摘要

本論文研究是以影像切割方法當中的距離正規化水平集演化(DRLSE)方法當作基礎，希望能針對該演算方法不夠完善與便利的部份做加強與改善，並建構了一套以此改良方法針對人類指骨的自動化處理圈選輪廓流程。
傳統的DRLSE方法需要手動設定輪廓線的初始位置，然而該位置是否接近目標邊緣將對於最終抓取的結果有著不小的影響。因此本論文提出以邊緣檢測函數距離邊界越近數值越小的性質，抓取到影像中物體的大略邊界位置，再配合一系列的影像處理流程使該結果成為更清晰明確的一條輪廓線。接著再以該輪廓線作為Level Set function初始位置進行傳統的DRLSE疊代運算，使輪廓線能更為平滑地貼近真實邊界。
除此之外，於本論文中也針對指骨影像獨有的影像特性，使上述依據邊緣檢測函數改良後的DRLSE方法能自動化處理資料庫中的所有影像，省去人工一一調整設定最佳化參數的成本。
研究結果以測試一般影像、單一指骨影像、自動化處理指骨資料庫等三個部份來呈現，並分別計算運算時間、疊代次數及ME、RFAE、MHD、EMM、NU及Mean Errors等六種錯誤測量指標，以最終結果輪廓位置圖與統計數值表示。
根據實驗結果，本論文提出的方法具有三個顯著的優點：第一是不需要人工依照每張影像中物體的形狀、大小、位置、數量等資訊來設定不同的DRLSE較佳初始位置；第二是整體進行的運算時間與DRLSE疊代次數能夠非常的精簡，可有效避免浪費運算資源；第三是經過本論文為指骨影像設計之自動化方法尋找到的良好邊界位置，確實可以幫助後來的DRLSE運算提升效果，降低錯誤率。

Abstract

Distance Regularized Level Set Evolution, which is abbreviated to DRLSE, is an algorithm for image segmentation. In this thesis, we aim to improve the traditional DRLSE algorithm using edge indicator function, and develop a standard process using the proposed algorithm to separate phalanx from its background automatically.
The key factor in the performance of DRLSE algorithm is the initial level contour. The closer the initial level contour and the outline of the target, the better result we get. The traditional DRLSE algorithm sets the initial level contour manually, which can be a tedious and inefficient work. In order to achieve better result, we propose a method using the predefined edge indicator function and several of image processing techniques to set the initial level contour automatically before traditional DRLSE evolution.
Besides the general image segmentation problem, this study focuses on the problem of phalanx segmentation in particular. We proposed a standard process to separate phalanx from its background automatically by using the proposed modified DRLSE algorithm. The parameters setting of the algorithm are simplified, without the need for manual setting.
The test images include some general images and a set of phalanx images. The result images, the execution time, the number of DRLSE iterations, and some measure of errors such as ME, RFAE, MHD, EMM, NU and Mean Errors are shown as experimental results. According to the experimental results, the proposed algorithm has three advantages over the traditional algorithm. First, the proposed algorithm sets the initial level contour automatically. It's more efficient and more convenient than the traditional one. Second, the whole computation time and the number of DRLSE iterations can be reduced drastically, which can avoid the waste of computing resources effectively. Third, the experimental results of the automated method for phalanx images have proved that the proposed algorithm works better. It can really help to enhance the effect of the DRLSE operation and reduces the error rates.

