臺灣博碩士論文加值系統

在這篇論文中我們提出了一個具有效的克服旋轉及縮放問題混
合型的分類器，這個分類器中包含了最佳特徵空間區域及賈伯濾波器
的多角度頻域演算法，在抗旋轉的部份我們利用Gobor Wavelets(GWs)
的多頻域解析特性來擷取圖形特徵，抗縮放方面則以特徵區域比對的
概念來進行空間相似性的匹配，在本論文中提出了一個針對縮放圖形
的演算法去找出最佳空間區塊當成我們的特徵區塊，並對其特徵空間
區塊計算出四個抗縮放的參數來比對每個空間區塊的相似度，經實驗
結果證明在抗旋轉及抗縮放辨識效果比其他方法來的好。

In this paper an effective classification algorithm based on
significant region and Gabor multi-frequency domain was proposed to overcome the identification problems of the texture image caused by the rotating and scaling. On rotation invariant, Gabor wavelets are used to acquire the texture’s features. On scaling resisting, the similarity between two textures was calculated by comparing the significant region. In our proposed algorithm the wavelet transform and clustering method were used to locate the significant region so that four scaling resistance
parameters were generated from the significant region. Then, support vector machines (SVMs) are introduced as the classifier. Experimental results reveal that this proposed algorithm outperforms existing design algorithms.

目錄
摘要.....................I
Abstract........................................II
誌謝...................................III
目錄..........................................IV
表目錄.......................................VI
圖目錄...................................VII
第一章、緒論..............................1
1.1 前言...............................1
1.2 研究動機目的................2
1.3 論文架構.........................4
第二章、知識背景..........................5
2.1 Sobel operator 二值化處理.............................5
2.2 物件像素統計.............................................................................7
2.3 灰度共生矩陣.............................................................................8
2.4 Gabor Filter.................................................................................11
2.5 支撐向量機器(Support vector machine) ..................................12
第三章、紋理分析與分類方法..............................................................15
3.1 紋理分析...................................................................................15
3.2 Gabor Filter 分析方法..............................................................16
3.3 最佳特徵區塊分析...................................................................22
第四章、特徵攫取方法..........................................................................30
4.1 所有特徵攫取方法...................................................................31
4.2 最大特徵點攫取方法...............................................................35
4.3 最佳特徵點攫取方法...............................................................39
4.4 最佳特徵演算法.......................................................................41
4.5 綜合步驟...................................................................................42
第五章、實驗..........................................................................................46
5.1 SVM 之效能評估......................................................................46
5.2 紋理分類之效能評估................................................................49
第六章、結論..........................................................................................64
參考文獻..................................................................................................65

表目錄
表3.2.1 由圖3.2.4 所得影像特徵比較表..............................................20
表3.3.1 經灰度共生矩陣計算特徵表....................................................25
表3.3.2 其最佳特徵區塊特徵值............................................................27
表(5.1.1)使用DWT 三層濾波評估SVM 和其他分類器比較.............47
表(5.1.2)使用DWFT 四層濾波評估SVM 和其他分類器比較...........48
表(5.1.3)使用DWFT 五層濾波評估SVM 和其他分類器比較...........49
表(5.2.1) 賈波濾波器與其他方法之比較.............................................52
表(5.2.2)抗旋轉與其他方法之比較.......................................................56
表(5.2.3)抗縮放與其他方法之比較.......................................................60

圖目錄
圖(2.1.1)不同強度對比圖形二值化比較.......................................6
圖(2.1.2)Sobel operator 二值化處理................................................6
圖(2.2.1)物件統計後X 和Y 向量...................................................7
圖(2.2.2)二值化圖形像素統計.........................................................8
圖(2.3.1)3×3 行列標號視窗...............................................................9
圖(2.3.2)紋理影像共生矩陣............................................................10
圖(2.3.3)PH、PV、PLD、PRD 共生矩陣..................................11
圖(2.4.1)Gabor 角度紋理訊息.......................................12
圖(2.5.1) ζi 變數(slack variables)示意圖....................................14
圖(3.1.1)紋理分類流程圖...............................................................15
圖(3.2.2)旋轉不變量的Gabor 示意圖........................................17
圖(3.2.3)Gabor Filter 圖示..............................................................18
圖(3.2.4)Gabor 旋轉不變之紋理影像.........................................21
圖(3.3.1)放大後取相同尺寸之圖形............................................23
圖(3.3.2)分別為圖(3.3.1)灰階共生矩陣....................................24
圖(3.3.3)最佳特徵區塊..................................................................26
圖(3.3.4)為最佳特徵區塊分析流程...........................................26
圖(4.1) 最佳特徵區塊詳細流程.................................................29
圖(4.1.1)Sobel operator 二值化處理...........................................30
圖(4.1.2)物件統計後X 和Y 向量..............................................31
圖(4.1.3)二值化圖形像素統計.....................................................32
圖(4.1.4)減去雜訊權重後之訊號如(a)(c) .................................33
圖(4.1.5)為欲攫取特徵影像.............................................................34
圖(4.1.6)攫取到的特徵點.................................................................34
圖(4.2.1)取最大特徵點.....................................................................35
圖(4.2.2)取得最大物件之影像.......................................................35
圖(4.2.3)影像放大局部攫取示意圖..............................................36
圖(4.2.4)等分四區塊示意圖............................................................37
圖(4.2.5)每個區塊所得到最大特徵點..........................................37
圖(4.3.6)找尋鄰近點示意圖............................................................39
圖(4.4.1)特徵區塊長寬.....................................................................42
圖(4.4.2)圖形放大後攫取的最佳特徵.........................................43
圖(5.2.1)tendfold cross-validation..................................................45
圖(5.2.1)Brodatz texture 紋理資料庫中的112 種紋理影……48

