臺灣博碩士論文加值系統

乳癌為我國婦女癌症發生率第1位，死亡率第4位，發生高峰約在45-69歲之間，每年約有1萬多位婦女罹患乳癌、約2千名婦女死於乳癌，相當於每天約31位婦女被診斷罹患乳癌、6位婦女因乳癌而失去寶貴性命。紅外線影像具備了無輻射、無疼痛感、低成本等優勢，而雙波段紅外線影像已被認為具評估化療療效和早期偵測腫瘤之潛力，可由長時間觀察乳房熱圖譜之間的差異，藉由腫瘤代謝率高和血管新生現象產生的高溫能量，乳房紅外線兼具功能性與解剖性，被視為一個有潛力的醫學輔助工具。
其中多時間點紅外線影像對位技術是一個必要的步驟，由於受測者每次拍攝紅外線的姿勢皆不同，也無法將標記點長期黏貼在受測者身上，使得不同時間點的紅外線影像對位時，不存在任何外在可靠的標記點，為克服此困難，本論文發展自動化多時間點乳房紅外線影像對位演算法。
本論文演算法以2011年Brox等人提出的Large Displcement Optical Flow(LDOF)為架構發展新的特徵點匹配策略，適用於乳房紅外線影像；而新的特徵點匹配策略分成新增與刪除匹配兩個階段，此策略整合2012年Ma等人提出的方法，發展兩階段迭代得到更好的匹配點對，其中初始匹配使用Shape Context描述的最小距離當做基準；特徵點偵測以Hessian矩陣為基礎發展之方法，可在平坦區域找到特徵點，也可簡單的與Shape Context做結合。本論文結果比較2017年Lee等人的方法與Free-Form Deformation(FFD)方法的邊緣，本論文的邊緣重疊度較好，表示影像的結構有對位好。特徵點偵測與特徵點匹配也較均勻，而高溫組織的量化曲線也與化療曲線一致，可用於之後的化療追蹤。

Breast cancer is the first occurrence of cancer in our country. the fourth death rate, Most cases occur in patients 45-69 years of age.Each year more than 10,000 women suffer from breast cancer, and approximately 2,000 women die of breast cancer. Every day about 31 women are diagnosed with breast cancer, and six women lose their precious lives to breast cancer. Infrared imaging has many advantages including, no radiation, no pain, low cost.Additionally, dual-band infrared images, which have been considered to assess the efficacy of chemotherapy and early detection of tumor potential, can be observed over a long time to detect changes in the breast heat map. Through high metabolic rate and angiogenesis of high temperature energy, breast infrared, both functional and anatomical, is considered a potential medical aid. Which is more longuitudinal infrared image registrartion technology is a necessary step, because the subjects each time shooting infrared posture are different, can not be long-term paste points in the subject, so that different time points of the infrared image alignment , There is no external reliable mark point, In order to overcome this difficulty, this paper has developed a Representative feature detection on flat region and iterative coherent spatial mapping for longitudinal IR breast image registration.
In this paper, the algorithm is based on Brox et al.'s Large Displacement Optical Flow (LDOF) and is proposed for the development of new feature point matching strategy for breast infrared imaging. The new feature matching strategy is divided into adding and eliminating feature pair stages, This strategy integrates the method proposed by Ma et al. in 2012 to develop a two-stage iteration in order to obtain a better match point pair.The initial match uses the minimum distance described by the Shape Context as the benchmark.The feature point detection, based on the Hessian matrix method, can be found in the flat area feature points, but also in the simple combination with the Shape Context. The results of this paper compare the method of Lee et al. and the Free-Form Deformation (FFD) method in 2017. Under the proposed method, the overlap between the boundary of the results of the propsed method and the boundary of the oringal image is more accurate than the FFD method, indicating that the structure is properly registered. . Feature point detection and feature point matching is also more uniform, and high temperature tissue quantification curve,consistent with the chemotherapy curve, can be used for subsequent follow-up chemotherapy.

