Abstract:A novel technique is proposed to make realistic facial animation driven by motion capture data. First, one's facial expressions are recorded using Oxford Metrics' VICON 8 optical motion capture system, where 23 optical markers are attached on one's face. The captured 3D motion data are then processed to eliminate 3D global head motion by finding the pivot point for rotation. These processed 3D motion data can be directly applied to a head model. A two and half dimension facial model is used here for implementation because it combines good features from both 2D mesh and 3D model: simple, vivid and natural, when small-scale rotation is applied. An interpolation function is employed to calculate the offsets of other unconstrained mesh vertices; furthermore, a face is divided into several regions and force constraints are applied to limit the displacement of vertices on the mesh depending on the region to make the animation more natural. The system has an update rate of over 30 frames/second on a Pentium III 500MHz PC with an OpenGL acceleration card.

Abstract:In this paper, we introduce linkages as new drawing tools and show that this tool is complete, i.e., all diagrams that can be described constructively can be drawn with linkages. This class includes the constraint problems with distance constraints only. As an application, we show that the simplest constrained graph which is beyond the scope of Owen and Hoffmann's popular triangle decomposition me thods can be transformed to pure geometric constructive form. To solve the equations raised from linkage constructions, we propose a geometric method which is based on dynamic locus generation.

Abstract:Traditionally paint work is always done on a piece of paper, canvas, on a wall or on the surface of some material. Augmenting it into 3D virtual reality space is a breakthrough. By the inspiration of Jackson Pollock's “action painting”, human body could act as a paintbrush to drive the paintings by his/her body motion. Fusing the concepts of action painting and virtual reality together not only introduces new aesthetic perspectives on paint art, but also raises potential multi-disciplinary researches over the end product. In this paper, a framework is presented for building an action studio to experiment a new form of paint art that utilizes human body as paintbrushes to drive the painting process and then visualizes the paint work in a 3D canvas in real-time. Further, based on its inherent correlation characteristic of human behavior and cognition, the system could be extended for a joint research with other potential disciplines such as music and psychotherapy.

Abstract:A geometric compression algorithm is presented in this paper to save the geometry model storage and transmission time. This method decomposes the model into a series of layers in a way of region growing. These layers are then encoded effectively by using inter-layers coherence, so that the topology information of the model is compressed dramatically. Experimental results show that it takes only an average of 1.42 bits per triangle. Additionally, a non-linear geometry predictor is designed to compress the geometric information. Compared with the previous work, this algorithm is of linear complexity, and it can be implemented effectively.

Abstract:Shamir gave the (t,n)-Threshold Secret Sharing Scheme of cryptography based on Lagrange interpolation. In this paper, the authors introduce it into image sharing and give a new kind of algorithm for image sharing based on moving lines. The mathematical scheme of moving lines is presented in this paper. Hereby, when an image is shared, the authors consider more original images, compute with rational curves of high degree, compare the differences between Lagrange interpolation and moving lines on image sharing of many frames, and give the application fault of Lagrange interpolation with high degree in image sharing. Finally, some experimental results for image sharing based on moving lines are presented.

Abstract:In this paper, the authors use a tree structure based on fractal dimension to organize the 3D-terrain mesh model and realize real time continuous view-dependent level of detail rendering of terrain data. The total algorithm can be divided into two stages: preprocessing and rendering. In preprocessing, it builds an adaptive tree structure by considering the complexity of terrain area which was evaluated by the fractal dimension. Then it generates triangular mesh approximation by dynamically determining reserved data points according to the view related factors and performing a real time restricted quadtree triangulation. Experimental results show that this algorithm is simple and efficient, and supports the interactive real time dynamic rendering of terrain model.

Abstract:Though smoothness of surfaces is one of the main goals in free-form surface modeling, non-smooth effects, such as sharp features, are required in some occasion. Generation of them is investigated in this paper. A method based on the construction of mesh topology is presented to unify the generation of the sharp features such as crease, corner, dart and cone into the method of boundary process. Thereby creating special subdivision rules is avoided for different effects, and thus the sharp features are realized without updating subdivision schemes.

Abstract:Accurate and efficient vectorization are the foundation of engineering drawing understanding. The existing vectorization methods have three main shortages: (1) A line segment is fitted using a chain of medial axis points so that the computational complexity is high; (2) A lot of time-consuming searching and judging operations are needed to combine the separated segments caused by noise or intersections; (3) The capability of processing slant lines is much lower than that of processing vertical or horizontal lines. In this paper, the authors propose a seed segment based global line vectorization method, which vectorizes an entire line in one step so that the combining step is avoided. Furthermore, it processes the lines of any orientation in the same way. Both the performance evaluation and experimental results show that the performance of line vectorization is significantly improved using the proposed method.

Abstract:In this paper, the authors study the multidegree reduction of Bézier curves with arbitrary degree interpolation conditions of two endpoint. For the given endpoint interpolation conditions, a new approximation method of multidegree reduction is presented. Using Chebyshev polynomial approximation theory, the nearly best uniform approximation under the interpolation conditions of endpoints can be obtained. This algorithm is easy to implement and simple for error estimation. The approximation effects of the degree reduction curves are very good. Combined with subdivision algorithm, it can reach a higher rate of error convergence.