[1] T.-S. Lee, “Image representation using 2D Gabor wavelets,” IEEE
Transactions on Pattern Analysis and Machine Intelligence, Vol.18,
No. 10, 1996, Pages 959–971.
[2] L.-L. Shen, J.-I. Zhen, ”Gabor wavelet selection and SVM
classification for object recognition,” Acta Automatica Sinica, Vol.
35, No.4, 2009, Pages 350-355.
[3] C. Koley, P. Purkait, S. Chakravorti ,”Wavelet-aided SVM tool for
impulse fault identification in Transformers,” IEEE Transactions on
Power Delivery, Vol. 21, No. 3, 2006, Pages 1283 - 1290
[4] S. Arivazhagan, L. Ganesan, ”Texture classification using wavelet
transform,” Pattern Recognition Letters, Vol. 24, 2003, Pages 1513 –
1521.
[5] C.-S. Lu, P.-C. Chung, C.-F. Chen, ”Unsupervised texture
segmentation via wavelet transform,” Pattern Recognition, Vol. 30,
Issue 5, May 1997, Pages 729-742.
[6] J.-H. Liu, C-Y Wang, “Research and implementation for texture of
handback skin quantitative analysis based on co-occurrence matrix,”
International Conference on Industrial Mechatronics and Automation
2009 ,Pages 158 – 161.
[7] M. Masotti, R. Campanini,“Texture classification using invariant
ranklet features,” Pattern Recognition Letters Vol. 29, 2008, Pages
1980–1986.
[8] J. Han, K.-K. Ma,”Rotation-invariant and scale-invariant Gabor
features for texture image retrieval,” Image and Vision Computing,
Vol. 25, 2007, Pages 1474–1481.
[9] Hassan, M. Khalid, N.E.A. Ibrahim , ”Evaluation of sobel canny shen
& castan using sample Line histogram method,” Information
Technology, Vol. 3, 2008, Pages 1 – 7.
[10] George M. Haley, B. S. Manjunath,“Rotation-Invariant texture
classification using a complete space-frequency model,” IEEE
Transactions on Image Processing, Vol. 8, No.2. 1999, Pages 255 –
269.
[11] J.K. Kamarainen, V. Kyrki, H. Kalviainen,“Invariance properties of
Gabor filter-based features-overview and applications,” IEEE
Transactions on Image Processing. Vol. 15, No.5, 2006, Pages
1088–1099.
[12] Y. Ma, X. Ding, “Face Detection Based on Hierarchical Support
Vector Machines,” IEEE Proceedings of the 16th International
Conference on Pattern Recognition, Vol. 1, 2002, Pages 222 - 225
[13] M. Unser, D.V.D. Ville,”The pairing of a wavelet basis with a mildly
redundant analysis via subband regression,” IEEE
Transactions on Image Processing, Vol. 17, No.11, 2008, Pages
2040 - 2052
[14] R. Manthalkar, P.K. Biswas,”Rotation and scale invariant texture
classification using gabor wavelets,” Indian Institute of Technology,
Kharagpur , West Bengal, India Pin 721 302.
[15] E. Avci,”An expert system based on wavelet neural network
Adaptive norm entropy for scale invariant texture classification,”
Expert Systems with Applications, Vol. 32, 2007, Pages 919-926.
[16] G.-H. Liu, J.-Y. Yang ,”Image retrieval based on the texton
co-occurrence matrix,”Pattern Recognition Vol. 41, 2008, Pages
3521-3527.
[17] R. Manthalkar , P.K. Biswas, B.N. Chatterji ,”Rotation and scale
invariant texture features using discrete wavelet packet transform,”
Pattern Recognition Letters Vol. 24, 2003, Pages 2455-2462.
[18] S. Li, J.-T. Kwok, H. Zhu, Y. Wang ,”Texture classi cation using the
support vector machines,” Pattern Recognition, Vol. 36, 2003, Pages
2883 - 2893.
[19] H. Xiong, T. Zhang, Y.-S. Moon, ”A translation and scale-invariant
adaptive wavelet transform,” IEEE Transactions on Image
Processing, Vol. 9, No. 12, 2000, Pages 2100 - 2108
[20] K. Muneeswaran, L. Ganesan, S. Arumugam, K. R. Soundar,”
Texture classification with combined rotation and scale invariant
wavelet features,” Pattern Recognition Vol. 38, 2005, Pages 1495
-1506.
[21] T. Celik, T. Tjahjad,”Multiscale texture classification using dual-tree
complex wavelet transform,” Pattern Recognition Letters, Vol. 30
2009, Pages 331–339.
[22] S. Yue, Q. Wang, R. Zhao ,” Robust wide baseline point matching
based on scale invariant feature descriptor,” Chinese Journal of
Aeronautics Vol. 22, 2009, Pages70-74.