近年來,由於多媒體技術的進步,有愈來愈多的影片是以數位化的方式製作和儲存。隨著資料量的增加,在今日繁忙的社會中,讓使用者沒有多餘的時間觀看過多的影片。因此,視訊摘要(Video Summarization)的發展讓使用者能夠在短時間內對影片有大概的認識,也就是從一整段視訊資料中,取出足以代表原本視訊資料的影像和聲音,讓使用者在較短時間及較少資料量的情況下,能夠了解整段視訊的原始意義的技術。我們為了讓觀看新聞的方式可以更加有效率,便利用新聞視訊當作分鏡轉換偵測(Shot Boundary Detection)的樣本,我們提出一套結合High-Level Fuzzy Petri Net能夠針對影片設定門檻值(Threshold Value)以及降低各種移動所造成分鏡誤判的分鏡轉換偵測。使用High-Level Fuzzy Petri Net來設定訂出影片分段的門檻值,本實驗結果減少耗費人工的判斷以及減少因個人對大量影片搜尋分鏡轉換時發生耗費時間的情況。
With the advent of digital era, the users have difficulty in using and absorbing overwhelming information brought out by technological advances in multimedia. Thus, the development of video summarization enables users to have a general idea about videos in a short time. This study focuses on the shot change, a part of the video summarization, to conduct an experimental sample on news programs. Moreover, a high-level fuzzy Petri net model is presented to describe boundary frames combination which indicates a shot boundary used for video frame sequence to detect both cut transitions and gradual transitions. This study uses a feature function to estimate the direct shot change in consideration of video shot boundary detection methods adopt the HLFPN method as a threshold value. The experimental results manifest it saves a lot of time and reduces the occurrence of improper shot change caused by the motions of objects and cameras when comparing this method with a manual method.