標題: 影像相似度評估函數之分析與比較Analysis and Comparison of Image Similarity Measure Functions 作者: 許哲魁Che-Kuei Hsu陳稔賈叢林Zen ChenTsorng-Lin Chia資訊科學與工程研究所 關鍵字: 影像相似度評估函數;分析;非參數轉換;混合式比對;Image Similarity Measure Function;Analysis;Non-parametric transform;Hybrid matching 公開日期: 2001 摘要: 二張影像間的關聯性比較及評估是許多影像處理應用中的關鍵技術。我們以一個列向量來表示中心位於目標點的子影像視窗，然後就可以由兩張子影像的列向量來計算其關係性。本文將以向量長度及角度兩種幾何觀點，探討六種常被使用的影像相似度評估函數。依據它們的幾何特性，對九種不同的狀況解釋及驗證其優缺點，用以提供使用者針對不同的應用環境，選擇適當的影像相似度評估函數，以公平應用在不同的影像量度上。此外也針對使用這六種影像相似度評估函數之間不等式關係作探討，以提供交互選取使用時門檻值設定的依據。The comparison and evaluation of the correlation measure between two images is the key component of many image processing applications. We represent the subimage in the window centered at the target point as a column vector. Then the correlation measure between two subimages is computed based on the two column vectors associated with the two subimages. This thesis discusses the six frequently used image similarity measure functions from two viewpoints of vector distance and angle. According to their geometric properties, nine cases are analyzed and their pros and cons are described. Thus, users may choose the proper image similarity measure function to deal with different images under different application environments. Besides, the inequality relations between these six image similarity measure functions are discussed to provide the guideline for the threshold setting. URI: http://140.113.39.130/cdrfb3/record/nctu/#NT900392038http://hdl.handle.net/11536/68452 顯示於類別： 畢業論文