目錄

摘要 i
Abstract ii
誌謝 iv
目錄 v
圖目錄 vii
表目錄 x
第一章 簡介 1
1.1 前言 1
1.2 研究動機 1
1.3 文獻回顧 2
1.3.1 參數式主動輪廓模型 2
1.3.2 GVF Snake 4
1.3.3 幾何式主動輪廓模型 6
1.4 研究目的 8
1.5 論文架構 8
第二章 理論與方法 10
2.1 傳統的Level Set方法 10
2.1.1 Level Set演化方式 10
2.1.2 重新初始化 13
2.2 距離正規化水平集演化(DRLSE) 15
2.2.1 能量函數 15
2.2.2 Distance Regularized Level Set Evolution 19
2.2.3 兩階段式收斂的Level Set方法 22
2.3 結合圖型邊界地圖資訊的DRLSE方法並全自動化圈選指骨輪廓 24
2.3.1 以影像的g值做Level Set處理 24
2.3.2 以影像處理技術選取適當的輪廓線 30
2.3.3 加上傳統的DRLSE 34
2.3.4 使DRLSE疊代次數自動化 36
2.3.5 使g值threshold選擇自動化 39
2.3.6 全自動化處理指骨資料庫 42
2.4 錯誤測量指標 45
2.4.1 人工繪製Ground Truth 46
2.4.2 ME 47
2.4.3 RFAE 47
2.4.4 MHD 48
2.4.5 EMM 48
2.4.6 NU 49
2.4.7 Mean Errors 49
第三章 實驗結果 50
3.1 資料準備 50
3.2 實驗結果 51
3.2.1 一般影像結果 52
3.2.2 指骨影像結果 62
3.2.3 指骨影像自動化處理統計結果 70
第四章 結論 79
4.1 結論 79
4.2 未來展望 80
參考文獻 81