摘要 4
Abstract 5
目錄 7
圖目錄 9
表目錄 12
第一章 緒論 1
1.1前言 1
1.2動機與目的 2
1.3文獻探討 3
1.3.1特徵點偵測 3
1.3.2特徵點描述 4
1.3.3特徵點匹配 5
1.3.4估計轉換函數 6
1.4論文架構 8
第二章 理論背景 9
2.1影像骨架化 9
2.2取骨架影像的交點 10
2.3 Harris角點響應 10
2.4取角點響應的局部極大值 12
2.5 Shape Context 12
2.6雙立方內插法 13
2.7 MR-CSM 13
2.8 LDOF原文 15
2.9雙波段熱圖譜分離(Dual Spectrum Heat Pattern Separation) 19
第三章 研究材料與方法 21
3.1研究材料：紅外線光譜照射系統 21
3.2研究方法 22
3.2.1前處理 22
3.2.2特徵點偵測 23
3.2.3特徵點描述 26
3.2.4特徵點匹配 26
3.2.5估計轉換函數 29
第四章 結果與討論 31
4.1特徵點稀疏度 31
4.2特徵點匹配均勻度 35
4.3邊緣驗證 42
4.4化療驗證 53
4.5本論文演算法限制 55
第五章 結論 59
參考文獻 60

[1]衛生福利部, 103年死因統計結果分析. 2015. p. 12-13.
[2]Balu-Maestro, C., et al. "Imaging in evaluation of response to neoadjuvant breast cancer treatment benefits of MRI." Breast cancer research and treatment 72.2 (2002): 145-152.
[3]Partridge, Savannah C., et al. "MRI measurements of breast tumor volume predict response to neoadjuvant chemotherapy and recurrence-free survival." American Journal of Roentgenology 184.6 (2005): 1774-1781.
[4]Zylberberg, B., et al. "Initial chemoimmunotherapy in inflammatory carcinoma of the breast." Cancer 49.8 (1982): 1537-1543.
[5]Lee, Chia-Yen, et al. "Spatiotemporal sharpening of sub-pixel super-resolution by means of two infrared spectrum cameras for early cancer detection." Proc. of SPIE Vol. 6979 (2008): 69790R-1.
[6]Lee, Chia-Yen, et al. "Evaluation of dual-spectrum IR spectrogram system on invasive ductal carcinoma (IDC) breast cancer." Biomedical Engineering: Applications, Basis and Communications 23.06 (2011): 427-433.
[7]Lee, Chia-Yen, et al. "Automatic Marker-free Longitudinal Infrared Image Registration by Shape Context Based Matching and Competitive Winner-guided Optimal Corresponding." Scientific Reports 7 (2017): 39834.
[8]Rosten, Edward, and Tom Drummond. "Machine learning for high-speed corner detection." Computer vision–ECCV 2006 (2006): 430-443.
[9]Smith, Stephen M., and J. Michael Brady. "SUSAN—A new approach to low level image processing." International journal of computer vision 23.1 (1997): 45-78.
[10]Mokhtarian, Farzin, and Riku Suomela. "Robust image corner detection through curvature scale space." IEEE Transactions on Pattern Analysis and Machine Intelligence 20.12 (1998): 1376-1381.
[11]Lindeberg, Tony. "Feature detection with automatic scale selection." International journal of computer vision 30.2 (1998): 79-116.
[12]Matas, Jiri, et al. "Robust wide-baseline stereo from maximally stable extremal regions." Image and vision computing 22.10 (2004): 761-767.
[13]Miao, Zhenwei, and Xudong Jiang. "Interest point detection using rank order LoG filter." Pattern Recognition 46.11 (2013): 2890-2901.
[14]Miao, Zhenwei, et al. "Contrast Invariant Interest Point Detection by Zero-Norm LoG Filter." IEEE Transactions on Image Processing 25.1 (2016): 331-342.
