Catálogo de publicaciones - libros

In this paper a simple approach is proposed to render crack-like patterns and animate cracking propagations on surfaces of 3D objects. With some controllable parameters, generating cracks on an object with variant visual effects such as bumping or carving is flexible. Besides, a proposed script system could generate various cracking animation on 3D objects.

A new adaptive tessellation method for general Catmull-Clark subdivision surfaces is presented. Development of the new method is based on the observation that optimum adaptive tessellation for rendering purpose is a recursive error evaluation and globalization process. The adaptive tessellation process is done by generating an inscribing polyhedron to approximate the limit surface for each individual patch. The inscribing polyhedron is generated through an adaptive subdivision on the patch’s parameter space driven by a recursive error evaluation process. This approach generates less faces in the resulting approximating mesh while meeting the given precision requirement. The crack problem is avoided through globalization of new vertices generated in the adaptive subdivision process of the parameter space. No crack-detection or crack-elimination is needed in the adaptive tessellation process. Therefore, no mesh element splitting to eliminate cracks is necessary. The new adaptive tessellation method can precisely measure the error for every point of the limit surface. Hence, it has complete control of the accuracy of the tessellation result.

It has been technically challenging to effectively model and simulate elastic deformation of spline-based, thin-shell objects of complicated topology. This is primarily because traditional FEM are typically defined upon planar domain, therefore incapable of constructing complicated, smooth spline surfaces without patching/trimming. Moreover, at least C ^1 continuity is required for the convergence of FEM solutions in thin-shell simulation. In this paper, we develop a new paradigm which elegantly integrates the thin-shell FEM simulation with geometric design of arbitrary manifold spline surfaces. In particular, we systematically extend the triangular B -spline FEM from planar domains to manifold domains. The deformation is represented as a linear combination of triangular B -splines over shell surfaces, then the dynamics of thin-shell simulation is computed through the minimization of Kirchhoff-Love energy. The advantages given by our paradigm are: FEM simulation of arbitrary manifold without meshing and data conversion, and the integrated approach for geometric design and dynamic simulation/analysis. Our system also provides a level-of-detail sculpting tool to manipulate the overall shapes of thin-shell surfaces for effective design. The proposed framework has been evaluated on a set of spline models of various topologies, and the results demonstrate its efficacy in physics-based modeling, interactive shape design and finite-element simulation.

This paper proposes a geometry-based technique to control the target shape and the motion of clouds in computer animation so that the synthetic appearance of the clouds resembles a specified three-dimensional shape. The technology for automatically generating this special effect has been desired by the movie industry for many years. Our method is based on ellipsoid decomposition. Firstly, ellipsoids are employed to approximate a given mesh model which indicates the target shape of cloud animation. After that, the target object is represented in a blobby implicit surface using ellipsoidal blobs. Finally, two geometry-based schemes are introduced to generate the cloud animations with target shape controlled in two different ways: aggregated from several pieces of clouds or diffused from one piece of cloud.

One of the main question addressed by paleoanthropologists is the recovery of plausible motions for extinct species whose knowledge is generally limited to incomplete bones and skeletons. Calculating locomotion for extinct species is mainly based on the direct application of captured trajectories to numerically reconstructed skeletons. The gait is judged realistic compared to another if it minimizes energy and preserves balance. In computer animation, adapting a motion to a skeleton is addressed by so-called motion retargeting techniques. The goal is then to ensure that the resulting motion is close to the initial one while dealing with new bones’ dimensions. In this paper, we adapt methods used in computer animation in order to solve such a problem. This approach is based on a two-steps framework: first, a reference motion of the feet is warped in order to optimize a set of biomechanical general laws, such as energy and Jerk minimization. Second, the remaining degrees of freedom (denoted DOF) are calculated thanks to inverse kinematics (denoted IK). This method is applied on a small woman, a tall man, a chimpanzee and Lucy ( Australopithecus afarensis ).

We present a technique for simulating variations in appearance of aerial images at different sun angles. The input to the algorithm is a single aerial image with information on the orientation of surfaces in the image. The location, date, and time at which the photograph was taken are also needed by the algorithm. The appearance of the aerial image at a new sun angle is synthesized by compensating for the direct sunlight component in the original image, and then relighting the image with sunlight from the new sun position. Techniques to remove cast shadows are also described. We demonstrate our results with images, and animations of changing lighting in aerial photographs as the sun follows a trajectory across the sky.

We introduce a new compression algorithm for complex animated meshes of constant connectivity based on the local principal component analysis. The algorithm segments the animated mesh into segments using a region growing algorithm and transforms the original vertex coordinates into the local coordinate frame of their segment. This transformation leads to a strong clustering behavior of vertices and makes each region invariant to any deformation over time. Then each segment is efficiently encoded using the principal component analysis. The set of basis vectors and coefficients corresponding to each segment are quantized and entropy encoded. Experimental results show that our algorithm yields a significant improvement upon some current coders.

We propose a framework to reconstruct human motion based on monocular camera video and motion database. In this framework, we use silhouettes for rough motion estimation based on a set of discriminative features and search motion database to find out the exact motion clips that meet with the video content. We model motion as a first-order Markov process. The transition probabilities between motion clips are preprocessed with consideration of the continuousness and smoothness of human motion. To eliminate the discontinuities between motion clips, we also adopt a seamless motion stitch method using multiresolution analysis technique. We verify the effectiveness of our method by reconstructing trampoline sports video as an example. The reconstruction results are visually comparable to those motions obtained by a commercial motion capture system in the premise that similar motions are included in the motion database.

In this work, we address visualization of spatio-temporal data for military application. Four different visualization prototypes have been developed to track the movement of military entities across a land surface over time; three more have been developed to track the occurrences of numerous war events. We have implemented the prototypes in a software system with a novel clock face GUI. Usability tests have been carried out and confirmed the effectiveness of the solution.

In this paper, a 3D city model generation method is described. At first, the ground images around the area of interest are acquired using a camera mounted on a GPS device and digital compass, which provide the initial inaccurate pose of each image. An adjacency graph that spatially represents the adjacency between images or buildings is built to handle efficiently a huge number of unordered image sequences. A set of images with views of the same building is automatically grouped, and an optimization algorithm based on SFM corrects their poses. Finally, a method of global pose estimation is outlined that can register 3D isolated building models in a global coordinate system. We validate our approach with a set of experiments on some urban sites.