- 學位論文
自動保留外形特徵的全域性網格最佳化法
Global Mesh Optimization with Automatic Shape Feature Preservation
摘要
隨著近年電腦硬體的蓬勃發展，無論是在學術、工業以及娛樂界，三維模型在各個領域已成了不可或缺的要件。而現今三維模型普遍是透過三維掃描硬體設備產生，或是由動畫師透過相關的編輯軟體繪製得來。無論是採用何種方式產生3D模型，基於任三點即可建立一平面的特性，建構的3D模型都會採用三角網格結構。然而，三角網格結構普遍會有不規則性的問題，這導致後續應用如重建、壓縮或成像等技術無法達到理想的結果。為了改善三維模型散亂的網格結構，提升後續應用的可能性及品質，網格重製、最佳化技術之研究已經成為電腦圖學重要的研究主題之一。 在網格重製與最佳化的研究中有兩個主要的目標：其一是保留原始模型之外形特徵，避免因網格處理而造成無法接受的外貌失真；其二是有效改善網格結構，提升整體三角片的規則度。在過往的網格規則度最佳化技術中，模型外形特徵主要是依靠人工選取的方式來保存，而網格規則度的提升則是依靠調整頂點位置來改善頂點分佈的均勻程度。傳統的作法雖然可以提升網格的規則度，但是需要耗費大量時間選取模型外形特徵，並且無法保證重要的特徵一定會被選取。而單純依靠調整頂點的方式來改善網格規則度，會因為原始模型相連關係不夠理想而無法達到預期的成效。因此，在本論文中提出了一個可以自動保留模型外形特徵的網格最佳化演算法。首先透過表面分群法萃取出模型的尖銳特徵並作為網格調整時的限制條件。接著利用線性系統求解的方式搭配幾何編輯技術，對頂點分佈和相連關係的規則度進行全域性的調整，讓處理後的模型網格規則度有更顯著的提升。
並列摘要
With the flourishing development of the computer hardware in recent years, regardless of academy, industry and amusement circles, the 3D model has already become the indispensable important document in each field. These 3D models generally produce now through the 3D scanner, or drawn and obtained through the relevant editing software by the animation technician. No matter which kind of way is adopted produces 3D model. Basis the characteristic of any three vertex could decide one plane, 3D model usually constructed by the triangle mesh. However, It is hard to get the regular triangle mesh at the beginning, this causes follow-up application like reconstruction、compression and rendering difficulty. For improve 3D model in disorder structure, rise possibility and quality that follow-up employs, the remesh and optimization technique had become one of the most popular subject in computer graphic research. There are two main goals in the research of remeh and mesh optimization：One is to keep the outward appearance, prevent the distortion from mesh processing. The other one is to improve the mesh structure and whole triangle regular. In passing mesh optimization technology, the appearance characteristic of the model mainly relies on the artificial way that chooses to be kept. The mesh regularity is dependents on the vertex adjust. Though traditional techniques can improve the mesh structure, but consume a large number of time and need to choosing the appearance characteristic of the model, and unable to guarantee the important characteristic is sure to be chosen. Because of the model’s connectivity is not ideal enough at the beginning, only depend on vertex adjustion is unable to reach the expected effect of the whole structure improvement. Therefore, in this paper we propose a mesh optimization rule which can keep the model’s sharp feature automatically. At first, algorithm through the surface partition method to extract sharp feature to be the limited condition. Then let the vertex distribution and connectivity being global adjustion by solving the linear system and though geometric editing. Let treated model have more apparent improvement finally.
參考文獻
延伸閱讀
- Chang, H. Y. (2019). 基於跨領域映射達成非監督式內容保留風格導向的圖片轉換 [master's thesis, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU201903078
- 林啟榮（2010）。一種簡約有效特徵選取方法在模式感知網路上之運用〔碩士論文，健行科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0022-2107201022442600
- Chang, C. H. (2013). 應用多項式模糊模型之網路控制系統強健H∞設計 [master's thesis, National Chung Cheng University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0033-2110201613562295
- Huang, M. H. (2008). Local Shape Feature Extraction for 3D Model Retrieval [master's thesis, National Tsing Hua University]. Airiti Library. https://doi.org/10.6843/NTHU.2008.00642
- 施詠禎（2010）。Regular Mesh Simplification〔碩士論文，中原大學〕。華藝線上圖書館。https://doi.org/10.6840/cycu201000335