簡易檢索 / 詳目顯示

回結果列表
研究生： 黃元瑞
Yun-Jui Huang
論文名稱： 藉由統計顏色空間的分割進行影像特徵的正規化
Image Feature Normalization by Statistical Color Space Partitioning
指導教授： 吳怡樂
Yi-Leh Wu
口試委員: 唐政元
Cheng-Yuan Tang
林彥君
Yen-Chun Lin
鄧惟中
Wei-Chung Teng
學位類別： 碩士
Master
系所名稱： 電資學院 - 資訊工程系
Department of Computer Science and Information Engineering
論文出版年： 2007
畢業學年度： 95
語文別： 英文
論文頁數： 45
中文關鍵詞： 影像註解 、BIC特徵 、Hilbert-scan 、顏色空間分割
外文關鍵詞： image annotation, BIC features, Hilbert-scan method, color space partitioning
相關次數： 點閱：210  下載：1 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  •  上一筆

    • 影像註解是將影像加上一組預定的關鍵字，主要使用於視覺資訊的管理。影像註解可以應用於不同的領域例如生物科學、軍事、網路影像分類、搜尋…等。
    邊緣/內部的像素分類特徵(BIC features)是非常簡潔和有效的特徵，因為它可以擷取顏色、形成和紋理資訊。但是BIC特徵對於每個特徵的利用率相差很大。因此我們提出利用Hilbert-scan和重覆分割的方法來改善利用率的問題進而提高影像註解的正確率。最後我們的實驗是註解60個不同種類共6000張的影像來證明我們提出的方法是有效。

    Image annotation refers to the labeling of images with a set of predefined keywords that is mainly used for visual information management. Image annotation can be applied in a variety of domains such as biomedicine, military, web image classification, search, etc. Image annotation employs low-level features to distinguish image contents such as color, shape, texture, etc.
    The Border/Interior pixel Classification (BIC) features [15] are very compact and efficient features that capture color, shape, and texture information. But the BIC features inherit the problem that the differences on utilization rates of each feature are high.
    We propose to employ the Hilbert-Scan method and the Iterative Partitioning Method (IPM) to improve the utilization rates of each feature which results in higher accuracy for image annotation. Finally, we show that our proposed method is effective in annotating 6000 images in 60 categories.

    Contents 1. Introduction 1 2. The Border/Interior pixel Classification (BIC) 4 3. Support Vector Machines (SVMs) 8 4. Color Space Partitioning 9 4.1 Z-scan Method 9 4.2 S-scan Method 10 4.3 Hilbert-scan Method 12 5. Adjusting the utilization rates 15 5.1 The Iterative Partitioning Method (IPM) 15 5.2 The Granularity of the Color Space 17 6. Experiments 23 7. Conclusion and Future Work 31 Appendix 32 References 34 List of Figures 1: Statistical normalization of the utilization rates of each feature. 2 2.1 BIC features 4 2.2: The utilization rate distributions in different color systems. 6 4.1: Use two dimensional (4x4) to illustrate the Z-scan method 9 4.2: The S-scan method 10 4.3: 64 color features by the S-scan method. 11 4.4: The utilization rates of the modified BIC features with the S-scan method. 11 4.5: The Hilbert-scan method 12 4.6: 64 color features by the Hilbert-scan method. 13 4.7: The utilization rates of the modified BIC features with the Hilbert-scan method. 14 5.1: The flowchart of the proposed Iterative Partitioning Method (IPM). 15 5.2: The process of the proposed Iterative Partitioning Method (IPM). 16 5.3: Visualization of color feature 15 by the S-scan method under different granularities. 18 5.4: The proposed IPM with the S-scan method 18 5.5: The utilization rates of the Hilbert-scan method in different granularities. 19 5.6: Visualization of the BIC features by the Hilbert-scan method in different granularities. 20 5.7: The proposed IPM with the Hilbert-scan method. 20 6.1: Accuracy distribution with different SVM parameter settings (the original BIC features). 24 6.2: Accuracy distribution with different SVM parameter settings (the modifies BIC features by the S-scan method). 24 6.3: Accuracy distribution with different SVM parameter settings (the modifies BIC features by the Hilbert-scan method). 24 6.4: Annotation AUC for different image categories. 26 6.5: Annotation positive accuracy of different image categories. 27 6.6: Annotation negative accuracy of different image categories. 28 6.7: Annotation precision for different image categories. 29 6.8: Sample image from the less color rich image categories 30 6.9: Sample image that cannot benefit from the proposed IPM. 30 List of Tables 4.1: The estimation of partitioning by the modified BIC features with the S-scan method. 10 4.2: The estimation of partitioning by the modified BIC features with 13 the Hilbert-scan method. 13 5.1: The utilization rates and color bin size of each feature. 22 6.1: The AUC for different image categories. 26 6.2: Positive annotation accuracy for different image categories. 27 6.3: Negative annotation accuracy for different image categories. 28 6.4: Annotation precision for different image categories. 29

