臺灣博碩士論文加值系統

在建置傳統的三維模型中，我們所需要許多繁複的手續。從攝影機影像中，點對點之對應、投影回三維空間中之特徵點確認、建製雲點、對性質相似之雲點合成、將雲點繪製成網圖、影像貼回網圖之中，到最後完成三維模型，這樣繁複的過程，有沒有其他替代的方式?
在本篇文章中，我們將使用有別於一般建立三維模型的方法，即為建立Ray-space。Ray-space的建立方法為將各攝影機所取得之影像，利用真實世界中每一個物體投影於Ray-space之中時，只會有其唯一的特性(固定之震幅、相位)，再來將影像中所有觀察到的物體進行討論。
首先將我們從各攝影機所取得的影像相互間找尋對應，但並非每一張影像中的每一個像素都會找到其對應點，所以從我們的觀察中發現了在同一張影像之中，當相鄰像素與像素間投影到ray-space時，會有部分的關聯性(震幅、相位)，進而得到未取得對應點的像素之對應關係。
了解各個影像與影像間對應關係後，再將所有取得之對應點投影回Ray-space之中，再經由EPI (Epipolar Plane Image)做出Ray-space的修補，進而完成建置完整的Ray-space。當得到完整的Ray-space時，我們就得到了近似建立三維模型的效果。

We need to have complicated flew in the traditional 3D-model.for example, getting matching points, projecting in the 3D space, building cloud points, matching cloud points, meshing cloud points, and taking image to mesh, all step are complicated. What can be changed in this method ?
In this thesis, we use a model,named Ray-space. It used the ray in the real world, and used the direction to build Ray-space. Everything in the real world had only one show points in the Ray-space.
First, we found the matching points by two cameras. Second, we use the matching points to get the parameter in the Ray-space.Third, we use the epipolar plane image to let Ray-space completely.

Chapter 1 Introduction 1
Chapter 2 Related Works 3
2.1 三維立體模型 3
2.1.1 雲點偵測 4
2.1.2 雲點合併 5
2.1.3 雲點網狀化 6
2.2 Ray-space空間模型 7
2.2.1 Orthogonal Ray-Space 8
2.2.2 Spherical Ray-Space 10
2.2.3 修補影像空洞 12
Chapter 3 Proposed Algorithm 14
3.1 影像對應與r及α之討論 15
3.1.1 影像對應 15
3.1.2 r及α之取得 18
3.1.3 r及α之修補與去除 20
3.2 Ray-space之修補 23
3.2.1 EPI之內差修補 24
3.2.2 EPI之遮蔽及雜訊修正 26
Chapter 4 Experimental Results 28
4.1 影像對應 29
4.2 r及α的討論 32
4.3 Ray-space之討論 33
Chapter 5 Conclusion 40
References 41

[1]Yebin Liu, Qionghai Dai, Senior Member, "A Point-Cloud- Based Multiview Stereo Algorithm for Free-Viewpoint Viedo " IEEE Transactions on visualization and computer graphics, vol. 16,no.3 ,may/june 2010.
[2]Takeyuki Kobayashi, Toshiaki Fujii, Tadahiko Kimoto and Masayuki Tanimoto, "Inerpolation of Ray-Space Data by Adaptive Filtering " In Three-Dimensional Image Capture and Applications III, Proceedings of SPIE Vol. 3958, 2000
[3]David G. LOWE, "Distinctive Image Features from Scale-Invariant Keypoints " International Journal of Computer Vision 60(2), 91–110, 2004
[4]Jean-Michel Morel, Guoshen Yu, "ASIFT: A new framework for fully affine invariant image comparison " SIAM Journal on Imaging Sciences, 2009
[5]Engin Tola, Vincent Lepetit, Pascal Fua, " DAISY: An Efficient Dense Descriptor Applied to Wide Baseline Stereo" IEEE Transactions on Pattern Analysis and Machine Intelligence Vol. 32, Nr. 5, pp. 815 - 830, May 2010
[6]Takeyuki Kobayashi, Toshiaki Fujii, Tadahiko Kimoto and Masayuki Tanimoto, " Interpolation of Ray-Space Data by Adaptive Filtering " In Three-Dimensional Image Capture and Applications III, Brian D. Conrner, Joseph H. Nurre, Editors, Proceedings of SPIE Vol. 3958, 2000
[7]Masayuki Tanimoto, Mehrdad Panahpour Tehrani, Toshiaki Fujii, and Tomohiro Yendo, " Free-Viewpoint TV " IEEE Signal Processing Magazine [67], January 2011