臺灣博碩士論文加值系統

一段影片裡，人物在移動的過程中，其局部或全部的影像被物體所遮蔽，欲移除造成遮蔽的物體，這時需修補人物在遮蔽物後的動作和背景，此方法稱為影片修補。

傳統的影片處理方法，在對於影片的修補方面，分為兩部分：a) 修補前景(人物)被遮蔽時的動作；b) 移除遮蔽物後，修補其位置的背景。針對修補前景這部分，我們提出一個新的影片修補方法，來解決傳統影片修補可能會造成的缺陷，例如人物身體部分的缺失、人物的跳動、非真實移動物體的陰影(ghost shadow)、人物動作不連續的現象等等。

我們的方法，利用影片中的人物影像來驅動三維模型，使我們三維模型的動作是自然且連續，並且產生完整的貼圖，使人物的影像具有完整性。

In some videos, objects might be occluded by barrier while moving. If we want to remove the barrier, needing to use the method called “video inpainting” to reconstruct the occluded regions.

In previous video inpainting methods, it could be divided into two part: a) reconstruct the motion of the moving objects; b) reconstruct the background after removing the occluded region.

For reconstructing motion of moving objects, we present a new method to deal with the defects occurred in previous method. For example, some parts of the objects might be lost or result in objects flicking, ghost shadow, discontinuous motion of objects and so on.

In our method, we use the character image in the video to drive 3D model to make the motion smooth, continuous. Then we get the texture remaining the integrity of the character. The result of video inpainting using our method conquers the artifact of previous method.

中文摘要.....i
Abstract.....ii
Table of Contents.....iii
1 Introduction.....1
1.1 研究背景與動機.....1
1.2 研究貢獻.....3
1.3 論文架構.....4
1.4 系統架構.....5
2 Related Work.....6
2.1 影像切割.....6
2.2 三維模型的變形.....6
2.3 影像修補.....7
2.4 影片修補.....8
3 變形之前處理和遮蔽物移除.....10
3.1 擷取前景.....10
3.2 動作檢索.....12
3.3 產生貼圖.....15
3.4 遮蔽物移除.....17
4 三維模型之變形與動作編輯.....20
4.1 三維模型變形方法.....20
4.2 建立三維模型動作資料庫.....21
4.3 變形三維模型.....23
4.3.1 輪廓的偵測與對齊.....23
4.3.2 尋找控制點與目標點的對應關係.....24
4.3.3 變形結果.....26
4.4 三維模型的動作編輯.....28
5 實驗結果與討論.....30
5.1 實驗結果.....30
5.2 操作與限制.....36
6 結論與未來展望.....37

[1] Connelly Barnes, Eli Shechtman, Adam Finkelstein, and Dan B Goldman. PatchMatch: A Randomized Correspondence Algorithm for Structural Image Editing. ACM Transactions on Graphics, 28(3):1, 2009.
[2] Scott D Cohen and Leonidas J Guibas. Partial Matching of Planar Polylines Under Similarity Transformations. (January):1--10, 1997.
[3] Soheil Darabi. Image Melding : Combining Inconsistent Images using Patch-based Synthesis. 2003.
[4] M. Granados, J. Tompkin, K. Kim, O. Grau, J. Kautz, and C. Theobalt. How Not to Be Seen - Object Removal from Videos of Crowded Scenes. Computer Graphics Forum}, 31(2pt1):219--228, May 2012.
[5] James Hays and Alexei~A Efros. Scene Completion Using Millions of Photographs. 1(212):1--7, 2007.
[6] Vivek Kwatra, Irfan Essa, Aaron Bobick, and Nipun Kwatra. Texture optimization for example-based synthesis. ACM SIGGRAPH 2005 Papers on - SIGGRAPH '05, page 795, 2005.
[7] Yin Li, Jian Sun, Chi-keung Tang, and Heung-yeung Shum. Lazy Snapping.pages 303--308, 2001.
[8] Kedar a Patwardhan, Guillermo Sapiro, and Marcelo Bertalmio. Video inpainting under constrained camera motion. IEEE transactions on image processing : a publication of the IEEE Signal Processing Society}, 16(2):545--53, February 2007.
[9] Patrick Perez, Michel Gangnet, and Andrew Blake. Poisson image editing. ACM SIGGRAPH 2003 Papers on - SIGGRAPH '03}, page 313, 2003.
[10] Olga Sorkine, Daniel Cohen-or Yaron, Marc Alexa, and R Christian. Laplacian Surface Editing. 2003.
[11] Jian Sun and Jiaya Jia. Image Completion with Structure Propagation. 2003.
[12] Yonatan Wexler, Eli Shechtman, and Michal Irani. Space-time completion of video. IEEE transactions on pattern analysis and machine intelligence, 29(3):463--76, March 2007.
[13] Johannes Zimmermann, Andrew Nealen, and Marc Alexa. SilSketch: automated sketch-based editing of surface meshes. Proceedings of the 4th Eurographics workshop on Sketch-based interfaces and modeling - SBIM '07, page~23, 2007.