以區域為基礎利用小波轉換和形態學運算的影像擷取

Abstract

近年來，隨著電腦科技發展迅速和生活數位化，數位影像資料庫的數量和容量正以極快的速度增加。如何有效的管理一個大型的數位影像資料庫也越來越重要。因此影像檢索便成為一個重要的研究領域。以區域為基礎的影像檢索法已經被證明比影像內容為基礎的影像檢索法來的有效率，因為它利用物件來描述影像而非使用整張影像內容。本研究所提出的方法先利用K-means分群法將影像切割成數各區域，並分別使用小波能量和形態學運算來描述區域的紋理、彩色和形狀特徵。再依此計算查詢影像和資料庫中影像的相似度，並選出最相近的影像。實驗結果顯示我們所提出的方法是快速且有強韌性的。

In recent year, with the rapid development of computer technology and digitize life, digital image databases have grown enormously in both size and number. How to effectively manage the large image databases is more and more important. So image retrieval has become an important research topic. Region-based image retrieval has been proven more effective than content-based image retrieval, because it overcome the deficiencies of content-based image retrieval by representing images at the object-level. The method proposed in this thesis utilizes the K-means clustering algorithm to segment the image into regions, and use the wavelet energy and morphological operator to extract texture, color and shape features to describe regions, respectively. Then we can retrieve the most similar images according the similarity between query image and database image. Experimental results indicate the proposed method is fast and robust.