透過您的圖書館登入
IP:18.218.254.122
  • 學位論文

最佳化視覺資訊與可視度完備性的模型全覽方法

Model overview by optimizing visual information and visibility completeness

指導教授 : 莊永裕
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


模型的視野參數選擇是一個被廣泛研究的主題。由於整個問題包 含許多的變數,而且所謂「理想的結果」通常取決於實際上的應用, 因此不同領域的研究者關心此的問題不同部份,而非嘗試為整個問題 提供一個一般解。此問題的其中一個應用是,如何自動產生數個視點 或者一段全覽影片(overview video),來幫助使用者了解一個三維模 型。這個應用的困難之處在於如何評估一個視點或是一段全覽影片的 好壞。在這篇論文中,我們介紹一個基於香農熵(Shannon Entropy) 的評估法,用以比較不同全覽影片所含有的總資訊量。對於一個固定 光照參數的三維模型,這個評估法同時考慮了視覺上的品質以及可視 度上的完備性。此外,該評估法可以簡單的以今日的繪圖硬體來實 作,因此可以在合理的時間內完成所需的計算。於是我們可以使用此 評估法加上攝影機移動的限制式,自動的產生一段平滑的攝影機路徑 來得到三維模型的全覽影片。

並列摘要


Selection of viewing parameters is a widely studied problem. Researchers from different areas pay attentions to different parts of the problem because the complete one involves too many variables, and the ideal results highly depend on the concrete applications. One of the topics is to understand a 3D model by automatically generated views or overviewvideos. The difficulty comes from how the quality of views and videos be evaluated. In this thesis, we introduce a measure based on Shannon entropy to compare the amount of information for different overview videos. The measurement considers both visual quality and visibility completeness of the target 3D model with fixed lighting parameters. Furthermore, the measurement can be easily implemented on graphics hardware so the computation can be completed in reasonable time. We then combine the measurement with camera movement constrains to automatically generate a smooth camera control path that provides an overview for the model.

參考文獻


[1] C. Andujar, P. P. Vazquez, and M. Fairen. Way-finder: Guided tours through complex
[2] J. Assa, D. Cohen-Or, I.-C. Yeh, and T.-Y. Lee. Motion overview of human actions.
ACM Transactions on Graphics, 27(5):1–10, 2008.
[3] P. Barral, G. Dorme, and D. Plemenos. Scene understanding techniques using a
[4] M. Christie and P. Olivier. Camera control in computer graphics. In Eurographics

延伸閱讀