近年來，電腦視覺的技術被廣泛應用於解決各種影像處理的問題，例如：偵測、追蹤、立體匹配、光流、分類及資料檢索等。除了研發各種方法與改善技術之外，特徵描述子在電腦視覺中為不可或缺的基石，若能為某種應用設計出一套完善的特徵敘述子，便可有效的處理資料並提升正確率。我們提出了一套新的特徵敘述子敘述方式，可以應用於結構照片的影像匹配、分類與資料檢索之問題，並可有效的提升其辨別正確率。在本方法中使用了形狀上下文的概念架構，分析特徵點與其他相近的特徵點之間的關係，計算其特徵點分佈狀況之長條圖，利用此長條圖來產生新的敘述子。透過這樣的設計，不只考慮特徵點本身在影像上的資訊與分佈，更融合了地域資訊，將鄰近的特徵點彼此之間的關係一併敘述。在實驗結果中，我們所提出的特徵點敘述方式擁有一定程度的移動不變性、旋轉不變性、縮放不變性、光影不變性及視角不變性，在基準的資料庫中表現相當優異，更能幫助提升結構影像檢索重排序的正確率。

關鍵字：特徵敘述子、結構影像、形狀上下文結構、影像匹配、影像檢索重排序。

We present a method to improve interest point descriptor on structure images that do not contain too many repeated patterns. In the proposed approach, interest point features are combined together based on shape context structure to maximize the feature descriptor capability based on extracted interest point information. The method involves constructing a shape structure for combining the geometry information into a single feature vector, generating histogram of the interest point distributions in the neighborhood, and incorporating the statistics information into the feature descriptor. At run time, features are transformed into a set of shape context to generate the feature descriptor. We apply the improved interest point descriptor to structure image matching and retrieval problems. Evaluation on a set of distortion images from a benchmark dataset shows that the proposed method outperforms the state-of-the-art feature descriptor methods. The proposed descriptor incorporates additional local geometry information with the shape context structure, and yields improvement in an image retrieval re-ranking system.

Additional Key Words and Phrases: Feature descriptor, structure image, shape context structure, image matching, and image retrieval re-ranking.

Chapter 1. Introduction 1
1.1 Background 1
1.2 Problem Description 2
1.3 Main Contribution 4
1.4 Thesis Organization 4
Chapter 2. Previous Works 6
Chapter 3. Proposed Method 10
3.1 Problem Statement 10
3.2 Improved Interest Point Descriptor 11
3.2.1 Shape Context Structure 12
3.2.2 Feature Histogram 15
3.2.3 Feature Filter 17
3.3 Image Retrieval Re-ranking System 20
Chapter 4. Experimental Results 24
4.1 Performance Evaluation of Local Descriptor 24
4.2 Image Retrieval Re-ranking 43
4.2.1 Holidays Dataset Retrieval Re-ranking 44
4.2.2 Flickr 11k Dataset Retrieval Re-ranking 47
Chapter 5. Conclusion 50
References 51

[1]. D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International journal of computer vision, vol. 60, no. 2, pp. 91–110, 2004.
[2]. H. Bay, T. Tuytelaars, and L. Van Gool, “Surf: Speeded up robust features,” in Computer Vision–ECCV 2006. Springer Berlin Heidelberg, 2006, pp. 404–417.
[3]. S. Belongie, J. Malik, and J. Puzicha, “Shape matching and object recognition using shape contexts,” Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 24, no. 4, pp. 509–522, 2002.
[4]. E. Rublee, V. Rabaud, K. Konolige, and G. Bradski, “Orb: an efficient alternative to sift or surf,” in Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011, pp. 2564–2571.
[5]. N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” in Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on, vol. 1. IEEE, 2005, pp. 886–893.
[6]. T. Ojala, M. Pietikainen, and T. Maenpaa, “Multiresolution gray-scale and rotation invariant texture classification with local binary patterns,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 7, pp. 971–987, 2002.
[7]. Oliva and A. Torralba, “Modeling the shape of the scene: A holistic representation of the spatial envelope,” International journal of computer vision, vol. 42, no. 3, pp. 145–175, 2001.
[8]. J. Wu and J. M. Rehg, “Centrist: A visual descriptor for scene categorization,” Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 33, no. 8, pp. 1489–1501, 2011.
[9]. K. Mikolajczyk and C. Schmid, “A performance evaluation of local descriptors,” Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 27, no. 10, pp. 1615–1630, 2005.
[10]. C. Theriault, N. Thome, and M. Cord, “Dynamic scene classification: Learning motion descriptors with slow features analysis,” in Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on. IEEE, 2013, pp. 2603–2610.
[11]. F. von Hundelshausen and R. Sukthankar, “D-nets: Beyond patchbased image descriptors,” in Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on. IEEE, 2012, pp. 2941–2948.
[12]. H. Jegou, M. Douze, and C. Schmid, “Hamming embedding and weak geometric consistency for large scale image search,” in ECCV (1), 2008, pp. 304–317.
[13]. Y.-H. Kuo, H.-T. Lin, W.-H. Cheng, Y.-H. Yang, and W. H. Hsu, “Unsupervised auxiliary visual words discovery for large-scale image object retrieval,” in Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on. IEEE, 2011, pp. 905–912.
[14]. I. T. Jolliffe, Principal Component Analysis. Springer-Verlag, 1986.