Dr. Hua Wei     

Assistant researcher. He got Ph.D in applied mathematics from Zhejiang University in 2002. He joined the CAD&CG State Key Lab in 2002. His main interests include real-time simulation and rendering, virtual reality and software engineering.

Contact: huawei@cad.zju.edu.cn

Current Interests

  • Virtual Reality
  • 3D Game Engine
  • Real-Time 3D Simulation & Rendering
  • Highly Efficient 3D Modeling
  • Software Engineering


Recent Technical Reports

Demo Video, about 6M, 58sec

Hierarchical Layered Assembled Billboard Packs for Fast Display of Large-Scale Forest with High Fidelity

Feb. 2004

The goal of this paper is fast display of large-scale forest with high fidelity by using fair storage. We propose a new representation for forest, Hierarchical Layered Assembled Billboard Packs (HLABPs). In the HLABPs, forest is represented by a scene graph, where each node representing a cluster of trees comprises a group of Layered Assembled Billboard Packs (LAB-Packs) arranged either in detail levels or in aspects. That is, a LAB-Pack represents an aspect of a cluster of trees in a detail level. It consists of a stack of parallel “billboards”, and each “billboard” is assembled by a layer of textured quadrilaterals with the same orientation but with the different size and depth. To construct HLABPs efficiently, we introduce a multiresolution structure of layered depth images, upon which the relevant assembled “billboards” in different detail levels can be generated avoiding sampling the original model time after time. Meanwhile, all textures are compressed by an occlusion-inclusive compression approach. Our rendering procedure traverses the HLABPs and sends the LAB-Packs whose view-dependent disparities are just less than a user-specific tolerance to the rendering pipeline. Moreover, we devise an elaborate blending scheme to mitigate the visual “popping” caused by detail level transition. Shown by the experiments, the rendering complexity of our approach is close to O(log(N)) (N is the number of trees), the complexity of storage is O(N) and the image quality is comparable to that of ray-tracing. The performance of the current implementation is able to meet the demands of interactive applications.








Plant Synthesis with Plantons

May. 2004

Inspired by the concept of synthesizing texture from textons, we propose a fast plant modeling approach for synthesizing new plants based on a 3D plant sample, whidh is directly recovered from photographs with image-based modeling technique. Our plant model is defined as a set of plantons and their transitions in a Markov model, named planton graph. A planton is represented as a group of similar organs while the transitions between plantons implicitly indicate the intrinsic growth patterns of a plant. We design a two-step analysis to learn plantons and their transitions from the sample. With learnt planton graph, it is easy and fast to synthesize new plants that are statistically similar to but visually different from each other.










Constructive Grammar:

An Approach for Rapid Architecture Modeling

May. 2004

In this paper, we present a semantic rule-driven architecture modeling grammar, namely constructive grammar, which formalizes the knowledge of architecture construction in a set of semantic rules and parameterizes the primary features of buildings to control the generation of models. It is efficient and convenient to generate regular architectures or a class of architectures with the similar style. To demonstrate the utility of our approach, we have implemented a system for modeling Chinese Ancient Architectures based on the constructive grammar. More than one hundred different buildings were generated based on a common rule set comprising about 110 rules.





Recent Publications

l        Qing Wang, Wei Hua, Xueyin Qing, Hujun Bao, Potential Function Based Generalized Rational Parametric Curve, Proceeding in Natural Science (Chinese version) , Vol 14, No2, Feb, 2004.

l        QIN Xu-jia, HUA Wei, FANG Xiang, BAO Hu-jun, PENG Qun-sheng, GFFD: Generalized free-form deformation with scalar fields, Journal of Zhejiang University SCIENCE, V.4 , No. 6, P.623-629, Nov. 2003.

l        Wei CHENWei HUAHujun BAOQunsheng PENG, Real-time Volume Clipping of Ray Casting, JCST, Dec, 2003.

l        Xueying QinEihachiro Nakamae, Wei HuaYasuo Nagai, Qunsheng Peng, Anti-Aliasing of Reflection Effects of Water SurfaceProceedings of 8th International Conference on CAD/Graphic, Macau, 2003.

l        Jing WANG, Wei HUA, Qing WANG, Hujun BAO, Geometry-based Connectivity Compression of Triangle Mesh, Proceedings of 8th International Conference on CAD/Graphic, Macau, 2003.

