the object boundary are bound to the sweep surfaces and follow their deformation. Several sweep surfaces can be organized into a hierarchy so that they interact with each other in a controlled manner, supporting intuitively plausible shape deformation of objects of arbitrary topology with multiple control handles, along with volume preservation.

We present a sweep-based approach to human body modeling and deformation. A rigid 3D human model, given as a polygonal mesh, is approximated with control sweep surfaces. The vertices on the mesh are bound to nearby sweep surfaces and follow the deformation as the model bends and twists its arms, legs, spine and neck. Anatomical features including bone-protrusion, muscle-bulge, and skin-folding are supported by a GPU-based collision detection procedure. The volumes of arms, legs, and torso are kept constant using a volume integral formula for sweep surfaces.

We present a new surface representation scheme based on a manifold structure and displacement functions. Given a geometric model represented as a point cloud, we construct a domain manifold which is a smooth surface blended from simple local patches. The original points are projected onto the local patches and their displacements are adjusted so that the final displaced surface approximates the given point data with high precision. Applications include multi-resolution representations and skeleton-driven shape deformation.

We present a new approach to realistic hand modeling and deformation with real-time performance. The underlying shape of a human hand is modeled by sweeps following a simplified skeleton. The resulting swept surfaces are blended, and an auxiliary palm-control surface handles bulges at various poses. Palm lines are modeled as valleys in a displacement map and activated by appropriate joint movements. Self-intersections and collisions are detected using geometric proximity tests.