QR碼是二維條碼的一種，在生活中已被廣泛地使用，QR來自英文「Quick Response」的縮寫，即快速反應，搭配現今的手機的應用程式掃描，即可快速解碼。由於其使用範圍廣泛，因此如何使QR碼的使用更加便利顯然是相當重要的任務。
　　在掃描QR碼時，鏡頭需要正對著QR碼且只能掃描單張QR碼，若同時有多張QR碼，則需要各別對正拍攝，亦或是QR碼貼於彎曲的表面，則會造成QR碼的失真，導致無法偵測及解碼。本論文主考慮為位於柱面上之QR碼，透過射影幾何的概念校正失真現象，使得QR碼的使用不受限於平面且拍攝角度更加自由。
　　由於QR碼是由黑白兩色所組成，本論文嘗試將彩色區塊濾除，搭配尋找連接區域(connected component)，將圖片中的QR碼影像找出來。平面上之失真QR碼透過投影轉換(projective transform)校正視角的傾斜(tilt)，即可將QR碼回正，位於柱面上之失真QR碼則需再透過水平拉伸(horizontal stretch)校正柱面上的水平伸縮變化，最後搭配交比(cross ratio)特定線段比例不變性質取得採樣點，即可將QR碼內的每一個模點讀出，進而得到失真QR碼的資訊。依據上述方法進行校正，我們可以成功地解碼許多失真的QR碼。

　　QR code is a type of two dimensional barcode that has been widely used in our daily life. QR is the abbreviation of the “Quick Response,” which means instantaneous response. When it is scanned with applications installed in smartphones, quick decoding can be achieved. Therefore, how to improve the usability of the QR code is an important task.
　　During the scanning of a QR code, the lens has to directly face the QR code and only a single QR code can be scanned. If there are multiple QR codes, then each of them must be detected and separated. Moreover, if the QR code is attached on a curved surface, the QR codes will be distorted, and then quiet often they cannot be successfully decoded. This thesis focuses on correcting the distortion of QR codes on cylindrical surfaces based on the concept of projective geometry such that the use of QR codes is no longer limited to a flat surface, and allowing the filming angle to be more flexible.
　　Given that the QR code comprises two colors, black and white, this thesis tried to filter the color block and combine the connected component with searching to sort out QR code images from photos. A distorted QR code on a flat surface can be rectified using projective transform to correct the tilted vision, whereas a distorted QR code on a cylindrical surface has to further undergo horizontal stretch to correct the horizontal stretch variation on the cylindrical surface. Finally, through extracting sampling points based on the invariance property of the specific segment ratio of cross ratio, every module of the QR code can be read and information in the distorted QR code can be extracted. Based on the correction method mentioned above, numerous distorted QR codes can be successfully decoded.

[1] Denso Wave, “Information technology - automatic identification and data capture techniques - bar code symbology - QR code,” ISO/IEC 18004, 2000.
[2] Irving S. Reed and Gustave Solomon, “Polynomial Codes Over Certain Finite Fields,” Journal of the Society of Industrial and Applied Mathematics (SIAM), pp. 200-204, 1960.
[3] Robin Hartshorne, Foundations of Projective Geometry, 2nd edition, Ishi Press, 2009.
[4] David A. Brannan, Matthew F. Esplen and Jeremy J. Gray, Geometry, 5th edition, Cambridge Press, 2003.
[5] Rafael C. Gonzalez, and Richard E. Woods, Digital Image Processing, 3rd edition, Prentice-Hall, 2008.
[6] Nobuyuki Otsu “A threshold selection method from gray-level histograms,” Proc. sactions of the ASME - Journal of Basic Engineering Vol. 82: pp. 35-45, 1960 IEEE Trans. Sys., Man., Cyber. Vol.9, pp. 62-66, 1979.
[7] Jiejing Zhou and Yunfei Liu, Peng Li, “Research on binarization of QR code Image,” Proc. IEEE Int. Conf. Multimedia Technology, Vol.1, pp. 1-4, 2010.
[8] Kirsti Andersen, The Geometry of an Art, Springer-Verlag, 1st edition, 2007.
[9] Irwin Sobel, “History and Definition of the Sobel Operator,” 2014.
[10] 陳怡靜, “貼於平面及柱面上之帶有大標誌的QR碼之編解碼,” 國立臺灣科技大學電子工程研究所, 2015.
[11] 林鈺勛, “使用圓錐切割法將柱面上的QR碼影像回正,” 國立臺灣科技大學電子工程研究所, 2014.
[12] Bernard Kolman, David R. Hill, Elementary Linear Algebra with Applications, 9th edition, Prentice Hall, 2007.