以機器視覺系統獲得環境中物體的三維資訊的做法，必須先找出參考影像與匹配影像中兩兩對應的像素，才能求算視差、推算三維資訊，而這樣的工作即稱為影像匹配。於影像中不同空間頻率的區域，須以不同尺寸的匹配視窗進行匹配，存在選擇匹配視窗尺寸的問題。多重紋理映射匹配法(multiple mip-map level, MML)[3]，以多重尺寸的匹配視窗進行匹配的方式，解決選擇匹配視窗尺寸的問題。
MML匹配法為解決以多重尺寸的匹配視窗進行匹配時其運算量過於龐大的問題，以盒形濾波器(box filter) 加速匹配，但卻導致其匹配結果失去匹配的細節。所以，我們以積分影像法(integral image)取代box filter來加速匹配解決此問題。我們以高斯分佈函數，依匹配視窗的尺寸賦予其所對應的匹配曲線不同的權重，來改善MML匹配法於物體邊界處匹配結果不佳的問題。因此，我們提出了結合積分影像法與高斯權重實現多重尺寸匹配視窗的影像匹配法。於實驗中，我們證明了此方法能有效改善MML匹配法的匹配結果。

In order to get 3-dimensional information of objects in environment by machine vision system, we must find out the corresponding pixel pair between reference image and the image to be matched to get disparity value,3-dimensional information, and this kind of work is called image matching. In regions of different spatial frequency, we use a variety of sizes of image matching windows to image match, but it exists a problem of choosing image matching windows. Multiple mip-map level method solves the problem of choosing image matching windows by multi-size image matching windows.
The multiple mip-map level method reduce the much computational cost which is caused by image matching with multi-size matching windows. If we use box filter to speed image matching, it will lose details as a result. Due to the above problems stated, we use integral image method in place of box filter to speed matching to solve these problems. We use Gaussian distribution function in accordance with differences of matching windows to get weights of corresponding matching curves, and it can improve the weak results caused by multiple mip-map level method in edges of objects. Therefore, we propose an image matching method of integral image method in combination of Gaussian weights to achieve multi-size matching windows. In our experiment, we proved our method that can improve the result of multiple mip-map level method.

[1]D. Scharstein and R. Szeliski, “A taxonomy and evaluation of dense two-frame stereo correspondence algorithms,” International Journal of Computer Vision, vol.1, no.47, pp.7–42, May 2002.
[2]M. J. Hannah, “Computer matching of areas in stereo images,” PHD thesis, Stanford University, 1974.
[3]R. Yang and M. Pollefeys, “Multi-resolution real-time stereo on commodity graphics hardware,” IEEE Computer Vision and Pattern Recognition, vol.1, pp.211-217, 2004.
[4]Y. L. Murphey, J. Chen, J. Crossman and J. Zhang, “Depth finder, a real-time depth detection system for aided driving,” IEEE Intelligent Vehicles Symposium 2000 Dearborn (MI), USA. October 3-5, 2000.
[5]R. Zabih and J. Woodfill, “Non-parametric local transforms for computing visual correspondence,” European Conference on Computer Vision, vol.2, pp.151–158, 1994.
[6]D. Scharstein, “Matching images by comparing their gradient fields,” International Conference of Pattern Recognition, vol.1, pp.572–575, 1994.
[7]S. Birchfield and C. Tomasi, “Depth discontinuities by pixel to pixel stereo,” IEEE International Conference on Computer Vision, pp.1073–1080, 1998.
[8]T. Kanade, “Development of a video-rate stereo machine,” Image Understanding Workshop, pp.549–557, 1994.
[9]R. Yang, G. Welch and G. Bisop, “Real-time consensus-based scene reconstruction using commodity graphics hardware,” Pacific Graphics 2002, pages 225–234, Beijing, China, October 2002.
[10]M. Tekalp, “Digital video processing,” Prentice Hall, Upper Saddle River, NJ, 1995.
[11]Q. Tian and M. N. Huhns, “Algorithms for sub-pixel registration,” Computer Vision, Graphics and Image Processing, 35:220–233, 1986.
[12]P. Fua, “Aparallel stereo algorithm that produces dense depth maps and preserves image features,” Machine Vision and Applications, 6:35–49, 1993.
[13]P. Viola and M. Jones, “Robust real-time object detection,” Second International Workshop on Statistical and Computational Theories of Vision, Vancouver, Canada, July 13, 2001.
[14]H. Hirschmiiller and D. Scharstein, “Evaluation of cost functions for stereo matching,” IEEE Computer Vision and Pattern Recognition, 2007.
[15]P. Sand and S. Teller, “Video matching,” ACM Transactions on Graphics, vol.23, no.3, 2004.
[16]P. Phonsue and S. Oe, “Stereo matching method using color frequency information,” IEEE International Conference on Mechatronics & Automation Niagara Falls, Canada, July 2005.
[17]R. Feris, R. Raskar and L. Chen, “Multiflash tereopsis: depth edge preserving stereo with small baseline illumination,” IEEE Transactions on Pattern Analysis and Machine intelligence, vol.30, no.1, January 2008.
[18]Y. Tsin, S. B. Kang and R. Szeliski, “Stereo matching with linear superposition of layers,” IEEE Transactions on Pattern Analysis and Machine intelligence, vol.28, no.2, February 2008.
[19]Y. Boykov, O. Veksler, and R. Zabih, “Fast approximate energy minimization via graph cuts,” IEEE Transactions on Pattern Analysis and Machine intelligence, vol.11, no.23, pp.1222–1239, 2001.
[20]V. Kolmogorov and R. Zabih, “ Computing visual correspondence with occlusions using graph cuts,” IEEE International Conference on Computer Vision, vol2,pp.508–515, 2001.
[21]A. Singh, “Optical flow computation : a unified perspective,” IEEE Computer Society Press,1992.
[22]P. Anandan, “A computational framework and an algorithm for visual motion,” International Journal of Computer vision , vol.2, pp.283-310, 1989.
[23]鄭銘揚，即時影像伺服追蹤控制，「影像伺服即時追蹤控制應用」專題研究產學交流成果發表會，中華明國九十年十二月十四日