由於裸眼多視角3D電視顯示系統的崛起，即時的產生多個虛擬視角是必須的。因此，對於生成多視角的深度影像繪圖法(DIBR)系統，產生深度圖的立體匹配系統的速度以及正確性顯得非常重要。然而，儘管使用區域立體匹配方法加上GPUs的平行化運作的幫助，在傳統的方法上依然很難達到即時。因此，我們提出成本推算的方式，能夠將某一個視角的成本圖迅速的推算出另一個視角的成本圖，靠著這種方式，能夠省下大量花在成本聚集的步驟上。另外，因為使用簡單的區域立體匹配的方式，所以使用比較精準的修正方式。所以本論文提出可以對平滑區域的雜亂深度值進行修補的修正前處理，以及對遮蔽區域進行精確的十字區域投票修正和矩形區域投票修正。透過這些修正，初始的深度圖在邊界上會正確許多，此外，平滑區域的雜亂深度值也可以獲得改善。

With the rise of naked-eyes multi-view 3DTV display system, it is necessary to create multi views in real-time. Therefore, fast stereo matching, which could create the precise depth map for the depth-image-based-rendering (DIBR) to produce multi views, becomes very important. However, traditional stereo matching methods are hard to reach the goal of real-time even by using local stereo matching approaches and implementing on GPUs, which can obtain desirable speedup leveraging with parallel computing. Therefore, a cost reproduction method, which can immediately transfer the cost in one view to obtain the cost in the other view, is proposed in our system. Based on this concept, much time, which is consumed on cost aggregation, will be significantly reduced. Moreover, the pre-refinement method is proposed to deal with incorrect disparity values in smooth regions. Besides, the cross-based and window voting refinement algorithms, which can revise occluded pixels precisely, are suggested. By using the proposed refinements, the disparities on the edge will be more correct than the original disparity map. Also, the most incorrect disparities will be recovered.

[1] S. B. Kang. A survey of image-based rendering techniques. In VideoMetrics, SPIE vol. 3641, pp. 2–16, 1999.
[2] J. Lengyel. The convergence of graphics and vision. Technical report, IEEE Computer, July 1998.
[3] Wang, and S.P. Patrick, Pattern Recognition, Machine Intelligence and Biometrics. 2011.
[4] J. Sun, N. N. Zheng, and H. Y. Shum, “Stereo matching using belief propagation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 25, pp. 787–800, Jul. 2003.
[5] Y. Boykov, O. Veksler, and R. Zabih, “Fast approximate energy minimization via graph cuts,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 23, pp. 1222–1239, Nov. 2001.
[6] X. Sun, X. Mei, S. Jiao, M. Zhou, and H. Wang, “Stereo matching with reliable disparity propagation,” in Proc. Int. Conf. 3D Image, Model. Process. Vis. Transm., May 2011, pp. 132–139.
[7] S. B. Kang, R. Szeliski, and J. Chai, “Handling occlusions in dense multi-view stereo,” in Proc. IEEE Comp. Soc. Conf. Comp. Vis. Pattern Recognit., vol. 1, 2001, pp. 103–110.
[8] O. Veksler, “Fast variable window for stereo correspondence using integral images,” in Proc. IEEE Comp. Soc. Conf. Comp. Vis. Pattern Recognit., vol. 1, Jun. 2003, pp. 556–561.
[9] K. J. Yoon and I. S. Kweon, “Adaptive support-weight approach for correspondence search,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 28, no.4, pp. 650–656, 2006.
[10] K. Zhang, J. Lu, and G. Lafruit, “Cross-based local stereo matching using orthogonal integral images,” IEEE Trans. Circuits Syst. Video Technol., vol. 19, pp. 1073–1079, Jul. 2009.
[11] J. D. Owens, D. Luebke, N. Govindaraju, M. Harris, J. Kruger, A. E. Lefohn, and T. J. Purcell, “A survey of general-purpose computation on graphics hardware,” Comp. Graph. Forum, vol. 26, Mar. 2007, pp. 80–113
[12] A. Hosni, M. Bleyer, C. Rhemann, M. Gelautz, and C. Rother, “Real-time local stereo matching using guided image filtering,” in Proc. IEEE Int. Conf. Multimed. Expo, Jul. 2011, pp. 1–6.
[13] J. Kowalczuk, E. T. Psota, and L. C. Perez, “Real-time stereo matching on CUDA using an iterative refinement method for adaptive support-weight correspondences,” IEEE Trans. Circuits Syst. Video Technol., vol. 23, pp. 94–104, Jan. 2013.
[14] J. Lu, S. Rogmans, G. Lafruit, and F. Catthoor, “Stream-centric stereo matching and view synthesis: a high-speed approach on GPUs,” IEEE Trans. Circuits Syst. Video Techonol., vol. 19, no. 11, Nov. 2009.
[15] A. Fusiello, V. Roberto, and E. Trucco, “Efficient stereo with multiple windowing,” in Proc. IEEE Comp. Soc. Conf. Comp. Vis. Pattern Recognit., Jun. 1997, pp. 858–863.
[16] D. Scharstein, R. Szeliski, and R. Zabih, “A taxonomy and evaluation of dense two-frame stereo correspondence algorithms,” in Proc. IEEE Workshop Stereo Multi-Baseline Vis., Dec. 2001, pp.131–140.
[17] K. Zhang, J. Lu, Q. Yang, G. Lafruit, R. Lauwereins, and L. Van Gool, “Real-time and accurate stereo: a scalable approach with bitwise fast voting on CUDA,” IEEE Trans. Circuits Syst. Video Technol., vol. 21, pp. 867–878, Jul. 2011.
[18] X. Mei, X. Sun, M. Zhou, S. Jiao, H. Wang, and X. Zhang, “On building an accurate stereo matching system on graphics hardware,” in Proc. IEEE Int. Conf. Comp. Vis. Workshops, Nov. 2011, pp. 467–474.
[19] R. Zabih, and J.Woodfill, “Non-parametric local transforms for computing visual correspondence,” in Proc. Eur. Conf. Comp. Vis. , 1994, pp. 151–158.
[20] Middlebury Stereo Vision Page [Online]. Available: http://vision.middlebury.edu/stereo
[21] H. Hirschmuller, “Stereo processing by semiglobal matching and mutual information,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 30, pp. 328–341, Feb. 2008.
[22] R. Gupta and S.-Y. Cho, “Real-time stereo matching using adaptive binary window,” 3DPVT, 2010.
[23] K. Zhang, J. Lu, G. Lafruit, R. Lauwereins, L. Van Gool, “Real-time accurate stereo with bitwise fast voting on CUDA,” in IEEE Workshops Comp. Vis., Sept. 2009, pp. 794–800