陳志宏臺灣大學:生醫電子與資訊學研究所李韋均Li, Wei-ChunWei-ChunLi2010-05-262018-07-052010-05-262018-07-052009U0001-1308200913401500http://ntur.lib.ntu.edu.tw//handle/246246/184199全景合成的技術是為了讓人們在看圖片時可以有一個更大的視野，這樣的技巧可以被運用在相當多的領域中，而如何讓他更快，更簡單，夠精確，就變成了一個很重要的議題。這篇論文中，我們介紹了影像縫合的技術，又稱為全景合成。並且實作出來且將他們應用在各種不同類型的影像中，更進一步探討他們在生醫影像這個領域的應用，驗證我們的方法是有效且快速的。The technique of image stitching is made up for a better range of a picture’s view for humans. It can be applied to many specific fields. How to make it fast, simple and accurate thus becoming a very important issue. n this thesis, we introduce the image stitching technique also called panoramas. Then implement it and apply it to different types of images. Especially on the field of biomedical images to verify that our method is effective and efficient.Chapter 1 Introduction 1.1 Motivation 1.2 History of panoramas 2hapter 2 Related Work 4.1 Pre-step 4.2 Warping: Cylindrical projection 8hapter 3 Feature Point-Based Method 11.1 SIFT: Scale Invariant Feature Transform 11.1.1 Features 11.1.2 Scale-Space Extrema Detection 12.1.3 Keypoint Localization 13.1.4 Interpolation of Nearby Data for Accurate Position 14.1.5 Discarding Low-Contrast Keypoints 15.1.6 Eliminating edge responses 15.1.7 Orientation assignment 15.2 Kd Tree 16.3 RANSAC: RANdom SAmple Consensus 18hapter 4 Our Proposed Method 21.1 RGB Color Channel Balance 21.2 Error Matrix 24.3 Stitch Line Finding 26.2 Seamless Blending 30hapter 5 Experiment Results 33.1 Experiment Environment 33.2 Results 34.2.1 Medical Images 34.2.2 Normal Daily Photos 54.3 Comparision 68hapter 6 Conclusion and Future Work 72.1 Conclusion 72.2 Future Work 72eferences…………………………………… ……………………73application/pdf4839502 bytesapplication/pdfen-US全景合成生醫影像panoramabiomedical imagethesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/184199/1/ntu-98-R96945008-1.pdf