薛克民臺灣大學：數學研究所林保同Lin, Pao-TungPao-TungLin2007-11-282018-06-282007-11-282018-06-282007http://ntur.lib.ntu.edu.tw//handle/246246/59446Image registration, i.e. the way to find a suitable transformation so that an image becomes similar to another one, is one of the fundamental tasks in many image processing problems. Our goal in this thesis is to study two different numerical image registration approaches, namely, the diffusion and curvature registrations, and show results obtained using each of these methods for some sample problems.1 Introduction 7 1.1 Problem formulation . . . . . . . . . . . . . . . 7 1.2 Thesis Outline . . . . . . . . . .. . . . . . . . . 9 2 The Diffusion Registration 11 2.1 Mathematical Model . . . . . . . . . . . . . . . . 11 2.2 Numerical Method . . . . . . . . . . . . . . . . . 12 2.3 Numerical Examples . . . . . . . . . . . . . . . . 15 3 The Curvature Registration 29 3.1 Mathematical Model . . . . . . . . . . . . . . . . 29 3.2 Numerical Method . . . . . . . . . . . . . . . . 30 3.3 Numerical Examples . . . . . . . . . . . . . . . . 32 4 Conclusion 47638428 bytesapplication/pdfen-US影像註冊擴散註冊法曲度註冊法Euler-Lagrange方程式疊加性分離法離散餘弦轉換Image registrationdiffusion registrationcurvature registrationEuler-Lagrange equationadditive operator splittingdiscrete cosine transformationsthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/59446/1/ntu-96-R94221014-1.pdf