隨著網路與多媒體技術的進步，網路多媒體應用已經成為日常生活中主要的資訊傳遞方式，隨之而來產生巨量資料 ( Big data ) 使得網路資料量呈現爆炸性成長，透過網路連結成為一個巨量影像資料庫 ( Large-scale image database)。另外雲端運算平台平行處理的技術進步，結合多媒體資料與雲端運算的應用成為當前的重要發展趨勢。目前的影像檢索技術以基於內容之影像檢索 ( Content-Based Image Retrieval, CBIR ) 最為常見，大多的CBIR系統以單一種類特徵值進行比對，無法有效反應影像內容的特性，因而檢索精確度有限，而在巨量影像資料庫的情況下，使用單一硬體伺服器進行影像檢索，因為檢索比對所需要的運算量過大，需要較長的檢索時間。因此，如何有效運用雲端運算技術有效地提昇巨量影像資料庫檢索的速度以及精確度效能，對於雲端巨量資料運算的應用相當重要。
本論文提出一套雲端巨量影像資料庫檢索系統，使用不同功能之影像特徵，分別為比對用特徵值以及分類用特徵值，藉以解決使用單一種類特徵值精確度不佳的問題；我們提出基於權重反向索引之資料庫分類演算法 ( Database-Categorizing based on Weighted-Inverted Index, DCWII ) 以及資料庫篩選演算法 ( Database-Filtering Algorithm, DFA )，利用權重反向索引 ( Weighted-Inverted Index ) 將影像資料庫分類，再根據檢索影像之特徵權重，設立一個門檻值進行資料庫事前篩選 ( Pre-filtering )，只選擇正相關影像資料庫進一步詳細檢索；我們將提出的方法結合雲端運算架構，加快資料庫比對的速度，降低單一硬體伺服器之比對運算量。實驗結果顯示，本論文之雲端巨量影像資料庫檢索系統在mAP效能上維持在0.678的表現，而在PR曲線效能上優於其他文獻方法，同時能夠比其他文獻的方法減少約55%至70%的檢索時間，所提方法和系統能有效地改善巨量影像資料庫檢索系統之效能。

With the advance of multimedia technology and communications, images and videos become the major streaming information through the Internet. How to fast retrieve desired similar images precisely from the Internet scale image/video databases (Big Data) is the most important retrieval control target. In this paper, a cloud based content-based image retrieval (CBIR) scheme is presented. To speed up the features matching process for large scale CBIR, we proposed to perform Database-Categorizing based on Weighted-Inverted Index (DCWII) and Database Filtering Algorithm (DFA). In the DCWII, it assigns weights to DCT coefficients histograms and categorizes the database by weighted features. In addition, the DFA filters out irrelevant image in the database to reduce unnecessary computation loading for features matching. Experiments showed that the proposed CBIR scheme outperforms previous works in the Precision-Recall performance and maintains mean average precision (mAP) about 0.678 in the large-scale database comprising one mega images. Our scheme also can reduce about 55%~70% retrieval time by pre-filtering the database, which helps to improve efficiency of retrieval system.

[1] N. Naphade et al, “Large-scale concept ontology for multimedia,” IEEE Multimedia, vol. 13, no. 3, pp. 86-91, 2006.
[2] Q. Zhang et al, “Cloud computing: state-of-the-art and research challenges,” J. Internet Services and Applications, vol. 1, no. 1, 2010.
[3] Google Image Search ( https://images.google.com ).
[4] Yahoo! Image Search ( https://images.search.yahoo.com ).
[5] TinEye Image Search ( https://www.tineye.com ).
[6] 百度識圖 ( http://image.baidu.com ).
[7] M. Chen et al, “Big data: A survey,” Mobile Networks and Applications, vol. 19, no. 2, pp. 171-209, 2014.
[8] P. Poursistani et al, “Image indexing and retrieval in JPEG compressed domain based on vector quantization,” Mathematical and Computer Modelling, vol. 57, no. 5, pp. 1005-1017, 2013.
[9] F. Rajam et al, “A survey on content based image retrieval,” Life Science Journal, vol. 10, no. 2, pp. 2475-2487, 2013.
[10] S. Gandhani et al, “Content based image retrieval: survey and comparison of CBIR system based on combined features,” Int. Journal of Signal Processing, vol. 8, no. 10, pp.155-162, 2015
[11] S. S. Mukati, and N. Rastogi, “A survey on image retrieval techniques with feature extraction,” IRJET, vol. 3, no. 1, pp. 323-327, 2016.
[12] T. Tuytelaars et al, “Local invariant feature detectors: a survey,” Foundations and Trends® in Computer Graphics and Vision, vol.3, no. 3,pp. 177-280, 2008.
