- 學位論文
具廣告偵測與剪輯之影音播放器
Design and Implementation of Media Player with Advertisement Detection and Video Reproduction
摘要
電視臺業者為了保存其播出紀錄及內容，均已舊時的貯藏方式留存下來，隨著數位化時代的到來，許多較舊的影帶儲存方式已不復存在，現今將舊式類比的檔案轉換成數位化的檔案內容之後卻存在著許多與節目無關之非節目內容；時下的網路影音播放網站也越來越發達，根據西元2008年 PricewaterhouseCoopers的研究報告中就顯示出多媒體產業別市場從2005年起，隨著創新科技及多媒體服務型態之轉變，產業的競爭已進入白熱化的階段，尤其網際網路影音產業的年成長率高達百分之三十二，居全球之冠。影音媒體產業的快速變遷加上e世代的到來，改變了其觀眾的收視習慣，如果這時能夠將數位化後之檔案有效濾除其不必要之資訊，不僅幫助了電視臺業者儲存上的便利，更能有效將其影片上傳至影音網站中，使得觀眾能夠更方便的觀賞節目。 有鑒於此本論文提供了兩種不同的方法分別針對電視臺業者側錄影片以及其節目播出帶進行非節目偵測；側錄影片為利用臺灣節目中其特有特色：節目標誌圖形，進行非節目內容之偵測；於節目播出帶中並無商業性的廣告，電視臺業者會在節目播出帶節目與節目之間插入數十個至數百個黑幕，本論文將提出簡單而有效的演算法，進行非節目內之濾除；實驗結果顯示，本論文提出之方法可以偵測出影片中所有非節目內容之片段，此外偵測出來的準確率平均可達百分之九十五以上。
關鍵字
廣告偵測,標誌偵測,影音播放並列摘要
With the digital era coming, many storage methods of old tape is no longer exist. In order to keep the records and the contents of programs, TV industry mainly preserves them by old storage methods. However, there are many contents do not relate to the program while the digital files converted from analog files. Nowadays, audio and video broadcast on website is more and more convenient. According to the research report of PricewaterhouseCoopers in 2008, the competition of multimedia industry has been heated up with innovative technology and the changes of multimedia service types. In particular, the annual growth rate of internet video industry has reached to 32% and it’s the highest rate in the world. The rapid transitions in multimedia industry and the advent of E generation change the viewing habits of audiences. If we could efficiently filter out the unnecessary information from the program, it will not only help the television industry to store the program conveniently also make easier for audiences to watch program on the internet. In view of this, we proposed two different methods for the recorded type TV industry and program video to detect non-program fragment. It uses the characteristic of Taiwan TV program, the logo of program, and detect non-program fragment. However, the program video doesn’t content the commercial advertisement. TV company uses black frame detections to detect program video, there are dozens to hundreds black frames between program sections in the video. We proposed a simple and effective method for these cases. According to the experiment, our method can find out all non-program fragments. And the precision and recall are exceeding more than 95%.