完整後設資料紀錄
DC 欄位語言
dc.contributor.author連健琳en_US
dc.contributor.authorChien-Lin Lianen_US
dc.contributor.author李素瑛en_US
dc.contributor.authorSuh-Yin Leeen_US
dc.date.accessioned2014-12-12T02:20:17Z-
dc.date.available2014-12-12T02:20:17Z-
dc.date.issued1998en_US
dc.identifier.urihttp://140.113.39.130/cdrfb3/record/nctu/#NT870392018en_US
dc.identifier.urihttp://hdl.handle.net/11536/64038-
dc.description.abstract隨選視訊服務是最熱門的多媒體應用之一。不管在什麼時間或地點,使用者都可以透過隨選視訊服務看到電視影片、電影及新聞等。然而,隨著影片的數量達上千個小時,在查詢後我們仍會得到許多符合的影片。為了要在短時間內挑選出我們最想要的影片,我們需要一個有效率的影片摘要方法以萃取影片的概要內容。 考慮到影片是由一些故事單元所組成,我們提出一個影片摘要方法,從影片的每個故事單元中擷取出最重要的影片片段,並把這些片段合成一個影片摘要。這個影片摘要大約只有原來影片長度的十分之一,但重要的是,使用者在看過這段影片摘要後,便能快速的瞭解到原來影片的大致內容及架構。為了開發這個影片摘要系統,我們提出一些視訊處理的技術,包括先將影片做切割的視訊切割方法,將視訊片段叢聚成故事單元的視訊叢聚方法及從故事單元中擷取重要片段的方法。我們提出一個以GOP為基礎的視訊切割方法,這個方法比一般以frame為基礎的視訊切割方法節省相當多時間。在視訊叢聚上,我們的演算法可以有效率的將影片片段叢聚成故事單元。最後我們也定義了重要影片片段以及從故事單元中擷取這些片段的演算法。另外,我們也提供了一個整合的影片摘要環境,包括視訊影音分離系統、視訊編輯系統、影片摘要系統和視訊影音整合系統。系統□的全部工作都是自動的不需人工介入,而且透過實驗,我們也得到令人滿意的結果。zh_TW
dc.description.abstractVideo on Demand ( VOD ) service is one of the most popular multimedia applications. By VOD services people can watch TV programs, movies, or news whenever they want or somewhere far away from server. However, with the volume of videos growing to thousands of hours, there may be lots of matched videos in the query result. For getting the most desired video in a short time, we need an effective approach to summarize the video for providing the overall view of videos. Considering that the video is composed of story units, we propose a video summary system which extracts the significant segments from each story unit of the video and combines them into an abstracted video. This abstracted video is about one-tenth of the original video in time-length and more importantly, presents the critical information such that users can rapidly realize the synopsis and have an overall view of the original video. To implement the video summary system, some techniques of video processing have been proposed in our system, including segmenting the video as shots, clustering the shots into story-units and selecting the significant shots from each story-unit. In shot segmentation, we propose GOP-based scene change detection approach, which saves more processing time than scene change detection frame by frame. In shot clustering, we design a shot clustering algorithm to effectively cluster shots into story units. We also define the significant shots and present the algorithms to extract the significant shots in a video. In addition, we provide an integrated environment for video summary system, such as audio-video splitter, audio-video editor, video summary and audio-video combiner. All tasks in this system are done automatically and the results of experiment prove to be justifiable and satisfactory.en_US
dc.language.isoen_USen_US
dc.subject視訊摘要zh_TW
dc.subject視訊瀏覽zh_TW
dc.subject場景變化偵測zh_TW
dc.subject視訊叢聚zh_TW
dc.subject視訊特徵辨認zh_TW
dc.subjectVideo Summaryen_US
dc.subjectVideo Browsingen_US
dc.subjectScene Change Detectionen_US
dc.subjectVideo Clusteringen_US
dc.subjectVideo Characterizationen_US
dc.title隨選視訊故事單元擷取及影片摘要瀏覽之研究zh_TW
dc.titleVideo Summary and Browsing Based on Story-Unit for Video-on-Demand Serviceen_US
dc.typeThesisen_US
dc.contributor.department資訊科學與工程研究所zh_TW
顯示於類別:畢業論文