本文提出了一種由四個階段組成的局部立體匹配演算法。在初始成本計算階段，我們首先計算顏色SAD和Haar小波HL帶係數SAD，然後使用彩圖計算出canny邊緣圖來組合這兩種類型的初始成本。在成本聚合階段，我們使用自適應支持權重算法。 在視差決策階段，我們使用獲勝者通吃方法來獲得初步的視差圖。在細化階段，我們首先執行左右一致性檢查以標出錯誤點，接著使用所提出的雙向校正方法來校正誤差點，最後使用加權中值濾波器去除錯誤，並得到最終結果。
和其他方法相比，我們的方法計算出的深度圖在深度不連續區有較良好的結果，能更準確地計算出物體輪廓的深度，而後處理部分加入的中值濾波器能讓無法被偵測出的錯誤也能夠得到修正。像Cone圖片右上角窗格的例子，我們的方法較能切齊窗格形狀的深度，其他方法大多無法切齊窗格的邊界，甚至連窗格的洞的都算不出來。總之，我們方法的優點在於能夠得到較準確的前景深度。

This thesis proposed a local stereo matching algorithm consisting of four stages. In the initial cost computation stage, we first calculated the color SAD and Haar wavelet HL band Coefficient SAD, then combine these two types of initial cost based on the canny edge map calculated from the input color image. In the cost aggregation stage, we use the adaptive support weight algorithm. In the disparity decision stage, we use the winner-take-all method to determine the raw disparity map. In the refinement stage, we first perform the left right consistency check to find the error points, then use the proposed two-direction correction method to correct them and finally use the weighted median filter to denoise the result.
Compared with other methods, the depth map calculated by our method can obtain the depth of the contour more precisely. The weighted median filter used in the post processing stage can better correct the errors detected by the LRC. For example, in the upper right pane of the Cones image, we can more accurately calculate the shape of the pane which is usually smoothed out by other methods. In conclusion, the foreground depth map obtained by our method is often better than other methods.

[1] Middlebury Stereo Vision website [Online]. Available: http://vision.middlebury.edu/stereo/.
[2] S. Birchfield and C. Tomasi, "A pixel dissimilarity measure that is insensitive to image sampling," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 20, pp. 401-406, 1998.
[3] D. Chen, M. Ardabilian and L. Chen, "A fast trilateral filter-based adaptive support weight method for stereo matching," IEEE Transactions on Circuits and Systems for Video Technology, vol. 25, pp. 730-743, 2014.
[4] K. He, J. Sun and X. Tang, "Guided image filtering," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, pp. 1397-1409, 2012.
[5] A. Hosni, M. Bleyer, M. Gelautz and C. Rhemann, "Local stereo matching using geodesic support weights," IEEE International Conference on Image Processing pp. 2093-2096, 2009.
[6] A. Hosni, C. Rhemann, M. Bleyer, C. Rother and M. Gelautz, "Fast cost-volume filtering for visual correspondence and beyond," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, pp. 504-511, 2012.
[7] X. Huang, C. Yuan and J. Zhang, "Graph cuts stereo matching based on patch-match and ground control points constraint," Pacific Rim Conference on Multimedia, Springer, pp. 14-23, 2015.
[8] J. Jiao, R. Wang, W. Wang, S. Dong, Z. Wang and W. Gao, "Local stereo matching with improved matching cost and disparity refinement," IEEE MultiMedia, vol. 21, pp. 16-27, 2014.
[9] J. Ko and Y. S. Ho, "Stereo matching using census transform of adaptive window sizes with gradient images," IEEE Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, pp. 1-4, 2016.
[10] Z. Ma, K. He, Y. Wei, J. Sun and E. Wu, "Constant time weighted median filtering for stereo matching and beyond," IEEE International Conference on Computer Vision, pp. 49-56, 2013.
[11] X. Mei, X. Sun, M. Zhou, S. Jiao, H. Wang and X. Zhang, "On building an accurate stereo matching system on graphics hardware," IEEE International Conference on Computer Vision Workshops, pp. 467-474, 2011.
[12] M. Mozerov and J. Van Weijer, "Accurate stereo matching by two step global optimization," IEEE Transactions on Image Processing vol. 24, pp. 1153-1163, 2015.
[13] P. Ndjiki-Nya et al., "Depth image-based rendering with advanced texture synthesis for 3-D video," IEEE Transactions on Multimedia, vol. 13, pp. 453-465, 2011.
[14] B. Salehian, A. M. Fotouhi and A. A. Raie, "Dynamic programming-based dense stereo matching improvement using an efficient search space reduction technique," Optik, vol. 160, pp. 1-12, 2018.
[15] C. H. Wei, S. H. Lai and C. K. Chiang, "Video synthesis from stereo videos with iterative depth refinement," Journal of Visual Communication and Image Representation, vol. 47, pp. 48-61, 2017.
[16] Q. Yang, "A non-local cost aggregation method for stereo matching," IEEE Conference on Computer Vision and Pattern Recognition, IEEE, pp. 1402-1409, 2012.
[17] S. H. Yang and B. Wang, "Improved adaptive-support stereo matching using local features," Journal of Information Science and Engineering, vol. 31, pp. 1557-1575, 2015.
[18] K. J. Yoon and I. S. Kweon, "Adaptive support-weight approach for correspondence search," IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 28, pp. 650-656, 2006.
[19] R. Zabih and J. Woodfill, "Non-parametric local transforms for computing visual correspondence," European Conference on Computer Vision, Springer, pp. 151-158, 1994.
[20] Y. Zhan, Y. Gu, K. Huang, C. Zhang and K. Hu, "Accurate image-guided stereo matching with efficient matching cost and disparity refinement," IEEE Transactions on Circuits and Systems for Video Technology, vol. 26, pp. 1632-1645, 2015.
[21] K. Zhang, J. Lu and G. Lafruit, "Cross-based local stereo matching using orthogonal integral images," IEEE transactions on Circuits and Systems for Video Technology, vol. 19, pp. 1073-1079, 2009.