觀賞運動賽事長久以來一直是項重要的娛樂活動，而運動賽事的轉播在電視節目中也佔有相當大的比例。隨著數位錄影機日益普及，許多無法即時欣賞比賽的觀眾會使用數位錄影機錄下賽事以隨後觀賞。考慮到運動比賽通常持續數小時，而比賽精華只佔其中的一部份，若能讓數位錄影機具備自動擷取與提供觀眾有興趣之比賽精華的功能，將可為使用者帶來極大的便利，也因此在近幾年有不少運動比賽瀏覽或是精華擷取之相關研究。然而，目前的研究成果與實用性仍有一段不小的差距，主要原因在於許多研究將儲存的視訊資料解壓縮成一幅幅畫面，再施以影像處理及進階的演算以解構比賽內容，但龐大的視訊資料量讓有效率的分析變得極為困難，特別是製造消費者所使用的數位錄影機在成本方面有一定的考量與限制，在播錄功能之外很難騰出大量額外的計算能力供視訊分析使用，運動比賽分析研究的實用性因此打了很大的折扣。本研究旨在為運動比賽視訊分析提供簡化且實用的解決方案，希望由運動賽事轉播中篩選出可能是精華的片段以供進階的視訊分析使用。在現今的運動賽事轉播中，比賽精華片段會伴隨著慢動作重播，而轉播單位通常會在重播前後加上所謂串場效果以告知觀眾。本研究提出偵測與運動比賽重播相關之串場效果以利精華片段的擷取。我們的研究方法將直接處理經由數位廣播所傳來或由數位錄影機所錄製的MPEG 串流，從MPEG 串流中抽取及計算特徵值以節省訊號處理所需的計算時間與硬體需求。我們使用MPEG 串流中所包含的色彩與動作資訊，以及編碼模式等，利用串場效果所具有之時間短暫、大量顏色變化與快速移動等特性，分析影片以找出可能是串場效果的片段，然後利用這些可能的串場片段訓練出該場比賽的串場效果模式以供比賽的其他部分比對。所偵測的串場效果將協助我們把慢動作重播鏡頭由運動比賽影片中可靠準確地抽取出來。在有限的重播畫面資料中，我們可採用進一步的分析以對內容識別或分類，讓使用者能夠選擇有興趣的部份觀賞。由於使用者應對比賽精華部份以正常速度播放更感興趣，我們由所偵測到的重播鏡頭為出發點，向前尋找出所謂主視角畫面，作為每段精華片段的開始，以建立一套實用的運動比賽精華擷取系統。我們相信本研究所提出的機制應可對現有之運動比賽分析研究帶來實質的幫助。 ; Watching sports videos has always been an important and popular recreation and the sportscast takes a large portion of TV programs nowadays. With the rapid advancement of digital technologies, audiences can enjoy watching the sports games at home with their high-quality audio-visual facilities and even record the videos by using digital video recorders (DVR). When the audiences choose to record the video for time-shifting purposes, they may not be interested in watching the entire game but the video highlights only. The audiences may be benefited a lot if a novel DVR is equipped with the functionalities of extracting the highlights from sports videos automatically and accurately. Therefore, a lot of research work has been devoted on the highlight extraction in recent years. However, there exists a huge gap between the existing research works and their feasibility. The major challenge comes from the fact that the large volume of video data prohibits exhaustive processing on video frames, especially under the strict requirement of efficient implementation in consumers’ DVR. The objective of the research is to provide a practical way to bridge this gap. In the sportscast nowadays, the highlight parts are always followed by slow-motion replays. Besides, the broadcaster usually inserts a transition effect between the normal frames and the replaying frames to inform the audiences of the replay. Therefore, the appearance of a transition effect has a direct linkage to the video highlight. In this research, we propose to detect transition effects for sports videos highlight extraction. In order to reduce the computational cost of hardware, the proposed method processes the MPEG compressed bit-streams recorded by the DVR directly. We exploit the color/motion information and coding modes of MPEG streams and then analyze to determine whether the transition effects occur by using their characteristics. Furthermore, we collect the video segments of possible transition effects to train a template, which will be used to determine the existence of an effect in the remaining parts of video. The scheme has been designed as general as possible to cope with various kinds of effects. The detected transition effects will help us reliably and accurately locate the slow-motion replays. With limited amounts of replay data, we can apply advanced analysis more efficiently for content classification so that the users can choose the video segment of their interests. Since the users will be more interested in watching the scenes played at the normal speed, we trace back to search appropriate locations as starting points for highlights. A practical sports videos highlight extraction scheme can thus be constructed. We believe that our proposed scheme can act as a baseline framework and be beneficial to the existing research on analysis of sports videos. ; 研究期間 9708 ~ 9807
