本文主要的目的是改變影像拼接過程中尋找單應性矩陣所使用的隨機抽樣一致 法，而以迭代線性最小平方法取代。首先，我們以 Bay 等人所提出的加速穩健特 徵演算法為影像拼接的基礎，找出圖像上的特徵點進行特徵點匹配，再使用 Lowe 的比例檢測法初步排除錯誤的特徵點配對，對於剩餘的特徵點對我們替換一般經 常使用的隨機抽樣一致法，改採迭代線性最小平方法來尋找合適的單應性矩陣， 並利用此矩陣進行影像拼接所需的座標轉換。影像拼接時，接縫處可能會因為色 差導致色彩不自然，我們使用線性融合調整接縫處的顏色。最後，我們執行一系 列的數值實驗，驗證了所建立的單應性矩陣能夠高效率地完成影像拼接。

The main purpose of this thesis is to replace the random sampling consensus method with a simple iterative linear least-squares method for finding a suitable homogra- phy matrix used in the image stitching process. First, based on the speed-up robust features proposed by Bay, we seek the feature points on two source images and use Lowe’s ratio test to eliminate the wrong feature point pairs preliminarily. Then, to find the suitable homography matrix used in the image stitching, we replace the commonly used random sampling consensus method with a simple iterative linear least-squares method. The resultant homography matrix is then used to perform the coordinate transformation required for stitching. When stitching images, the color of the seams may be different due to chromatic aberration from two source images. Therefore, we employ a linear blending technique to make the color of the seams more natural. Finally, we perform a series of numerical experiments that verify the high performance of the resultant homography matrix for image stitching.

[1] H. Bay, T. Tuytelaars, and L. V. Gool, SURF: Speeded up robust features, In: A. Leonardis, H. Bischof, and A. Pinz (eds), Computer Vision - ECCV 2006, Lecture Notes in Computer Science, Vol. 3951, pp. 404-417, Springer, Berlin, Heidelberg, 2006.
[2] F. C. Crow, Summed-area tables for texture mapping, SIGGRAPH’84: Pro- ceedings of the 11th Annual Conference on Computer Graphics and Interactive Techniques, 1984, pp. 207-212.
[3] M. Brown and D. G. Lowe, Recognising panorama, Proceedings of the 9th In- ternational Conference on Computer Vision, 2003, pp. 1218-1225.
[4] R. C. Bolles and M. A. Fischler, Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography, Communications of the Association for Computing Machinery, 24 (1981), pp. 381-395.
[5] R. C. Gonzalez and R. E. Woods, Digital Image Processing, Fourth Edition, Pearson Education Limited, New York, 2018.
[6] M. Jones and P. Viola, Rapid object detection using a boosted cascade of simple features, Conference on Computer Vision and Pattern Recognition, 2001, pp. 511-518.
[7] D. Kriegman, Homography estimation, Lecture Notes: Computer Vision I (CSE 252A), 2007, pp. 1-3.
[8] T. Lindeberg, Feature detection with automatic scale selection, International Journal of Computer Vision, 30 (1998), pp. 79-116.
[9] T. Lindeberg and B. M. ter Haar Romeny, Linear scale-space I: basic theory, In: B. M. ter Haar Romeny (eds), Geometry-Driven Diffusion in Computer Vi- sion, Computational Imaging and Vision, Vol. 1, pp. 1-38, Springer, Dordrecht, Netherlands, 1994.
36
[10] K. Mikolajczyk and C. Schmid, Indexing based on scale invariant interest points, International Conference on Computer Vision, Vol. 1, pp. 525-531, 2001.
[11] C.-Y. Wei, A combined method of image stitching and contrast enhancement for constructing panoramas, Master Thesis, National Central University, 2021.
[12] https://read01.com/zh-tw/QLayA.html#.Yrgd1y9CYUs
[13] https://www.instagram.com/shiuan_0629/