標題: 卡氏積檔案的最佳化與圖樣資料的索引架構
A Study on Storage Organizations: Cartesian Product File Optimization and Iconic Indexing
作者: 周韻寰
Chou, Yun-Huan
楊維邦
Wei-Pang Yang
資訊科學與工程研究所
關鍵字: 合查詢;正交區間查詢;二維字串;二維階層字串;圖樣索引;Cartesian product file;partial match query;orthogonal range
公開日期: 1995
摘要: 決定了欲存取的資料,所以要設計一個好的資料儲存結構就必須要以查 詢 般來說,查詢可因查詢的條件型式不同而區分為:部份吻合查詢和正 交區間查詢兩種。所謂的部份吻合查詢葑囓韝丑A某些資料欄位被限定為 單一 查詢。若查詢條件中,資料欄位被限定為一範圍內的屬性值而不是 單一的屬性值時,這種查詢即稱為正交區間查詢(ORQ)。只能針對資料的 單一屬 及資料之多重屬性的查詢----PMQ和ORQ,來說,是無法保證一定 能有所助 有兩種特別的資料結構,稱為:分割醢序檔案和搜尋樹結構, 被提出來支 查詢的效率。這裡所說的『檔案』,係指由多重屬性資料所 組成的資料檔 醢序所得到的檔案結構中,卡氏積檔案是最典型的一種。 而正交區間查詢 查詢,且涵蓋了部份吻合查詢,因此,我們的研究即定 位在:設計與分析 的最佳卡氏積檔案上。我們的目標是將此問題以各種 角度來探討,包括: 找出益於正交區間查詢的最佳卡氏積檔案、萃取出 益於正交區間查詢的最 各種特性,以及發展出一個快速的演算法來找尋 近似最佳的卡氏積檔案。 對位置可以用來構成影像的另一種索引,這種 索引即稱為圖樣索引。這種 像的存取,也提供使用者另外一種查詢的方 式。有名的圖樣索引有二維字 等。由於圖樣索引儲存在主記憶體中,因 此如何設計一個更經濟的圖樣索 。在本論文中,我們提出了另一個新的 圖樣索引型式,它不僅節省更多的空間,同時在影像資料的存取上也較以 往的更有效率。 Users' queries determine what data will be retrieved, therefore good storage organization must be based on queries. In general, distinguished into partial match queries and orthogonal range partial match query (PMQ) is a query in which fields arengle values. If a query in which fields are restricted to a range of than a single value, then this kind of queries is called an query (ORQ). Because index structures only built on some onea, it is not guaranteed that those index structures are helpful for which queries involve more than one attribute of data. Thus, nds of data structures proposed to support systems where these types common: partitioned hashing file and search tree structure.t file (CPF) is a typical hashed files by partitioned hashing, and commonly used for contemporary queries and can be regarded asf PMQs. Hence our study intends to focus on the design andmal CPFs for ORQs. We intend to explore this problem in various find the optimal two-attribute CPF for ORQs, extract properties N-attribute CPF for ORQs, and develop a very fast algorithm to optimal CPFs for ORQs. The relative relationships among objects be organized into a kind of index for images. This kind ofonic index, not only speeds up the image retrieval, but also supports way. The most famous iconic indices proposed include 2D-string Iconic indices are stored in main memory, thus how to design a iconic index is very important. In this dissertation, we also new kind of iconic index, which saves more space and can be used image data more efficiently.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT840394071
http://hdl.handle.net/11536/60518
顯示於類別:畢業論文