本論文提出一個在點對點網路架構(Peer-to-Peer, P2P)上進行內容式影像檢索(Content-Based Image Retrieval, CBIR)的系統運作方法。在搜尋引擎的設計上，本論文提出以多詢例多特徵(Multi-Instance with Multiple Features,MIMF)檢索的方法，發展適用於P2P-CBIR架構上的通用檢索機制。所提出的點對點影像檢索系統相較於傳統主從式(server-client)或者中央化(centralized)的檢索系統架構，在檢索效率以及運算效能都能獲得較好的結果。本論文提出的P2P-CBIR系統，由檢索端點(query peer)開始，由上層至下層以樹狀結構形式建立檢索端點連線，收到檢索訊息的每個子端點會在其局部現狀資料庫中進行影像相似性排名，並傳送高相似度的檢索影像至上層端點，同時並接收來至下層端點所傳來的高相似影像。所有因應此一檢索訊息之工作端點都同時進行這兩個程序，如此整個P2P-CBIR所有之工作端點同時不斷的進行檢索與傳送，可以在層層檢索與傳送的過程中，不斷的過濾檢索影像相似度，在不增加平均運算與傳輸負擔的情況下，可以讓檢索端點在最短的時間接收到最好的檢索結果。另外，我們在此一P2P-CBIR系統架構上，提出系統調適與更新的方法 (system reconfiguration)，以因應網路端點資料庫為時變(time variant)特性的實際狀況。藉由在端點閒置時，啟動更新程序，首先更新現時局部資料庫的特徵參數，並進一步更新端點間相似性連結的參數，如此可以讓網路搜尋引擎在較為正確的系統參數下進行檢索程序。相較於之前的煙火式(firework)和廣度優先搜尋(Depth First Search, DFS)式兩種檢索方法，在檢索效能(回取率/檢索範圍) 以及平均頻寬負載皆能獲得改善。模擬結果顯示，本論文所提出之reconfigurable 機制，相較於原先未盡行更新之P2P-CBIR系統方法，回取率(recall-rate)可以提高7%至38%倍。而本系統對於選擇多特徵描述影像資訊執行影像檢索時，能適時調整多特徵的權重關係，以充分描述影像內容資訊，相較於選擇較少的特徵描述影像，可有效提升回取率9.46%。

A peer-to-peer content-based image retrieval system (P2P-CBIR) that provides scalable retrieval function has been proposed. The proposed P2P-CBIR network image search engine, which adopted multi-instance query with multi-feature types (MIMF), effectively reduced average network traffics while maintaining high retrieval accuracy on the query peer. The scalable retrieval function can adaptively control the query scope and progressively refine the accuracy of retrieved results. It improves the query efficiency (recall-rate/query-scope) by effectively combining the: (1) forwarding query message (forward phase) to reduce the query scope and; (2) transmitting retrieval results (backward phase) that activated peers keep filtering high similarity images on the linking-path toward the query peer. Experiments showed that the query efficiency of the scalable retrieval approach is better than previous methods, i.e., firework query model (FQM) and breadth-first search (BFS). We also proposed to update network database feature characteristics inside one peer, which would be used to update the peer linking information between peers to reconfigure the P2P-CBIR system. The system reconfiguration procedure can be carried out regularly, such that the network image search engine can operate on the most updated system parameters to yield the highest recall rate. Simulations demonstrate that, with reconfiguration, the recall rates can be improved to 7% to 38% as compared with the original P2P-CBIR without reconfiguration. The MIMF query mechanism has also verified that, by adopting multiple features types and use them intelligently, the recall rates can be improved up to 9.46%, as compared to that with few feature types.

[1] D. B. Terry, “Caching hints in distributed system,” IEEE Trans. Software Engineering, vol. se-13, no. 1, Jan. 1987.
[2] M. Batko, V. Dohnal, D. Novak, and J. Sedmidubsky, “MUFIN: A multi-feature indexing network,” IEEE Second International Workshop on Similarity Search and Application, 2009.
[3] MUFIN, http://mufin.fi.muni.cz/imgsearch/
[4] D. Novask, M. Batko, and P. Zezula, “Web-scale system for image similarity search: when the dreams are coming true,” IEEE Internal Workshop on Contenet-Based Multimedia Indexing, 2008.
[5] Napster, http://music.napster.com/napsterhomemain.htm
[6] I. Stoica et al., “Chord : A Scalable Peer-to-Peer Lookup Protocol for Internet Applications,” IEEE/ACM Trans. Networking, vol. 11, no. 1, Feb. 2003.
[7] S. Ratnasamy, P. Francis, M. Handley and R. Karp, “A Scalable Content-Addressable Network,” ACM Int. Conf. on SIGCOMM, pp. 161-172, Aug. 2001.
[8] S. Ratnasamy, S. Shenker, and I. Stoica, ”Routing algorithm for DHTs: some open question,” in Proc. First International Peer-to-Peer Workshop, pp. 45-52, 2002.
[9] The Gnutella protocol specification v0.41,
http://rfc-gnutella.sourceforge.net/src/rfc-0_6-draft.html, Jun. 2002.
