利用影像技術進行空間資料之獲取一直是空間資訊技術中核心工作之一,例 如使用立體對影像重建三維模型、影像校齊、影像鑲嵌、影像檢索…等工作中,
共同的特點是必需從影像中提取適當數量之特徵點作為相關處理之依據。以立體 對影像重建三維模型為例,傳統上一般均使用區域基礎(像元灰度值)作為影像匹 配運算之基礎,可以達到次像元等級高精度量測之目的,但其中仍存在有不易解 決之困難,如兩影像間解析度(比例尺)、攝影傾角及光照條件之差異及影像中之 均調(Homogeneous)區等均可能造成影像匹配之失敗。
由學者David Lowe 於 2004 年所提出的 SIFT 基本演算法,實驗證明對於影 像的尺度、旋轉角度及光照等差異具有良好不敏感性(Robust)與不變性(Invariant) 特性,在影像點特徵提取(Point Feature Extraction)及影像匹配(Image Matching)中 可提供重要的應用,可以用於解決目前立體對影像自動匹配的問題,再依功能之 需要,修改SIFT 演算法中之步驟,也可有效應用於其他相關的工作。本研究利 用SIFT 演算法進行立體對影像匹配與遙測影像檢索之應用,已獲得具體之成果。
一、立體對影像匹配應用
1. SIFT 演算法屬基於特徵的描述與特徵匹配之方法,由實驗得知對紋理相 似之均調區進行匹配仍具有良好之效果。
2. 本研究對 SIFT+RANSAC 基本運算之精度驗證,由三組實驗結果顯示,
匹配點總均方根誤差分別為0.56 像素、0.63 像素與 0.54 像素,且匹配品 質小於1 個像素分別為 95.36%、90.49 與 97.06%,最多僅有 2 個點之誤 差大於三個像素,證明SIFT+RANSAC 匹配具有良好精度。
3. 本研究修改 SIFT 演算法匹配方式,於演算法後使用核線約束並搭配 RANSAC 進行相關實驗,由三組實驗結果顯示,匹配總均方根誤差分別 由0.56 像素降為 0.37 像素、0.63 像素降為 0.35 與 0.54 像素降為 0.27 像 素,匹配點位分別由474 點提升至 647 點、305 點提升至 318 點與 204 點 提升至 247 點,甚至沒有大於 2 像素之點位出現,實驗結果證明加入線 約制除了提高匹配可靠度之外,還可提高匹配之精度,甚至可以增加匹 配數量。
4. 本研究於 SIFT 演算法匹配中利用核線幾何束方法,讓自動匹配過程更正 確快速,與提高點位可靠度。利用此觀念配合其它仿射不變之角點檢測 法,並利用基礎矩陣之轉換關係進行核線匹配,以補足SIFT 演算法匹配 點位非角點之問題。
二、影像檢索應用
本研究另外提出SIFT 演算法於影像檢索之應用,在本實驗中利用遙測影像 為檢索應用之目標影像,利用不同尺度及旋轉影像進行測試,經SIFT 與 RANSAC 演算法運算,並延伸用於多目標索引之檢測應用,獲得具體之成果。
(1) SIFT 演算法於航拍影像檢索之應用,實驗證明該演算法用於同時其影像 檢索方面有極佳之效果。
(2) SIFT 演算法於影像區快速定位之應用,實驗證明利用該演算法計算影像 重疊區之紋理明顯(如建物、道路)及均調區(如水體、植被等)區域仍然有 良好之匹配效果。
(3) 本研究修改 SIFT 演算法用於多目標檢索,並利用象棋進行相關測試,實 驗結果證明利用SIFT 用於多目標及旋轉特徵檢索亦有極佳之效果。
三、建議:
本研究成功的援引SIFT 基本演算法於立體對影像匹配與影像檢索之應用,
證明該方法具有相當之應用潛力,未來可進一步研究之方向包括:
(1) 結合其它角點之偵測方法,補足 SIFT 演算法匹配點位非角點之問題。
(2) 演算法與除錯機制之改善,例如搜尋策略、匹配判斷法則等,可加速特 徵點提取之運算效率並提高匹配成功率。
(3) 其他影像處理工作之應用,例如影像校齊、影像快速鑲嵌、變遷偵測等。
誌謝:
本研究感謝海軍大氣海洋局在行政與實務經驗方面的指導與協助。感謝國 科會補助本計畫(計劃編號:NSC 97-2623-7-151-001-D)之執行。
計畫成果自評:
本計畫執行與原計畫預定內容及進度均相符,已達成預期之目標,本研究 在基於內容之影像檢索上提出新的方法,以SIFT演算法為基礎之特徵點選取與描 述進行影像之匹配與檢索,實驗例證明其效果相當良好。本研究結果正進一步整 理並發表於國際學術期刊。
參 考 文 獻
[ 1] David G. Lowe, 2004, "Distinctive image features from scale-invariant keypoints," International Journal of Computer Vision, 60, 2 (2004), pp. 91-110 [ 2] David G. Lowe, 1999, "Object recognition from local scale-invariant
features," International Conference on Computer Vision, Corfu, Greece (September 1999), pp. 1150-1157
[ 3] Loncomilla, P.[Patricio], Ruiz-del-Solar, J.[Javier], 2005, "Improving
SIFT-Based Object Recognition for Robot Applications, " LECTURE NOTES IN COMPUTER SCIENCE, 2005, NUMB 3617, pages 1084-1092.
