於MPEG格式教學影片上進行自動擷取主要畫面研究

Abstract

主要畫面擷取是在進行建立影片索引及影片內容探尋時,非常重要的前處理步驟。而教學影片,其主要畫面定義為投影片切換的畫面。本研究提出一套有效率的演算法,針對MPEG格式教學影片進行主要畫面擷取,能在不需還原壓縮的格式下,直接分析畫面本身的差異性,進而找出教學影片中的主要畫面。本研究由MPEG格式中的I畫面取出代表亮度的Y畫面,再以Y畫面中的DC值形成YDC畫面,依教學影片的特性,將IYDC畫面分成內、外兩個區域,比較連續兩張YDC畫面的內外區域差異,來找出發生投影片切換的時間點。本研究並已運作於目前實際進行的教學課程,實驗數據顯示本研究的方法能有良好的效能。
Key-frame extraction is an important pre-process step before video indexing and retrieval. For lecture videos, key-frames are defined to be those involving slide-changing frames. This thesis proposes an efficient algorithm for automatic slide-change detection of MPEG format lectures videos. The proposed algorithm is based on analyzing the regional differences of the dc-values of the Y-channel of I-frames of the MPEG compressed video. Experimental results show that the proposed algorithm is fast and effective in detecting slide-changing frames while suppressing those involving intensity changes due to non-slide-changing activities.

Description

Keywords

主要畫面偵測, 教學影片, 視訊分割, MPEG格式影片, 場景變換偵測, keyframe detection, lecture video, video segmentation, MPEG format video, shot change detection

Citation

Collections

Endorsement

Review

Supplemented By

Referenced By