人類之所以會有立體的感知主要是因為我們使用左右眼觀賞同一個畫面時，我們的左右眼接受到的畫面是有所差距的，這個現象稱之為雙眼視差，當這兩個有著視差的畫面同時進入到我們的左右眼時，我們的大腦就會有一些機制來產生立體的感知。但是如果我們使用這個概念來製作3D影像時，整體的影像大小會是2D影像的兩倍，不僅是儲存成本增加對傳輸頻寬來講也是一個瓶頸。
我們使用DIBR演算法來解決以上的問題，DIBR演算法只需要傳統的2D影像搭配相對應的深度資訊就可以合成出我們所需要的左右眼圖，而深度資訊是灰階的，相對於彩色資訊只需要三分之一的儲存空間，可以大大的降低我們的頻寬需求，達到即時播放的目的。
但DIBR並非沒有缺點，我們最需要解決的問題是影像在位移模擬新視角後，會存在一些沒有色彩資訊的像素，我們稱之為孔洞，尤其在深度變化劇烈的地方會產生大的孔洞，存在著孔洞的影像無法使大腦合成出完整的立體圖像。最常見的解決方式是對深度圖做平滑濾波，此方法雖可降低孔洞的大小，但破壞了原始的深度資訊，造成立體效果的衰減。
本論文將改善上述缺點，只對深度圖的邊緣做優化，消除彩色圖和深度圖邊緣錯位的問題，不僅消除鬼影效應，因為只對邊緣的部分做深度值的改變，立體效果可以大大的增加。但因為沒做前處理的關係，孔洞的範圍會比較大，故在本篇論文將提出一個新的孔洞填補演算法，運用背景紋理的相關性來填補孔洞，降低圖片的失真。最後在依據此DIBR系統提出電路架構，達到即時播放的目的。

The reason why human beings have stereoscopic perception is that our two eyes receive two image with slight difference. simultaneously. This phenomenon is called binocular parallax. When these two images with slight difference are received by our two eyes, our brain will start some mechanism to produce the 3D perception. When we use the concept mentioned above to produce 3D video, the total capacity will be twice of the conventional 2D video and the bandwidth will be the bottleneck.
We use DIBR algorithm to figure out the problem mentioned above. DIBR algorithm use only color image and its associate depth map to synthesize the right- and left-eye image. The depth map stores the 8-bit value for depth, the total image capacity will be one third of the color image and the requirement of the bandwidth will be reduced for the real-time application.
DIBR algorithm is not perfect, the most serious problem we have to solve first is that the color image pixels are shifted to synthesize the new view point, after that, there will be some pixels without color information, this phenomenon is called holes. Especially the place where the depth map change intensively. Smoothing the depth map is the common method to solve this problem. Although it can reduce the scale of holes, the initial depth information is destroyed, the 3D perception will also be reduced.
In this paper, we will solve the problem mentioned above. We only refine the depth map to solve the boundary mismatch problem and the ghost effect. Because we only refine the edge pixels not the whole depth map, the 3D perception will be better. But we take the pre-processing step away in our DIBR system, the hole-region will be larger. We also propose a new hole-filling algorithm using the background texture to fill the holes and reducing the distortion of image. Finally, according to the DIBR system we proposed, we propose our hardware architecture for real-time application.

[1] http://www.3d-display-info.com/tags/market-reports
[2] C. Fehn, “Depth-image-based-rendering (DIBR), compression and transmission
for a new approach on 3D-TV,” in Proc. International Society for Optical Engineering Conf. Stereoscopic Displays and Virtual Reality Systems XI, San Jose, CA, vol.5291, pp.93–104, Jan. 2004.
[3] J. Compen, “3D graphics rendering for multi-view displays: Using programmable
shaders on graphics cards,” Technical note TN-2004/00920, Koninklijke Philips
Electronics N.V. 2005.
[4] http://www.panasonic.com/business/provideo/ag-3da1.asp
[5] Y.-H. Chen, “Using Expanding Parallax Map to Preserve Edge Information for 3DTV,” Master’s Thesis, Dept. of Electrical Engineering, National Cheng Kung University, July 2012.
[6] C. Fehn, “A 3D-TV approach using depth-image-based rendering (DIBR),” in Proc. Vis., Imaging, and Image Process., Sept. 2003, pp. 482-487.
[7] L. Zhang and W.J. Tam, “Stereoscopic Image Generation Based on Depth Images for 3D TV,” IEEE Trans. On Broadcasting, vol. 51, No.2, pp. 1191-199, June 2005.
[8] C. Vázquez, W. J. Tam and F. Speranza, “Stereoscopic Imaging: Filling Disoccluded Areas in Depth Image-Based Rendering,” Proc. Of SPIE, Vol. 6392, Canada, 2006.
[9] K-J. Oh, S. Yea and Y-S. Ho, “Hole filling method using depth based in-painting for view synthesis in free viewpoint television and 3-d video,” in PCS, Piscataway, NJ, USA, 2009, pp. 233–236, IEEE Press.
[10] https://code.google.com/p/ct2-inpainting/
[11] http://www.researchgate.net/publication/3208254_Tutorial_time- multiplexed_stereoscopic_computer_graphics
[12] Xuyuan Xu, Lai-Man Po, Kwok-Wai Cheung#, Ka-Ho Ng, Ka-Man Wong, Chi-Wang Ting,” A FOREGROUND BIASED DEPTH MAP REFINEMENT METHOD FOR DIBR VIEW SYNTHESIS,” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2012.
[13] H.-J. Chen, F.-H Lo, F.-C. Jan, and S.-D. Wu,“Real-Time Multiview Rendering Architecture for Autostereoscopic Displays,” Proc. IEEE Intl, Symp. Circuits and Systems (ISCAS), pp. 1165-1168, May 2010.
[14] H.-C. Chen “Real-time Stereoscopic Image Generation Using Depth Image Based Rendering with Virtual View Point Estimation,” Master’s Thesis, Dept. of Electrical Engineering, National Cheng-Kung University, July 2010.
[15] W.-Y. Lee “Improved Hole-filling Algorithms and Hardware Implementation for Virtual View Generation,” Master’s Thesis, Dept. of Electrical Engineering, National Tsing-Hua University, July 2011.
[16] http://research.microsoft.com/en-us/downloads/5e4675af-03f4-4b16-b3bc- a85c5bafb21d/