[ 4] F. Rothganger, S. Lazebnik, C. Schmid, and J. Ponce., 2006, "3D Object
Modeling and Recognition from Photographs and Image Sequences",Toward Category-Level Object Recognition, Springer-Verlag Lecture Notes in Computer Science vol. 4170, J. Ponce, M. Hebert, C. Schmid, and A. Zisserman (eds.), 2006, pp. 105-126.
[ 5] Yan Ke, Rahul Sukthankar, Larry Huston, 2004, "Efficient Near-duplicate Detection and Sub-image Retrieval," Proceedings of ACM International Conference on Multimedia (MM), 2004, IRP-TR-04-07.
[ 6] Stoettinger, J. Hanbury, A. Sebe, N. Gevers, T., 2007, "Colour Interest Points for Image Retrieval," Image Processing, 2007. ICIP 2007. IEEE International Conference on, 16 2007-Oct. 19 2007 I - 169-I – 172.
[ 7] H. Su, D. Crookes, A. Bouridane, and M. Gueham, 2007, "Shoeprint Image Retrieval Based on Local Image Features," School of EEE and CS, Queen’s University Belfast, pages 387-392.
[ 8] Moravec, http://www.cim.mcgill.ca/~dparks/CornerDetector/mainMoravec.htm [ 9] M. Brown and D. G. Lowe, 2003, "Recognising Panoramas," 13-16 Oct. 2003
Computer Vision, 2003. Proceedings. Ninth IEEE International Conference on, 1218-1225 vol.2.
[10] Autopano-SIFT, http://user.cs.tu-berlin.de/~nowozin/autopano-sift/
[11] hugin - Panorama photo stitcher, http://hugin.sourceforge.net/
[12] Stephen Se, David Lowe, and James J. Little.,2000, "Vision-based mobile robot localization and mapping using scale-invariant features,". submitted to ICRA-00, 2000. http://citeseer.ist.psu.edu/se01visionbased.html
[13] Dr. Lisa Spencer, 2006 , "Writing Video Applications", Opencv Learning Opencv, http://www.cs.ucf.edu/~lspencer/vid_app.pdf
[14] Chunrong Yuan, 2006, "Markerless Pose Tracking for Augmented Reality,".
ISVC (1) 2006: 721-730
[15] K. Mikolajczyk and C. Schmid.,2003, "A performance evaluation of local descriptors,". In Proceedings of Computer Vision and Pattern Recognition, June 2003.
[16] Y. Ke and R. Sukthankar,2004 , "PCA-SIFT: A More Distinctive
Representation for Local Image Descriptors," Computer Vision and Pattern Recognition, 2004
[17] Herbert Bay, Tinne Tuytelaars and Luc Van Gool, 2006, "SURF: Speeded Up Robust Features", Proceedings of the 9th European Conference on Computer Vision, Springer LNCS volume 3951, part 1, pp 404--417, 2006.
[18] Moravec H P., 1979, "Visual mapping by a robot rover", Proc. of 6th IJCAI, 1979:589-600
[19] Harris C, Stephens M., 1988, "A combined corner and edge detector", Proc.