[10] I. Clarke, Q. Sandberg, B. Wiley, and T. W. Hong, “Freenet: a distributed anonymous information storage and retrieval system,” in H. Federrath, editor, Designing Privacy Enhancing Tech., no. 2009 in LNCS, pp. 46-66, Jul. 2000.
[11] I. King, C. Hang Ng, and K. Chenung Sia, “Distributed content-based visual information retrieval system on peer-to-peer network,” ACM Trans.. Information. System, vol. 22, no. 3, pp. 477-501, 2004.
[12] Overview of the MPEG-7 Standard (version 7), ISO/IEC JTC1/SC29/WG11 N4674, Mar. 2002.
[13] D. Zeinalipour-Yazti, V. Kalogeraki, and D. Gunopulos, “Information retrieval in peer-to-peer networks,” Department of Computer Science and Engineering University of California – Riverside Riverside, CA 92521, USA.
[14] Y. Liu, X. Zhou, “Automatic texture segmentation for texture-based image retrieval,” IEEE Proceedings of the 10th International Multimedia Modeling Conference (MMM’04), 2004.
[15] M. Kokare, P. K. Biswas, and B. N. Chatterji, “Texture image retrieval using rotated wavelet filters,” Pattern Recognition Letters, vol. 28, pp. 1240-1249, 2007.
[16] Y. Rui, T. S. Huang, M. Ortrga, and S. Mehrotra, “Relevance feedback: a power tool in interactive content-based image retrieval,” IEEE Trans. Circuits and Systems for Video Technology, vol. 8, pp. 644-655, 1998.
[17] J.-J. Chen and C.-Y. Liu, “A universal query mechanism for similarity retrieval based on shape information in image database,” IEEE International Conference on Acoustics, Speech, and Signal Process, vol. 4, pp. 3676-3679, 2002.
[18] J.-J. Chen, C.-Y. Liul, Y.-S. Hunang, and J.-W. Hsieh, “Similarity retrieval in image databases by boosted common shape features among query images,” IEEE The 2nd Pacific-Rim Conference on Multimedia, PCM 2001, pp. 285-292, Oct. 2001.
[19] S. Datta, C. R. Giannella, and I. Kargupta, “Aprroximate distributed k-means clustering over a peer-to-peer network,” IEEE Trans. Knowledge and Data Engineering, vol. 21, no. 10, Oct. 2009.
[20] Y. Rui, T. S. Huang, and S.-F. Chang, “Image retrieval: current techniques, promising directions, and open issues,” IEEE Journal of Visual Communication and Image Representation, vol. 10, pp. 39-62, 1999.
[21] C. Carson, S. Belongie, H. Greenspan, and J. Malik, “Blobword: image segmentation using expectation-maximization and its application to image query,” IEEE Trans. PAMI, vol. 24, no. 8, Aug. 2002.
[22] J. R. Smith and S. F. Chang, “An image and video search engine for the world-wide web,” Storage and Retrieval for Image and Video Database (Sethi, I K and Jain, R C, eds), Proc SPIE 3022, pp. 84-95, 1997.
[23] J. R. Smith, “Integrated spatial and feature image system: retrieval, compression and analysis,” PhD thesis, Graduate School of Art and Sciences, Columbia University, Feb.
[24] A. Pentlandm, R.W. Picard, and S. Sclaroff, “Photobook: content-based manipulation of image databases,” IEEE Journal of Computer Vision, vol. 18, pp. 233-254, 1996.
[25] J. Kim, Peer-to-Peer overlay networks, http://www.cs.fsu.edu/jungkkim/P2P.html
[26] D. Tsoumakos and N. Roussopoulos, “Adaptive probabilistic search for peer-to-peer network,” Peer-to-Peer Computing, 2003, pp. 102-109, 1-3 Sept. 2003.
[27] Y. Zhu et al, “Making search efficient on Gnutella-like P2P system,” IPDPS’05, pp. 56a-56a, 04-08 April 2005.
[28] Z. Liu and M. S. Drew, Fundamentals of Multimedia, New Jersey: Prentice Hall, 2004.
[29] Virtual Drums, http://www.virtual-drums.com/frozen-cameleon.php
[30] B. S. Manjunath, J.-R. Ohm, V. V. Vasudevan, and A. Yamada, “Color and texture descriptor,” IEEE Trans. circuits and systems for video technology, vol. 11, no. 6, Jun. 2001.
[31] Course, http://course.cug.edu.cn/21cn/%B6%E0%C3%BD%CC%E5%BC%BC%CA%F5/Mmt09_06_3.htm
[32] 連國珍 著, “數位影像處理”, 儒林出版社 五版 2008
[33] J.-J. Chen, C.-J. Hu, and C.-R. Su, “Scalable image retrieval with optimal p2p configuration,” IEEE Trans. Multimedia, vol. 10, Feb. 2008.
[34] Corel Image, http://www.brothersoft.com/downloads/corel-image-database-dvd.html
[35] Caltech 101, http://www.vision.caltech.edu/Image_Datasets/Caltech101/
[36] Flickr, http://www.flickr.com/groups/tw/