    [1] Y. Bazi and F. Melgani, “Toward an Optimal SVM Classification System for
    Hyperspectral Remote Sensing image”, IEEE Trans. on Geoscience and Remote
    Sensing, Vol. 44, NO. 11, November 2006
    [2] C.-C. Chang, C.-J. Lin, LIBSVM: A library for support vector machines, 2001,
    software available at http://www.csie.ntu.edu.tw/˜cjlin/libsvm
    [3] Y. Chen, J. Z. Wang, “A Region-Based Fuzzy Feature Matching Approach to
    Conten-Based Image Retrieval”, IEEE Trans. on Pattern Analysis and Machine
    Intelligence, Vol. 24, NO. 9, September 2002
    [4] Y. Chen, J. Z. Wang and R. Krovetz, “CLUE：Cluster-Based Retrieval of Image
    by Unsupervised Learning”, IEEE Trans. on Image Processing, Vol. 14 NO.8,
    August 2005
    [5] E. Chang, K. Goh, G.. Sychay, and G. Wu, “CBSA：Content-Based Soft
    Annotation for Multimodal Image Retrieval Using Bayes Point Machines”, IEEE
    Trans. on Circuits and Systems for Video Technology, Vol. 13, NO 1,January
    2003.
    [6] K. S. Goh, E. Y. Chang and B. Li, “Using One-Class and Two-Class SVMs for
    Multiclass image annotation”, IEEE Trans. on Knowledge and Data Engineering,
    Vol. 17, NO 10, October 2005.
    [7] Isabelle Guyon, “An Introduction to Variable and Feature Selection”, Journal of
    Machine Learning Research 3, pp. 1157-1182, 2003
    [8] Li J, Wang J Z, Wiederhold G., “IRM: Integrated region matching for image
    Retrieval”, In Proc. ACM Multimedia 2000, Los Angeles, CA, 2000, pp.147-156.
    [9] S. Kamata, R. O. Eason and Y. Bandou, “A New Algorithm for N-Dimensional
    Hilbert-scan”, IEEE Trans. on Image Processing, Vol.8, NO. 7, July 1999
    [10] C. Y. Lee and V. W. Soo, “The Conflict Detection and Resolution in Knowledge
    Mergin for Image Annotation”, In Proc. ACM Information Processing and
    Management, Vol. 42, pp. 1030-1055, 2006
    [11] F. Jing, M. Li, H. J. Zhang and B. Zhang, “A Unified Framework for Image
    Retrieval Using Keyword and Visual Feature”, IEEE Trans. on Image
    Processing, Vol. 14, NO. 7, July 2005
    [12] K. H. Queh, C. Quek, “MCES：A Novel Monte Evaluative Selection Approach
    for Objective Feature Selections”, IEEE Trans. on Neural Networks, Vol. 18,
    NO. 2, March 2007
    [13] X. Qi, Y. Han, “Incorporating multiple SVMs for automatic image annotation”,
    Pattern Recognition Society, vol. 28, Apr. 2006.

    [14] A. Rakotomamonjy, “Optimizing AUC with Support Vector Machine (SVM)”,
    Proceedings of European Conference on Artificial Intelligence Workshop on
    ROC Curve and AI, Valencia, 2004.
    [15] R. O. Stehling, M. A. Nascimento and A. X. Falcao, “A Compact and Efficient
    Image Retriveal Approach Based on Border/Interior Pixel Classification”,
    in Proc. Int. Conf. ACM CIKM’02, November 4-9 2002.
    [16] D. Tao, X. Tang, X. Li and X. Wu, “Asymmetric Bagging and Random Subspace
    for Support Vector Machines-Based Relevance Feedback in Image Retrieval”,
    IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol. 28, NO. 7, July
    2006
    [17] L. Wang, L. Khan, “Automatic Image Annotation and Retrieval Using Weighted
    Feature Selection”, Multimedia Tools and Applications , Vol. 29, pp. 55 – 71,
    April 2006
    [18] R. Zhang, Z. Zhang, M. Li, W. Y. Ma and H. J. Zhang, “A Probabilistic Semantic
    Model for Image Annotation and Multi-Modal Image Retrieval”, in Proc. Int.
    Conf. Tenth IEEE International Conference on Computer Vision(ICCV’05)

    QR CODE