l        Qing Wang, Wei hua, Guiqing Li, Hujun Bao, Generalized NURBS Curves and Surfaces, GMP 2004, IEEE Computer Society,

l        Wei HUA, Hujun BAO, Qunsheng PENG, Hierarchical Surface Fragments, Proceeding in Natural Science Vol 13, No 9, 2002

l        Wei HUA, Hujun BAO, Qunsheng PENG, A.R. FOREST, The Global Occlusion Map : A New Occlusion Culling ApproachACM VRST’2002, Hong Kong200211

l        Wei HUAQunsheng PENG, Real-time Mirror Reflection Computation for Virtual Scene WalkthoughJournal of Software (Chinese)No.92000.












Patent Applications

  • Cheekong CHUI, Wei HUA, et. al. Extract and Editing Three Dimensional Objects Associated with Volume Images. Filed on 7 July, 2000 (Filing number - 10/332,429). The international patent application is accorded with the International No. PCT/SG00/00100. 
  • A pyramid-quad tree-based texture mapping method, Chinese patent application (200310109176.2), application date Dec 5, 2003.
  • A view-dependent mesh generation method based on quad-queue for terrain data, Chinese patent application (200310108882.5), application date Nov 24, 2003
  • A view-dependent error metric for terrain mesh generation, Chinese patent application (200310108883.x), application date Nov 24, 2003.
  • A Method of Hierarchical Layered Assembled Billboard Packs for Fast Display of Large-Scale Forest, Chinese patent application (200410017677.2), application data April 12, 2004. 





Relievo Picture Modeler®


Demo video

(about 25M, 1.32min)


As an image-based modeling software, it can recover 3D models from single or multiple photographs very efficiently. It provides many topological modifiers to help user construct arbitrary shapes, which are not standard parameterized primitives in the package. It also supports 3D spatial constraints between shapes to reduce the number of photographs.







Recovered geometries

Recovered geometries and textures




Recovered geometries, where the roof in the arbitrary shape is recovered by using topological modifiers

Recovered geometries and textures






MacroDimension® Large-scale Terrain Rendering Package

It consists of two components. One is a real-time rendering engine for large-scale terrain. The other is a visual editing tool. Additionally, a SDK for C++ is also included and it can be used to develop applications that the real-time visualization of large-scale terrain is required.


A visual editing tool for large-scale terrain is provided. User can use it to interactively model or edit terrain, construct constrained texture mapping, rapidly synthesize terrain texture, and optimize either height field or texture of terrain for real-time visualization purpose. It can support height fields with 16k X 16k vertices, textures with 64k X 64K pixels.


Height Filed Size 8193×8193

Texture Size64K×64K


The interface of visual editing tool.

Demo video, about 42M, 2.52min

The rendering result of our engine

Demo video, about 23M, 1.0min


Visual Scenario

Editing And Walkthrough Environment for 3D Scene

It is a multi-purpose software, as a powerful tool to edit and fast visualize large-scale 3D scene in real-time. It can be used in architecture demonstration, landscape design and evaluation, urban planning, 3D geography information integration, environment protection, ancient architecture protection, etc.

Demo video, about 25M, 39Sec







Recent Research Projects



l        The Rapid Modeling, Processing and Real-time Visualization Technique for Large Scale Multiresolution Spatial Information from Multiple Sources”, granted by National 863 High Technology Program, from Jan. 2002 to Dec. 2003.


It is to research/develop some essential and sound techniques for modeling, processing and visualization large-scale multiresolution 3D spatial information from multiple sources, and to develop a software system. We have achieved a certain technological breakthroughs in real-time terrain visualization and image-based modeling. We have developed three prototyped subsystems so far, such as a general geometry modeling system, an image-based modeling system, and a real-time terrain visualization system.




l        Geometric Signal Processing Techniques for Large Scale Scene”, granted by National Science Foundation, from Jan. 2003 - Dec. 2006.


While people are looking forward to more realistic synthesized image in real-time, the rendering technique for large scale scene, as a fundamental technique in Virtual Reality, engineering simulation and other relevant fields, is confronting a serious challenge. This project is dedicated to proposing a novel geometric signal processing method to study a unified processing framework for large scale scene, the optimization approaches for rapid rendering and the appearance-preserving simplification for non-manifold models. It will be of great benefit to the development of real-time large-scale scene processing system either in theory or in practice.