Abstract:Due to the visibility change and surface enlargement in the new view as viewpoint moves, 3D reprojection from just one reference image inevitably produces holes in the destination image. In tackling the problem, a common solution is to conduct image synthesis from multiple reference images. However, how to extract the information needed in the novel view from these images with plenty of redundance information becomes one of the key techniques in image-based modeling and rendering. In this paper, a new method of image synthesis from multiple (currently two) reference images is proposed. The method combines forward warping and inverse warping to fulfil the image composition task for a novel viewpoint. In the algorithm, vision constraint-epipolar constraint is employed to search the pixel from other reference images in filling the hole, so as to avoid depth comparison and re-mapping of the whole reference image. Besides, the searching process is accelerated by taking advantage of pixel ordering on pairs of epipolar lines and the information already in the destination image. Without increasing the complexity of previous warping method, this algorithm greatly reduces the computation required in image synthesis from multiple reference images.

Abstract:A novel constructive algorithm is presented to model free-form surfaces based on general skeletons in this paper. Each skeleton is first used to construct a distance field, then a smooth implicit surface is generated by using the implicit function blending and CSG (constructive solid geometry) representation technique. The final surface mesh is achieved by polygonizing the implicit function. By using the general skeletons as primitives, the authors break the limitation of traditional implicit method which uses points as primitives. The distance surface fits the shape of the original skeleton well, which makes users be able to interactively design complex surfaces intuitively. Furthermore, the introduced warping function greatly enriches the modeling means. The experimental results demonstrate that the prototype system based on this algorithm can conveniently and intuitively construct complex free-form surfaces.

Abstract:A novel algorithm for fitting surface reconstruction of unorganized data points is presented in this paper. Since the local geometric and topological properties can be clearly described with its neighboring points for a given point, the algorithm takes full advantage of this fact to generate a local triangular mesh based on Delaunay triangulation technique in two dimensions. The main contribution of the proposed algorithm is to progressively integrate those local triangular meshes into a resulting mesh by deleting the illegal triangles. In addition, holes and gaps can be detected automatically during reconstruction, and the algorithm works well for models with arbitrary topology. Experimental results show that this algorithm is effective, robust and can be easily implemented.

Abstract:In this paper, the authors present a physically based model for curtainreal-time animation, which not only significantly reduces the time of rendering, but also maintains visually appealing results. The curtain model is represented as an object composed of mass points connected by semi-rigid rods whose behavior is governed by non-rigid dynamics. The longitudinal (vertical) and latitudinal (horizontal) directions of the model are decoupled and processed separately, and later combined to generate the final curtain.

Abstract:To seek a smallest circle containing all the point of a given point set is an interesting problem in both practice and theory. In this paper, an algorithm of finding a smallest circle containing all the points given is presented. The time complexity of the algorithm is O(|lg(d/R)|*n), where R is the radius between the smallest circle, d is the smallest distance between the points of the set that are not on the circle and the circle.

Abstract:Based on the overlay technique, the authors propose a novel algorithm for interactive walkthrough of complex polygonal virtual environments with global specular reflection effects in this paper. This algorithm integrates the global specular reflection preprocessing with visibility preprocessing. The problem of computing global specular reflection is then converted to that of determining the potential visible polygon set (PVS) through each mirror. A mirror-PVS-mapping G is constructed, associating the potential incidence beams with their respective PVS. At the stage of walkthrough, the proposed algorithm can synthesize the global specular reflection in real time by G mapping while calculating the other intensity components of the environment.

Abstract:In this paper, the authors present a new multiresolution modeling algorithm, which generates the full set of simplified polygonal models by progressively removing insignificant edges from previous models. This algorithm first divides the polygonal mesh of the model into regions and interpolates each region with a variation implicit surface. When an edge is collapsed, its target point is obtained by sampling the relative implicit surface. A weighted control function is adopted to determine the order of edge collapse operation. The algorithm dynamically adjusts the weights of the control function in different simplification phases so that the less important edges for appearance preservation are always removed earlier than other edge candidates. Furthermore, independent sets are defined to avoid excessive local simplification. Experimental results show that the proposed algorithm can achieve satisfactory result.

Abstract:The interpolation of some given moving frames (origin position+orientations of three coordinate axes) of a moving object is an important problem in the fields of computer graphics, robotics, etc. In this paper, a new method is presented to interpolate and approximate the moving frames of moving object by employing B-splines. By parameterizing the rotation transformation part of the orthogonal moving frame with four Euler parameters, a simplified optimization function is derived. An iteration method is presented to solve the optimization function for approximating the rotation transformation part of the pseudo moving frame at any position of the motion trajectory. The error caused by interpolation and approximation is proven to be controllable. Since the computation of moving frame is only related with order 2 or 3 linear equations, the proposed approach can be implemented with high efficiency.

Abstract:Based on the analysis of the classifical 4 point linear interpolatory subdivision scheme introduced by Dyn, a functional nonlinear discrete subdivision scheme is presented. This scheme has the preserving convexity property, i.e., for any given convex discrete data, when some conditions are satisfied, the subdivision polygon curve produced in any step by this scheme is convex, so the limit curve is also convex. Some numerical examples show that the limit curves are fractal like when the smooth condition is not satisfied.

Abstract:Automatic parallel compilation provides important supports for parallelizing “legacy” programs and designing new parallel programs. It has been focused on for two decades. In recent years, the developments of automatic parallel compilation technique, including dependence analysis, program transformation, data distribution and redistribution, and scheduling, have gradually made it practically applicable. In this paper, the new development of automatic parallel compilation technique and the challenges of future research are presented.

Abstract:In this paper, the concept of action oriented belief updating is introduced based on the PMA (possible model approach). It is proved that a first order formula holds in a situation if and only if it belongs to the belief set of the situation provided the situation calculus system is informational complete. This result makes it possible to avoid the appeal to second induction axiom in the entailment of situation calculus and also provides a theoretical foundation for modeling agents with intention driver and the agent-oriented programming language AOPLID (agent-oriented programming language with intention driver).