本論文提出一個即時從真實場景的連續視訊，自動建立全景影像的編輯系統。使用者可直接手持攝影機進行拍攝，不需另外使用特別的硬體裝置如魚眼鏡頭、腳架等，即可訊速產生所需的全景影像。假設，將原本已壓縮過的5.72MB影片（大小為320x240），在CPU 為PII-233的機器上執行，只需22秒就能產生完整的全景影像，而其大小只有123KB。
由於連續視訊中的資料量非常多，本論文利用階層式來達到即時的估算位移量，同時也修正了可能因為階層式而造成影像定位錯誤的情形。而估算位移量的方法，一般是利用明暗度的差值來估算，但是攝影機光圈的變化可能會造成比對上的失誤，於是，本論文提出可以解決此問題的方法，使得估算位移量能更準確的適合更多的情況。同時，攝影機光圈的變化也可能造成結果的全景影像中，會有整塊較黑的區域出現，無法只利用調整整個全景影像的亮度或比對來克服；因此本論文利用動態的方式，調整每個畫面的明暗度，以解決因光圈的快速變化，造成全景影像出現整塊較黑區域的問題。一旦使用者拍攝環狀場景超過360度時，本系統可自動尋找拍攝360度中的重複場景，同時能接合頭尾影像，使整個系統能真正達到完全的自動接合。
接下來，可以將產生的全景影像輸出成虛擬實境瀏覽器所支援的格式，或可輸出成VRML97的檔案格式；如此一來，使得原本已壓縮過的5.72MB影片，變成只有123KB左右的全景影像，大小只有原來的三十四分之一，非常適合透過網際網路的傳輸，而參觀者還可以將視窗前後左右移動，瀏覽整個全景影像。

An authoring system is proposed to construct panoramic images of real-world scenes from video clips automatically in real-time. Instead of using other special hardware devices such as fish-eye lens or tripod, a user can just pan a hand-held camcorder simply. Therefore, a user can construct a panoramic image more convenient and more rapidly. Our implementation of constructing a panoramic image from CIF format video clip takes 22 seconds on a Pentium II-266 PC.
There are so much data in video clips, and so a hierarchical method is proposed to get transformation between successive frames. The traditional method uses intensity errors to get the above transformation, but that may cause error due to change in aperture. To make the system more robust, we propose a method to solve the problem. At the same time, the change in aperture may generate a dark area in the panoramic image, which can not be overcome by adjusting brightness and contrast. Therefore, we use adaptive brightness adjustment of every frame to solve this problem. Once a user captures a scene over 360 degrees, our system will detect the same area automatically, and stitch them together. As a result, this system can not only generate popular image-based virtual reality data format, but also output the panoramic images in VRML97 format. Compared with the original video data, our proposed method reduces the data size from 5MB to 150KB, and is very suitable for transmission through Internet.

第一章 緒論4
1.1 概說4
1.2 動機與目標5
1.3 系統概觀7
第二章 相關研究10
2.1 利用照片來接合環狀的全景影像10
2.2 利用照片來接合球狀的全景影像12
2.3 利用連續視訊來接合環狀的全景影像13
第三章 位移量的估算16
3.1 估算位移量的基本概念16
3.2 幾種估算位移量的方法18
3.3 階層式的估算位移量19
第四章 接合影像與輸出29
4.1 消除縫隙的接合畫面29
4.2 接合頭尾成為環狀全景影像35
4.3 輸出全景影像40
第五章 實驗結果與討論41
5.1 實驗結果41
5.2 討論60
第六章 結論與未來展望61
6.1 結論61
6.2 未來展望61
參考文獻66
附錄一 Video VR 編輯器使用者手冊69

