三維多邊形處理、檢索及驗證方法

Abstract

本論文提出了三種適於計算機圖學之相關應用的多邊形處理技術。首先，我們提出一種利用視覺感知特徵為引導的三維多邊形物件分解方法。所提方法利用視覺感知相關的特徵引導三維多邊形物件的分解過程。由於特徵擷取法係根基於認知心理學中的視覺特點理論，因此所提分解方法能夠適切地仿傚人類視覺對物件之構成要素的感知能力。透過物件的拆解/分解，有許多計算機圖學為主的應用能夠達到效能的提升。這些應用包括：碰撞偵測、熱幅射能量成像模擬、強健式傳輸與連續式傳輸、材質貼圖、三維形變、多邊形化簡與壓縮法、以骨架為主的動畫製作以及三維物件檢索系統。透過研究考察，我們發現既有的三維物件檢索系統缺乏心理學理論的支撐與連結。反觀心裡學領域，許多認知心裡學專家已提出了許多重要的理論(或規則)闡述人類視覺對物件或型式的感知過程。有鑑於此，在第二項研究中，我們提出了一套以認知心理學為基礎的三維物件檢索系統。我們將具體化源自於認知心理學領域的概念性理論，用以設計特徵擷取及比對方法，並進而實現一套能仿傚人類對物件辨識與認知處理的三維物件檢索系統。在第三項研究中，為了讓使用者能驗證與偵測所擷取之三維物件的篡改與否，我們提出一種利用新的脆弱型浮水印技術以達到篡改偵測與驗證的目的。所提方法所嵌入之浮水印除了能抵抗一些非惡意的幾何處理(如頂點座標量化及頂點順序重組)，尚能利用視覺檢視三維多邊形物件的篡改處。

In this dissertation, we propose three mesh processing techniques for 3-D graphics-related applications. First, we propose a novel mesh decomposition scheme called ``visual salience-guided mesh decomposition,'' which uses visually salient features to guide the mesh decomposition process. Since the features adopted are closely related to the psychology-based theory of visual salience, the decomposition process can appropriately mimic the function of a human visual system. There are a variety of applications that benefit from breaking up a 3-D object into components. These applications include collision detection, radiosity simulations, robust transmission and streaming, texture mapping, metamorphosis, 3-D shape retrieval, simplification and compression, watermarking, and control skeleton extraction for key-frame animation. In our investigation, the existing 3-D shape analysis and retrieval algorithms are lack of explicit link to psychology-based principles, while cognitive psychologists have found a set of principles (or properties) that are fairly important in perception of a form or a shape. This motivates us to conduct our second work called ``a cognitive psychology-based approach for 3-D shape retrieval.'' In this work, we incorporate a set of principles, which is originated from cognitive psychology, into the design of 3-D shape retrieval system. The proposed approach is intended for mimicking human visual perception and recognition based on the psychology-based rules proposed by Hoffman and Singh. In addition, our system realizes a ``recognition-by-components'' and ``recognition-by-visually-salient-components'' search strategies. Moreover, use of combined search strategies can perform a coarse-to-fine retrieval task. In our third work, to provide a user with an ability to verify the integrity of the content he/she received (or retrieved), we propose a new fragile watermarking scheme for authenticating 3-D polygonal meshes. The proposed scheme can not only achieve localization of malicious modifications in visual inspection, but also is immune to certain incidental data processings, such as quantization of vertex coordinates and vertex reordering.