植基於顏色和結構特徵的影像查詢系統

Abstract

隨著網際網路以及電腦技術的快速發展，人們對於多媒體及影像的需求越來越多，因此，多傾向於將其數位化來保存，以供日後查詢檢閱之用。傳統以文字為基礎的檢索研究經過多年的發展已有顯著的成果，人們可以藉由自行定義的關鍵字，進而找到所關心的內容。此種方法或許也可運用在多媒體方面的檢索，由人力一步一步的對這些數位資料建立文字索引，但可想見的是這將是一件非常繁瑣的工作。 針對這樣一個問題，近年來興起一個熱門的議題：以內容為導向的擷取技術，研究的目的即在於如何有效管理一個龐大的數位資料庫並能提供使用者找出資料庫內和欲查詢的數位媒體有著相似內容的資料。在本篇論文，我們是對影像資料庫做擷取設計出一個系統出來，將重點放在顏色特徵和外形特徵的研究。實驗結果發現，我們的系統對於各類型的彩色影像都有不錯的表現，特別是對於有明確輪廓的影像有特別好的效果。

With the fast development of internet and computer technologies, the demand of multimedia and images are more and more. We prefer to digitize and store them for query and retrieval another day. Traditional information retrieval research, such as text-based information retrieval has remarkable achievement. People can use key words which they define themselves to search interesting information they care about. This method, may be we can use in multimedia information retrieval. Indexing step by step by person is a very tedious job. In the light of this problem, content-based retrieval technologies nowadays have gotten more attraction. An intention of research is how to manage a large digital database effectively and can support user to find similar information which they query by example in database. In this thesis, we design an image retrieval system based on content. We focus on color features and structure features in our system. According to our experiment, our system has a good performance in different kinds of color images. In particular, it has obvious improvement in clear contour which image has.