[1] James R. Bergen, P. Anandan, Keith J. Hanna, and Rajesh Hingorani. Hierarchical Model-Based Motion Estimation. In Second European Conference on Computer Vision (ECCV''92), Santa Margherita Liguere, Italy, May 1992, pages 237-252.
[2] Bernd Jahne, Digital Image Processing : Concepts, Algorithms, and Scientific Applications. Springer-Verlag, Berlin, New York, 1992.
[3] Shenchang Eric Chen. QuickTime VR - An Image-Based Approach to Virtual Environment Navigation. Computer Graphics (SIGGRAPH''95), Los Angeles, California, USA, Aug. 1995, pages 29-38.
[4] Ju-Wei Chen, Shu-Cheng Huang, Tse Cheng, Der-Lor Way, and Li-Sheng Shen. Building Full View Spherical Panoramas via an Image-Based Approach. International Computer Symposium Workshop on Computer Graphics and Virtual Reality, N.C.K.U., Tainan, Taiwan, Dec. 1998, pages 112-119.
[5] Richard Szeliski and Heung Yeung Shum. Creating Full View Panoramic Image Mosaics and Environment Maps. In Computer Graphics Proceedings, Annual Conference Series, Proc. SIGGRAPH''97, Los Angeles, California, USA, Aug. 1997, pages 251-258.
[6] Richard Szeliski, Video Mosaics for Virtual Environments. IEEE Computer Graphics and Applications, Mar. 1996, pages 22-30.
[7] Zhengyou Zheng, RachidDeriche, Olivier Faugeras, and Quang-Tuan Luong. A robust technique for matching two uncalibrated images through the recovery of the unknown epipolar geometry. Artificial Intelligence 78, 1995, Pages 87-119.
[8] Richard I. Hartley. In Defence of the Eight-Point Algorithm. In IEEE Transactions on Pattern Analysis and Machine Intellugence, Vol. 19, N0. 6, June 1997, pages 580-593.
[9] Ding-Yuen Chen, Ming Ouhyoung, VideoVR: A real-time system for automatically constructing panoramic images from video clips. IFIP 1998 Workshop on Modelling and Motion Capture Techniques for Virtual Environments (CAPTECH98), Geneva, Switzland, Nov. 1998.
[10] Harpreet S. Sawhney, Steve Hsu, and R. Kumar. Robust video mosaicing through topology inference and local to global alignment. In Proc. European Conference on Computer Vision, 1998, pages 103-119.
[11] Shmuel Peleg and Joshua Herman, Panoramic Mosaicing with VideoBrush. DARPA Image Understanding Workshop, May 1997, pages.261-264.
[12] Shmuel Peleg and Moshe Ben-Ezra, Stereo Panorama with a Single Camera. To appear in CVPR''99
[13] P.J. Burt and E.H. Adelson, A Multiresolution Spline with Application to Image Mosaics. ACM Transaction on Graphics, Vol. 2, No. 4, Oct. 1983, pages 217-235.
[14] J.J. Su, Z.Y. Zhuang, S.D. Lee, J.R. Wu, and M. Ouhyoung, Photo VR: An Image-Based Panoramic View Environment Walk-Through System. Proc. of IEEE International Conference on Consumer Electronic (ICCE''97), pp. 224-225, Chicago, USA 1997.
[15] 李協德碩士論文, 台灣大學資訊工程研究所, 1998.
[16] 洪志坤碩士論文, 台灣大學資訊工程研究所, 1999.
[17] Video Brush, http://www.videobrush.com
[18] QuickTime VR, http://qtvr.quicktime.apple.com
[19] IPIX, http://www.ipix.com
[20] Be Here, http://www.behere.com
[21] Smooth Movie, http://www.smoothmove.com
[22] Live Picture, http://www.livepicture.com
[23] OmniCam, http: //www.cs.columbia.edu/CAVE/omnicam
[24] PhotoVR, http://www.cmlab.csie.ntu.edu.tw/cml/g/pvr/PVR.html
[25] IBM, http://www.software.ibm.com/net.media/solutions/panoramix
[26] VRML97, The Virtual Reality Modeling Language, ISO/IEC 14772-1:1997
[27] J. Meehan, Panoramic Photography. Watson-Guptill, 1990.
[28] 廖敏華, 印刷攝影學. 五洲出版社, 1985.
[29] 林會承, 清末鹿港街鎮結構. 境與象出版社, 1991.