參考文獻

[1] C. Gonzalez and E. Woods, “Digital Image Processing, 3rd Edition” pp. 689-787, 2007
[2] R. Ohlander, K. Price, D. R. Reddy, "Picture Segmentation Using a Recursive Region Splitting Method". Computer Graphics and Image Processing 8 (3): 313–333 1978
[3] R. Adams and L. Bischof, "Seeded region growing," Pattern Analysis and Machine Intelligence, vol. 16, no. 6, pp. 641-647, 6 1994.
[4] J. Canny, "A Computational Approach To Edge Detection," IEEE Trans. Pattern Analysis and Machine Intelligence, 8:679-714, 1986
[5] M. Kass, A. Witkin and D. Terzopoulos, "Snake: active contour models," International journal of computer vision, vol. 1, no. 4, pp. 321-331, 1 1988.
[6] S. Osher and J. A. Sethian, "Fronts propagating with curvature-dependent speed: Algorithms based on Hamilton-Jacobi formulations," Journal of computational physics, vol. 79, no. 1, pp. 12-49, 11 1988.
[7] S. C. Zhu and A. Yuille, "Region competition: unifying snakes, region growing, and Bayes/MDL for multiband image segmentation," Pattern Analysis and Machine Intelligence, vol. 18, no. 9, pp. 884-900, 9 1996.
[8] C. Xu and J. L. Prince, "Snakes, shapes, and gradient vector flow," Image Processing, vol. 7, no. 3, pp. 359-369, 3 1998.
[9] L. D. Cohen, "On active contour models and balloons," CVGIP: Image Understanding, vol. 53, no. 2, pp. 211-218, 3 1991.
[10] W. D J and S. M, "A fast algorithm for active contours and curvature estimation," CVGIP:Image Understanding, vol. 55,no. 1, pp. 14-26, 1992
[11] L. D. Cohen and I. Cohen, "Finite-element methods for active contour models and balloons for 2-D and 3-D images," Pattern Analysis and Machine Intelligence, vol. 15, no. 11, pp. 1131-1147, 11 1993.
[12] C. Xu and J. L. Prince, "Gradient vector flow: A new external force for snakes," Computer Vision and Pattern Recognition, pp. 66-71, 1 1997.
[13] V. Caselles, F. Catte, T. Coll, and F. Dibos, "A geometric model for active contours in image processing," Numerische Mathematic. 66: 1-31, 1993.
[14] R. Malladi, J. A. Sethian and B. C. Vemuri, "Shape modeling with front propagation: A level set approach," Pattern Analysis and Machine Intelligence, vol. 17, no. 2, pp. 158-175, 2 1995.
[15] N. Paragios and R. Deriche, “A PDE-based Level Set approach for Detection and Tracking of moving objects,” ICCV,1139-1145, 1998
[16] T. Chan and L. Vese, “Active contours without edges,” IEEE Trans. On Image Processing, 10: 266-277, 2001
[17] D. Enright, S. Marschner and R. Fedkiw, “Animation and Rendering of Complex Water Surfaces,” SIGGRAPH, ACM TOG 21: 736-744, 2002
[18] S. Osher and N. Paragios, “Geometric Level Set Methods in Imaging Vision and Graphics,” Springer Verlag, 2003
[19] A. Lefohn, J. Kniss, C. Hansen and R. Whitaker, “A Streaming Narrow-Band Algorithm: Interactive Deformation and Visualization of Level Sets,” IEEE Transactions on Visualization and Computer Graphics, 10(40):422-433 , 2004
[20] H. Yamauchi, S. Gumhold, R. Zayer and H.P. Seidel, “Mesh segmentation driven by Gaussian curvature,” Source Visual Computer, 21(8-10 Special Issue SI): 659-668, 2005
[21] Y. Shi and W.C. Karl, “Real-time tracking using level sets,” Computer Vision and Pattern Recognition, 2:20-25, 2005
[22] V. Caselles, R. Kimmel and G. Sapiro, "Geodesic active contours," International journal of computer vision, vol. 22, no. 1, pp. 61-79, 2 1997.
[23] K. Saddiqi, Y.B. Lauziere, A. Tannenbaum, and S.W. Zucker. Area and length minimizing flows for shape segmentation [J]. IEEE transactions on Image Processing, 7(3): 433-443, 1998
[24] Bertalmio M, Sapiro G, Randall G. Region tracking on level-set methods. IEEE Transactions on Medical Imaging, 18(5): 448-451, 1999
[25] Paragios N, Deriche R. Geodesic active contours and level sets for the detection and tracking of moving objects. IEEE Transactions on Pattern Analysis and Machine Intelligence, 22(3): 266-280, 2000
[26] Samon C, Blanc-Feraud L, Aubert G, Josiane Z. Level set model for image classification. International Journal of computer Vision, 40(3): 187-197, 2000
[27] Masouri A-R, Sirivong B, Konrad J. Multiple motion segmentation with level sets. Proceedings of SPIE, Vol.3974, pp.584-595, 2000
[28] J. A. Sethian, Level Set Methods and Fast Marching Methods: evolving interfaces in computational geometry, fluid mechanics, computer vision, and materials science, Cambridge Univ. Press, 1999.
[29] S. Osher and R. Fedkiw, Level set methods and dynamic implicit surfaces, Springer, 2002.
[30] C. Li, C. Xu and M. D. Fox, "Level set evolution without re-initialization: a new variational formulation," Computer Vision and Pattern Recognition, vol. 1, pp. 430-436, 6 2005.
[31] C. Li, C. Xu, C. Gui and M. D. Fox, "Distance Regularized Level Set Evolution and Its Application to Image Segmentation," Image Processing, vol. 19, no. 12, pp. 3234-3254, 12 2010.
[32] A Brief Introduction to Level Set, Narrow Band, and Fast Marching Methods
[33] Fatemi E ,Sussman M, "An efficient interface preserving level-set re-distancing algorithm and its application to interfacial incompressible fluid flow[J]" SIAM J Sci Statist Comput ,158(1):36-58, 1995.
[34] Sussman M ,Smereka P, "Axi-symmetric free boundary problems[J]" J Fluid Mech , 341(1):269-294, 1997.
[35] Sussman M ,Almgren A S ,Bell J B , et al, "An adaptive level-set approach for incompressible two-phase flows[J]" J Comput Phys ,148(1):81-124, 1999.
[36] C. Xu, A. Yezzi, Jr and J. L. Prince, "On the relationship between parametric and geometric active contours," Signals, Systems and Computers, vol. 1, pp. 483-489, 11 2000.
[37] J. Gomes and O. Faugeras, "Reconciling distance functions and Level Sets", J. Visiual Communic. and Imag. Representation, vol. 11, pp. 209-223, 2000.
[38] L. Evans Partial Differential Equations, Providence: American Mathematical Society, 1998.
[39] 林益弘, “使用兩階段收斂的Level Set,” 私人通訊.
[40] M. Sezgin, “Survey over image thresholding techniques and quantitative performance evaluation,” Journal of Electronic imaging, vol. 13, no. 1, pp. 146-168, 6 2004.
[41] C.C. Liu and P.C. Chung, “objects extraction algorithm of color image using adaptive forecasting filters created automatically,” International Journal of Innovative Computing, Information and Control, vol. 7, no. 10, pp. 5771-5757, 10 2010.