本論文的主要研究是在於邊界追蹤 (edge tracking) 與小波轉換 (wavelet transform) 的結合。利用轉換後多重解析度的高頻影像 (HL，LH，HH bands) 做邊界追蹤，可以有效解決邊模糊化及不連續的問題。追蹤演算法可分為下列五個步驟：影像前置處理、高頻資訊擷取、小波分解，起始點擷取及多重解析度追蹤。
前置處理包含影像去除雜訊及強化對比，目的是要盡量減少分支 (branch) 的發生。高頻資訊的擷取是利用 Haar 做小波轉換，去除 LL 波段，再反轉換得到梯度影像 (gradient image)。之後的邊界追蹤就是以梯度影像為主，配合小波分解多重解析度的特性來做。但光靠上述特性並不能保証追蹤的正確性，所以論文中還會提出其他追蹤方向的判斷式 (criteria) ，結合多方面的考量以獲得更好的追蹤效果。
在實做方面，除了採用影像處理常用的影像範例外，我們試著處理醫學影像，發現對於器官組織的輪廓擷取，仍有不錯的效果。

A novel edge detection approach based on the wavelet transformation and edge tracking is proposed. Wavelet transform provides multiresolution representation of images for robust tracking. The proposed approach consists of four modules: (i) image preprocessing, (ii) starting point extraction and purgation for tracking, (iii) wavelet decomposition, and (iv) multiresolution edge tracking. Image preprocessing includes band-pass and high-pass filterings. The band-pass filter is used to remove noise and eliminate regular and violent textures; the high-pass filtering is used to generate a gradient image for multiresolution tracking.
The starting points may affect the performance and tracking results. The results is dependent on applications; thus the starting points are extracted from the gradient image by specifying threshold values or using default values for a specified application as user’s desire. Before tracking, the gradient image is decomposed twice by a wavelet transform to generate two coarser-scaled gradient images for multiresolution tracking.
The proposed approach doesn’t need post-processing. Experiments with several commonly used images and medical images are conducted to evaluate performance of the proposed approach. Based on the human visual inspection, the proposed approach always generates the proper results.

[1]Beltrán, J. R., J. García-Lucía, and J. Navarro, “Edge detection and classification using Mallat’s wavelet,” in IEEE Proc. 1st Int. Conf. Image Processing, Austin, Texas, Nov.13-16, 1994, pp.293-297.
[2]Busch, C., “Wavelet based texture segmentation of multi-modal tomographic images,” Comput. & Graphics, Vol.21, No.3, pp.347-358, 1997.
[3]Chang, T. and C.-C. Jay Kuo, “Texture analysis and classification with tree-structured wavelet transform,” IEEE Trans. on Image Processing, Vol.2, No.4, pp.429-440, 1993.
[4]Chun, S. LU, P. C. Chung, and C. F. Chen, “Unsupervised texture segmentation via wavelet transform,” Pattern Recognition, Vol.30, No.5, pp.729-742, 1997.
[5]Climent, J., A. Grau, J. Aranda, and A. B. Martínez, “A high precision operator to determine edge orientation,” in IEE Int. Conf. Control, University of Wales Swansea, UK, Sep.1-4, 1998, pp.95-99.
[6]Falcão, A. X., J. K. Udupa, and F. K. Miyazawa, “An ultra-fast user-steered image segmentation paradigm: live wire on the fly,” IEEE Trans. Medical Imaging, Vol.19, No.1, pp.55-62, 2000.
[7]Feng, L., C. Y. Suen, Y. Y. Tang, L. H. Yang, “Edge extraction of images by reconstruction using wavelet decomposition details at different resolution levels,” Int. Journal Pattern Recognition and Artificial Intelligence, Vol.14, No.6, pp.779-793, 2000.
[8]Fukuda, S. and H. Hirosawa, “A wavelet-based texture feature set applied to classification of multifrequency polarimetric SAR images,” IEEE Trans. Geoscience and Remote Sensing, Vol.37, No.5, pp.2282-2286. 1999.
[9]Heijden, F., “Edge and line feature extraction based on covariance models,” IEEE Trans. Pattern Analysis and Machine Intelligence, Vol.17, No.1, pp.16-33, 1995.
[10]Hohne, K.H., B. Pflesser, A. Pommert, M. Riemer, T. Schiemann, R. Schubent, and U. Tiede, “A ‘virtual body’ model for surgical education and rehearsal,” IEEE Computer, Vol.29, pp.25-31, 1996.
[11]Hsieh, J. -W., M.-T. Ko, H.-Y. Mark Liao, and K.-C. Fan, “A new wavelet-based edge detector via constrained optimization,” Image and Vision Computing, Vol.15, No.7, pp.511-528, 1997.
[12]Hsieh, J. -W., Wavelet-based Image Analysis, Ph.D. dissertation, Elect. Eng. Dept., National Central Univ., Chung-li, Taiwan, 1995.
[13]Lorensen, W. E.,” Marching through the visible man,” in IEEE Proc. Conf. Visualization, Atlanta, GA, Oct.29-Nov.3, 1995, pp.368-373.
[14]Mallat, S. G., “A theory for multiresolution signal decomposition: The wavelet representation,” IEEE Trans. Pattern Analysis and Machine Intelligence, Vol.11, No.7, pp.674-693, 1989.
[15]Moon, P. and G. De Jager, “An heuristic graph searching algorithm to find the boundary of apple images,” in IEEE Proc. Communications and Signal Processing, University of Cape Town, South African, Sep.11, 1992, pp.233-238.
[16]Morsy, K. A. and Y. Kanayama, “A new straight edge detection algorithm using direction-controlled edge tracking and random hitting,” IEEE Int. Symposium Computational Intelligence in Robotics and Automation, pp.398-405, 1997.
[17]Pal, N. R. and Pal, S. K., “A review on image segmentation techniques,” Pattern Recognition, Vol.26, No.9, pp.1277-1294, 1993.
[18]Ruaon, M. A. and C. Tomasi, “Color edge detection with the compass operator,” in IEEE Conf. Computer Vision and Pattern Recognition, Fort Collins, Colorado, 1999, pp.160-166.
[19]Sonka, M., V. Hlavac., and R. Boyle, eds., Image Processing, Analysis, and Machine Vision, Thomson Learning, Stamford Connecticut, 1998, Ch.5, pp.148.
[20]Van der Zwet, P. M. J., and Reiber, J. H. C., “A new algorithm to detect irregular coronary boundaries: the gradient field transform,” in IEEE Proc. Computers in Cardiology, Los Alamitos, CA, 1992.
[21]Xu, X. G., T. C. Chao, and A. Bozkurt, “Vip-man: an image-based whole-body adult male model constructed from color photographs of the visible human project for multi-particle Monte Carlo calculations,” Health Physics Society, Vol.78, No.5, pp.476-486, 2000.