Abstract
摘要:網路圖學 (Web Graphics) 為一個重要且新生的平台 (platform),因為有越來越多的使用者在網路世界 (Web) 中也能擁有三維圖學 (3D graphics) 的支援。以三角形方式呈現之三維幾何模型 (geometric 3D model) 是目前電腦圖學界應用最廣泛之三維模型呈現法,而如何將三維幾何模型有效率的在網路上傳輸也成了一個重要的課題,因為三維幾何模型的使用為三維圖學不可或缺的要素之一。但為了表現極纖細的特徵,三維幾何模型的資料量通常都非常的龐大,雖然許多時候使用者並不需要如此纖細的三維幾何模型。因此,如何以少量的資料來表現三維幾何模型並還能以維持其形狀與特徵便成為一件很重要的課題,即使是組成該三維模型的網格 (mesh) 是未結構化 (unstructured) 的。
在本項計畫中,為了保有原來三維幾何模型的形狀與特徵,我們首先偵測 (detect) 三維幾何模型的特徵,並且根據該特徵來將此三維模型分割 (segment) 成數個部分。三維幾何模型的特徵可依照不同的材質 (material property) 或是此三維模型製作者所指定的特徵來做偵測。
Abstract: Web Graphics has become an important and new platform today, since there are more and more people that would like to have 3D graphics supports on the Web. How to transmit geometric 3D models efficiently over the Internet is an important task in Web Graphics, since there are many users using geometric 3D models on the Web. The data size of a geometric 3D model is usually large to enable more detail to be represented, although the user on the Internet usually does not need to use such a detailed model in most cases. Hence, it is necessary to represent the geometric 3D model while keeping the data size small and preserving the shape and features, even if the meshes that constitute the 3D model are unstructured.
In this project, we first detect the features of the geometric 3D model in order to preserve the shape and features of the original model, and then segment the 3D model into several parts according to the detected features. The features of a geometric 3D model can be found as the parts of different material properties or due to the features defined by the 3D model provider. The meshes that constitute the 3D model are called unstructured if there is no such feature in it, hence the features of this kind of 3D model should be detected by the curvature of the 3D model’s surface. Therefore, to simplify each part of the 3D model iteratively will not have the over-simplification problem, which is a common problem of other mesh simplification methods. Even the meshes consisted by the model is unstructured; since the feature detection method we will use can achieve this kind of models, our method can also get a good result efficiently. For the model simplification method, we will use a reversible mesh simplification method, so that the simplified 3D model can be reconstructed by adding some patches which contain the removed information while simplifying. Moreover, due to the reversible mesh simplification method, we can provide multi-resolution 3D models for showing different 3D models with suitable resolution on different kinds of the display devices, for transmitting the 3D model over the Internet according to the unstable network bandwidth, and for LOD (level-of-detail) uses.
There are many methods for simplifying geometric 3D models; most of them are generally time-consuming due to model optimization. These kinds of time-consuming methods are hardly used on the Web page directly. Moreover, the simplified models created by some previous methods can not be easily recognized nor used to reconstruct the original model. Therefore, in this project we will implement a mesh simplification method and the simplified 3D model generated by our method will be suitable for transmitting over the Internet. In order to provide a client/server system for evaluating the method, we will also contribute a 3D development environment for Web Graphics.
Keyword(s)
幾何模型簡化
網路圖學
模型特徵辨識
LOD
mesh simplification
web graphics
model feature detection
level-of-detail