[13] K. Grauman, and B. Leibe, “Visual object recognition,” Synthesis Lectures Artif. Intell. Mach. Learn., vol. 5, no. 2, pp. 1-181, 2011.
[14] F. Bellavia et al, “Improving Harris corner selection strategy,” IET Computer Vision, vol. 5, no. 2, pp. 87-96, 2011.
[15] C. Harris, and M. Stephens, “A combined corner and edge detector,” Alvey vision Conf., vol. 15, pp. 147-151, 1988.
[16] D. G. Lowe, “Object recognition from local scale-invariant features,” in Proc. IEEE Int. Conf. on Computer vision, vol. 2, pp. 1150-1157, 1999.
[17] M. J. Swain et al, “Color indexing,” Int. J. Computer Vision, vol. 7, no. 1, pp. 11-32, 1991.
[18] 鍾國亮， “影像處理與電腦視覺”，五版，pp. 471-472，2012.
[19] M. A. Stricker, and A. Dimai, “Color indexing with weak spatial constraints,” Electronic Imaging: Science & Technology. Int. Society for Optics and Photonics, pp. 29-40, 1996.
[20] 余秋忠，“以內容為基礎之影像與視訊處理技術於鑑識科學之應用”，國立中央警察大學鑑識科學研究所博士學位論文，2012。
[21] G. Raghuwanshi, and V. Tyagi, “A survey on texture image retrieval,” Springer Proc. 2nd Int. Conf. Computer & Communication Technologies, pp. 427-435, 2015.
[22] Z. Zhao, et al, “Content based image retrieval scheme using color, texture and shape features,” Int. J. Signal Processing, Image Processing and Pattern Recognition, vol. 9, no. 1, pp. 203-212, 2016.
[23] S. Agarwal et al, “Content based Image retrieval using color edge detection and discrete wavelet transform,” ICICT, pp.368-372, 2014.
[24] E. Gupta, and R.S. Kushwa, “Combination of local, global and K-Mean using wavelet transform for content base image retrieval,” Int. Journal of Computer Applications, vol. 116, no. 14, 2015.
[25] J. Canny, “A computational approach to edge detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 679-698, 1986.
[26] C. Iakovidou et al, “Searching images with MPEG-7 (& mpeg-7-like) powered localized descriptors: the SIMPLE answer to effective content based image retrieval.” CBMI, IEEE 12th Int. Workshop, pp. 1-6, 2014.
[27] A. Alzu’bi et al, “Semantic content-based image retrieval: a comprehensive study,” J. Visual Communication and Image Representation, vol. 32, pp. 20-54, 2015.
[28] X. Y. Wang et al, “A new SVM-based active feedback scheme for image retrieval,” Engineering Applications of Artificial Intelligence, vol. 37, pp. 43-53, 2015.
[29] S.Y. Irianto, “Content based Image Retrieval in the Compressed Domain,” Int. J. Computer Applications, vol. 99, no. 13, pp. 18-23, 2014.
[30] Z. M. Lu et al, “A content-based image retrieval scheme in JPEG compressed domain,” Int. J. Innovative Computing, Info. & Control, vol. 2, no. 4, pp. 831-839, 2006.
[31] D. Edmundson, and G. Schaefer, “Fast JPEG image retrieval using optimised Huffman tables,” IEEE ICPR, pp. 3188-3191, 2012.
[32] L. Zheng et al, “Fast image retrieval: query pruning and early termination,” IEEE Transactions on Multimedia, vol. 19, no. 5, pp. 648-659, 2015.
[33] Y. Zhang et al, “Image retrieval with geometry-preserving visual phrases,” IEEE CVPR, pp. 809-816, 2011.
[34] Color layout descriptor - Wikipedia
( http://en.wikipedia.org/wiki/Color_layout_descriptor ).
[35] Corel Image ( http://corel.brothersoft.com/corel-image-database-dvd.html ).
[36] Caltech 101 ( http://www.vision.caltech.edu/Image_Datasets/Caltech101/ ).
[37] MIR-Flickr 1M ( http://press.liacs.nl/mirflickr/ ).
[38] M. J. Huiskes et al, “New trends and ideas in visual concept detection: the MIR flickr retrieval evaluation initiative,” ACM Proceedings of the Int. Conf. on Multimedia information retrieval, pp. 527-536, 2010.
[39] Flickr ( https://www.flickr.com ).
[40] K. Kishida, “Property of average precision and its generalization: An examination of evaluation indicator for information retrieval experiments,” Tokyo, Japan: National Institute of Informatics, pp. 1-19, 2005.