數位影片上自動化的場景切割技術

Abstract

近年來數位影片是最受歡迎的多媒體資訊，也廣泛的被使用在多媒體系統中。為了達到快速搜尋或存取使用者所須的資訊，第一步就是要對這些影片做場景的偵測與切割。在我們的論文裡將提出一個很有效率的場景切割的技術，能有效的偵測在一般影片中較常出現的場景變換。藉由影片差異性的統計，可以將此方法應用在各種類型的影片上。同時，為了容易研究與使用，我們也為本技術發展了一個方便的使用的系統。

Recently, digital video is the most popular source of multimedia information. And it has also been used widely in multimedia systems. In order to search and access user-desired information quickly, the first task is to segment and detect the frames in which shot changes. In this thesis, an approach for efficiently detecting shot changes will be proposed. It could effectively detect various shot transitions, which are common in videos and include abrupt, fade-in, fade-out, dissolve and some with graphical editing effects transitions. Since the method is based on the statistics of video difference. It could be applied to various kinds of videos. Furthermore, in order to make use conveniently a shot change detection system is also provided.