瑕疵抑制之虛擬視角生成映射演算法

Artifact Reduction Warping Algorithm for Virtual View Generation

王鈺崇(Yu-Chung Wang)

指導教授 ： 李佩君

國立暨南國際大學/科技學院/電機工程學系/碩士(2013年)

摘要

為了滿足影像系統在立體視訊(Three-Dimension Video，簡稱3DV)與自由視角視角(Free-ViewPoint TV，簡稱FTV)上的應用，因此近年來有不少自由視角合成演算法逐漸被提出，其中大多數是採用2D彩圖搭配深度圖(Depth map)資訊來重建任意虛擬視角且根據Depth image-Based Render(DIBR)基礎理論進行系統的轉換。近幾年所發表的自由視角合成系統只針對於密集且平行相機陣列系統進行開法，拍攝影像時攝影機之間擺設位置可根據人眼的視差(disparity)係數進行調整，如果相機間的距離太大或太小將會影響到觀賞者的立體視訊效果。為了達到產生新的虛擬視角影像。
本篇論文以稀疏的相機陣列開發以DIBR系統為基礎之自由視角合成系統，一開始我們會先分析3DVC標準所提供的虛擬視角合成軟體(View Synthesis Reference Software，簡稱 VSRS)，其中，加入相機參數的優點用以提升輸出影像的品質，但是卻會使得整體的運算過程更加複雜化且運算量增加。因此論文將探討如何不影響輸出影像之品質的情況下，捨棄VSRS中的相機參數以達到運算流程進行簡化與減少整體的運算量，提出自由視角合成系統來生成新的虛擬視角。此外，在影像Warping的過程運算中，如果影像深度值資訊不太準確，將導致虛擬視角影像邊緣區域產生瑕疵(artifact)效應，本論文也將會針對瑕疵的區域進行消除。
最後，會將原本輸入的影像資訊轉換成含有正、負視差影像，來搭配虛擬視角影像輸出至3D顯示器上，觀賞具有正、負視差立體影像。

關鍵字

影像合成 ； 虛擬視角 ； 瑕疵抑制 ； 立體影像 ； Stereo Matching

並列摘要

In order to satisfy the application of Three-dimensional video (3DV) and Free Viewpoint Video (FVV) in video system, there are many free view synthesis algorithms are proposed. Using of 2D color with depth map information to reconstruct any virtual perspective is a concept of Depth image-Based Render (DIBR) for 2D to 3D conversion. The free viewpoint video synthesis system is designed for parallel cameras array system, ball view images can be adjust by cameras distance which can be all human eye parallax disparity coefficients, if the camera distance is too large or the too small will affect the viewer stereoscopic video experience.
In order to achieve a new virtual view images system this thesis develops a free camera position synthesis system. Base on DIBR in sparse arrays cameras configure, we study the virtual view synthesis software method, which can improve output image quality by using camera parameters. However, the computation time will increase. Therefore, this thesis proposes a disparity estimation algorithm to instead of the camera to reduce computation on camera parameter estimate. In addition, this thesis proposes a virtual image warping to the artifact in edge area.
The synthesized image includes positive and negative parallax images to match 3D video and give an real 3D experience for viewers.

並列關鍵字

View Synthesis ； Artifact Reduction ； Stereo Matching ； 3D Video

參考文獻

[1] M. Tanimoto, “FTV (Free Viewpoint TV),” in proc. of International Conference on Image Processing, September, pp.26-29, 2010.

[2] M. Tanimoto, “Overview of Free Viewpoint Television,” in proc. of Signal Processing: Image Communication,vol.21,no.6,pp.454-461,July 2006

[3] S. Zinger, L. Do, and P.H.N. de, “Free-Viewpoint Depth Image Based Rendering,” in proc. of Journal Visual Communication and Image Representation, pp. 533-541, 2010.

[4] A. Smolic and McCutchen, “3DAV Exploration of Video-Based Rendering Technology in MPEG,” in proc. of IEEE Trans. Circuits Syst. Video Technol., Special Issue on Immersive Commun., vol.41,no.9,pp.348-356, March 2004.

[5] A. Smolic, K. Mueller, P. Merkle, C. Fehn, P. Kauff, P. Eisert and T. Wiegand, “3D Video and Free Viewpoint Video - Technologies, Applications and MPEG Standards,” in proc. of IEEE International Conference on Multimedia and Expo, pp. 2161-2164, 9-12 July 2006.