Alvey Vision Conf., 1988: 147-151
[20] J.J. Koenderink, 1984, "The structure of images", Biological Cybernetics,50:363-396, 1984.
[21] T. Lindeberg, 1994, "Scale-space theory: A basic tool for analysing structures at different scales", Journal of Applied Statistics, 21(2):224-270, 1994.2(60):91-110
[22] Martin A. Fischler and Robert C. Bolles, June 1981, "Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography". Comm. of the ACM 24: 381–395.
[23] Z.Zhang, R. Deriche, O. Faugeras, and Q. Luong., 1995, "A Robust Technique for Matching two Uncalibrated Images through the Recovery of Unknown Epipolar Geometry", Artificial Intelligence, 78:87—119, 1995.
[24] A. W. Moore, Nov. 1990, "Efficient Memory-Based Learning for Robot Control", University of Cambridge Computer Laboratory, University of Cambridge
[25] Wikipedia , 2008 , "RANSAC", http://en.wikipedia.org/wiki/RANSAC [26] Beis, J. S. and Lowe, D. G., 2003, "Shape indexing using approximate
nearest-neighbor search in high-dimensional spaces", In <EM>
Conferenceon Computer Vision and Pattern Recognition (CVPR)</EM>
(2003),pp. 1000--1006.
[27] O. Faugeras and Q.-T. Luong, 1996 , "The fundamental matrix : Theory Algorithms, and Stability Analysis", Int'l J. Computer Vision, 17, pp. 43-45, 1996.
[28] Zhang Z Y, 1998, "Determining the Epipolar geometry and its uncertainty:a review", International Journal of Computer Vision, 1998 , 27(2):161-198.
[29] Brigham Anderson, Andrew Moore, Dan Pelleg, Alex Gray, Bob Nichols, Andy
Connolly,2008, "Scalable Data Mining",The Auton Lab, Carnegie Mellon University, www.autonlab.org
[30] M.A. Fischler and R.C. Bolles. "Random sample consensus: A paradigm for model Fitting with applications to image analysis and automated
cartography". Communications of the ACM, 24(6):381–395, 1981.
[31] Google Inc. Picasa, 2008, http://picasa.google.com/
[32] K. Mikolajczyk and C., 2004, "Schmid, Scale and Affine invariant interest point detectors"., In IJC V 60(1):63-86, 2004.
[33] J.Matas, O. Chum, M. Urban, and T, 2002, "Pajdla, Robust wide baseline stereo from maximally stable extremal regions", In BMVC p. 384-393, 2002.
[34] T.Tuytelaars and L. Van Gool, 2004, "Matching widely separated views based on affine invariant regions" . In IJCV 59(1):61-85, 2004.
[35] T. Kadir, A. Zisserman, and M. Brady, 2004, "An affine invariant salient region detector". In ECCV p. 404-416, 2004.
[36] Edward Rosten and Tom Drummond, 2005, "Authors: Edward Rosten and Tom Drummond", IEEE International Conference on Computer Vision 1508--1511
[37] ward Rosten and Tom Drummond, 2006, "Machine learning for high-speed corner detection", European Conference on Computer Vision,430—443.
[38] FAST Corner Detection,http://mi.eng.cam.ac.uk/~er258/work/fast.html [39] Paul Viola, Michael Jones, 2001, "Rapid Object Detection using a Boosted
Cascade of Simple Features", ACCEPTED CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION 2001
[40] 周明全,耿國華,韋娜,2007, "基於內容圖像檢索技術", 清華大學出版社, 7/1/2007
[41] 吳銳航 , 李紹滋 , 鄒豐美, 2008, "基於 SIFT 特徵的圖像檢索," 計算機應 用 研究 App lication Research of Computers, 100123695 (2008) 0220478204.
[42] 李家欣,2005, “多視點輔助定位系統”, 國立臺灣科技大學機械工程系碩士論 文.
[43] 張洪剛 陳光 郭軍, 2006, "圖像處理與識別", 北京郵電大學出版社, pp1-pp20
[44] 簡大淵, 2002, "內視鏡影像序列之自動校正、重構與病灶量測", 碩士論 文, 國立成功大學資訊工程學系碩士班