[15]Su, Ran, et al. "Junction detection for linear structures based on Hessian, correlation and shape information." Pattern Recognition 45.10 (2012): 3695-3706.
[16]Lowe, David G. "Object recognition from local scale-invariant features." Computer vision, 1999. The proceedings of the seventh IEEE international conference on. Vol. 2 (1999): 1150-1157.
[17]Bay, Herbert, et al. "Surf: Speeded up robust features." Computer vision–ECCV 2006 (2006): 404-417.
[18]Belongie, Serge, et al. "Shape context: A new descriptor for shape matching and object recognition." Nips. Vol. 2 (2000): No. 2000
[19]Scott, Guy L., and H. Christopher Longuet-Higgins. "An algorithm for associating the features of two images." Proceedings of the Royal Society of London B: Biological Sciences 244.1309 (1991): 21-26.
[20]Shapiro, Larry S., and J. Michael Brady. "Feature-based correspondence: an eigenvector approach." Image and vision computing 10.5 (1992): 283-288.
[21]Carcassoni, Marco, and Edwin R. Hancock. "Spectral correspondence for point pattern matching." Pattern Recognition 36.1 (2003): 193-204.
[22]Leordeanu, et al. "A spectral technique for correspondence problems using pairwise constraints." Computer Vision, 2005. ICCV 2005. Tenth IEEE International Conference on. Vol. 2. IEEE (2005): 1482-1489.
[23]Zaslavskiy, Mikhail, et al. "A path following algorithm for the graph matching problem." IEEE Transactions on Pattern Analysis and Machine Intelligence 31.12 (2009): 2227-2242.
[24]Zhou, Feng, and Fernando De la Torre. "Factorized graph matching." IEEE transactions on pattern analysis and machine intelligence 38.9 (2016): 1774-1789.
[25]Duchenne, Olivier, et al. "A tensor-based algorithm for high-order graph matching." IEEE transactions on pattern analysis and machine intelligence 33.12 (2011): 2383-2395.
[26]Fischler, Martin A., and Robert C. Bolles. "Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography." Communications of the ACM 24.6 (1981): 381-395.
[27]Ma, Jiayi, et al. "Mismatch removal via coherent spatial mapping." Image Processing (ICIP), 2012 19th IEEE International Conference on. IEEE (2012): 1-4.
[28]Reddy, B. Srinivasa, and Biswanath N. Chatterji. "An FFT-based technique for translation, rotation, and scale-invariant image registration." IEEE transactions on image processing 5.8 (1996): 1266-1271.
[29]Chui, Haili, and Anand Rangarajan. "A new point matching algorithm for non-rigid registration." Computer Vision and Image Understanding 89.2 (2003): 114-141.
[30]Thirion, J-P. "Image matching as a diffusion process: an analogy with Maxwell's demons." Medical image analysis 2.3 (1998): 243-260.
[31]Cachier, Pascal, et al. "Iconic feature based nonrigid registration: the PASHA algorithm." Computer vision and image understanding 89.2 (2003): 272-298.
[32]Vercauteren, Tom, et al. "Diffeomorphic demons: Efficient non-parametric image registration." NeuroImage 45.1 (2009): S61-S72.
[33]Rueckert, Daniel, et al. "Nonrigid registration using free-form deformations: application to breast MR images." IEEE transactions on medical imaging 18.8 (1999): 712-721.
[34]Liu, Ce, et al. "Sift flow: Dense correspondence across scenes and its applications." IEEE transactions on pattern analysis and machine intelligence 33.5 (2011): 978-994.
[35]Brox, Thomas, and Jitendra Malik. "Large displacement optical flow: descriptor matching in variational motion estimation." IEEE transactions on pattern analysis and machine intelligence 33.3 (2011): 500-513.
[36]Arbelaez, Pablo, et al. "From contours to regions: An empirical evaluation." Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. IEEE (2009): 2294-2301