Image retrieval

本文藉由收集分析網路使用者實際進行個人圖像需求之檢索過程資料，並輔以觀察、問卷、及訪談等方法，共分析34名不同領域大學院校學生之圖像檢索行為特性，及其圖像相關判斷準則及線索，並觀察檢索前後的變化。主要研究結果包括：檢索需求多為人物之專指性主題，且檢索目的以個人休閒需求為主；圖像相關判斷中最重要的準則為「主題性」、「正確性」、及「美感性」，且檢索前後的變化不大，而其它準則如「聯想性」及「技術性」則呈現上升，「可及性」則下降；使用者背景、檢索需求類型、及資料類型對相關判斷行為具有影響。最後本文也針對圖像使用者與圖像資訊檢索系統提供一些改善建議。

The study collected and analyzed data on the process by which web users conduct real image search tasks. With the additional methods of non-obtrusive observation, questionnaire, and interview, this study investigated the image search behavior and relevance judgment of 34 college students from different subject domains. Pre- and post-search analysis was also conducted. The results include: users mostly search for specific images like people to fulfill their recreational needs; topicality, accuracy, and sense of aesthetics are the top three relevance judgment criteria; associativeness and technical details increase in importance, whereas accessibility decreases in importance from pre-search to post-search; user background, image needs, and data request types also have an influence on relevance judgment. To conclude, some suggestions on image literacy and image retrieval system design were provided.

隨著數位影像資源日益增多，愈來愈多的使用者透過圖像檢索系統來查詢圖像資訊，然而根據本人先前針對網路圖像檢索行為研究結果發現，圖像檢索需求與圖像檢索技術的相關研究，似乎呈現各自發展、雙軌並行的現象，欠缺有效的聯繫與整合；此外，過去對於圖像需求的分析也多以特定學科領域中的少數族群為主，欠缺一通用性的架構或模式，來作為圖像使用者研究的基礎。因此本研究擬提出一通用性之檢索需求分析模式，並以此為基礎，分析圖像需求與圖像檢索技術之間的關聯，以探討其作為瞭解圖像需求與提昇圖像檢索效益的可行性。

With substantial increase of image resouces, ,many users use image retrieval systems to access image resources. However, based on the author's previous research, there lacks an integration between image user study and image retrieval technology. Besides, most image user studies were done with special collection of images, or special groups of users. Few research focuses on a large scale and applies to more general-purpose retrieval environment. This paper proposes a general-purpose query analytic model to understand image needs and explore the limitation of current image retrieval technology and feasible enhancements.