本研究提出一個快速的投影片影像比對的方法。在不需要辨識出投影片內容，如每一個文字或圖片的情形之下，進行投影片影像之比對。比對之方法分為三個階段：首先，找出每一張投影片影像前景物件，如文字、圖片等物件在該影像中所在的位置和範圍，將前景與背景分離(foreground-background separation)。第二階段則利用橢圓曲線線段，針對整張影像進行取樣，並將每一張影像取樣的結果以一組Lattice特徵向量表示。最後，從任兩張影像間的Lattice特徵進行信賴值的計算，以信賴值最高的投影片影像作為比對結果。
本研究利用41組教學簡報，共1980張投影片影像，進行比對實驗，比對從數位攝影機拍攝該簡報撥放的過程後，所匯出的影片檔案。利用所提出之比對演算法，在僅利用11條取樣曲線的情況下，即可達到 97% 以上之比對正確率。不僅解省計算的時間和記憶體使用的空間，對因鏡頭所造成影像傾斜或變形的問題亦能有效地克服。

This thesis proposes a fast image matching algorithm for matching up video taped slide presentations against original slides. Our approach uses only local features and neither text-figure segmentation nor text-recognition is needed. The algorithm consists of three steps. First, background and foreground images are segmented using motion occlusion zones detection technique. Second, local features are sampled from virtual vertical elliptic lines on every slide images and lattice feature vectors are computed. Finally, the lattice feature vectors are matched and the least square error is computed for each matching images. Experiment was conducted using forty-one sets of lecturing slides and video tapes, which consisted of 1980 slides all together. The experimental results show that with only 11 elliptic sampling lines and 64 feature samples per line, a 97% precision rate can be attained. The average computation time for matching a set of slides is less than one second. It could not only reduce the cost of computed time and memory space, but also overcome the distorted problems from camera lens effectively.

[1] Ackermann, F., High precision digital image correlation, In: Proceedings
of 39th Photogrammetric Week. Institute of Photogrammetry, University
of Stuttgart, pp. 231-243, 1983
[2] Alfredo F., Giovanni G., Best-match retrieval for structured images,
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 23(7),
July 2001
[3] Boyer, K.L., Kak, A.C, Structural stereopsis for 3-d vision, IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 10, pp. 144-
166, 1988
[4] Chuang G., Li-de W., Structural matching of multiresolution for
stereo vision, IEEE, 1990
[5] De Vel , O., Aeberhard S., Object recognition using random imagelines,
Image and Vision Computing, vol. 18(3), pp. 193-198, February 2000
[6] El Ansari, M., Masmoudi, L., Radouane, L., A new region matching
method for stereoscopic images, Pattern Recognition Letters, vol. 21(4),
pp. 283-294, April 2000
[7] Gimel’Farb, Georgy L., Jain, Anil K., On retrieving textured images
from an image database, Pattern Recognition, vol. 29(9), pp. 1461-1483,
September 1996
59
[8] Greenfeld, J.S., Schulte-Hinsken, S., Muller, W., A strategy for
automated stereo model orientation using a feature-based matching
procedure, Proc. ACSM-ASPRS Annual Conv., pp. 131-140, 1991
[9] Johann W., Axel P., Layout and analysis: finding text, titles, and
photos in Digital Images of Newspaper Pages, IEEE, 1993
[10] Jun Wei H., Lei G., A shape-based image retrieval method using
salient edges, Signal Processing: Image communication, vol. 18, pp. 141-
156, February 2003
[11] Lemmens, M., A survey on stereo matching techniques, Int. Arch.
Photogramm. Remote Sensing, vol. 27, pp. 11-23, 1988
[12] Li, M., High precision relative orientation using feature based techniques,
Int. Arch. Photogramm. Remote Sensing, vol. 27, pp. 456-465, 1988
[13] Meier T., Ngan K. N., Video segmentation for content-based coding,
IEEE Transactions on Circuits and Systems for Video Technology, vol.
9, pp. 1190-1203, December 1999
[14] Minerva, Y., Boon-Lock, Y., Segmentation of video by clustering
and graphy analysis, Computer Vision and Image Understanding, vol.
71(1), July 1998
[15] Nagy, G., A prottype document image analysis system for technical
journals, IEEE Transactions on Computers, vol. 25, pp. 10-22, 1992
60
[16] Pass G., Zabih R., Histogram refinement for content-based image
retrieval, Workshop on Applications of Computer Vision, pp 96-102, 1996
[17] Radu H., Thomas S., Structural Matching for Stereo Vision, IEEE,
1998
[18] Richard C. W., Edwin R. H. (ed.)., Structural matching by discrete
relaxation, IEEE Transactions on Pattern Analysis and Machine Intelligence,
vol. 19, 1997
[19] Sriprakash S., Sudeep S., An approximate algorithm for structural
matching of images, IEEE, 1998
[20] Strouthopoulos C., Papamarkos N., Atsalakis A.E., Text extraction
in complex color documents, Pattern Recognition, vol. 35(8) pp. 1743-
1758, August 2002
[21] Strouthopoulos C., Papamarkos N., Chamzas C., PLA using RLSA
and a neural network, Engineering Application of Artificial Intelligence,
vol. 12, pp. 119-138, 1999
[22] Strouthopoulos C., Papamarkos N., Text identification for document
image analysis using a neural network, Image and Vision Computing,
vol. 16(12-13), pp. 879-896, August 1998
[23] Yi-Long C., Hiromasa N., Image region correspondence by structural
similarity, IEEE, 1992
61
[24] Younian W., Principles and applications of structural image matching,
ISPRS Journal of Photogrammetry & Remote Sensing, vol. 53, pp. 154-
165, 1998
[25] Yan L., Wen G., Feng W., Sprite generation for frame-based video
coding, IEEE, 2001
[26] Whal, F., Wong, K., Casey, R., Block segmentation and text extraction
in mixed text/image documents, Computer Vision Graphics
and Image Processing, vol. 2, pp. 327-352, 1982
[27] Wang, D., Shihari, S.N., Classification of newspaper image blocks
using texture analysis, Computer Vision Graphics and Image Processing,
vol. 47, pp. 327-352, 1989
[28] Ze-Nian, Li., Xiang, Z., Mark S. D., Illumination-invariant image
retrieval and video segmentation, Pattern Recognition, vol. 35(8), pp.
1687-1704, August 2002
[29] Ze-Nian, Li., Xiang, Z., Mark S. D., Spatial-temporal joint probability
images for video segmentation, Pattern Recognition, vol. 35(9),
pp. 1847-1867, September 2002