透過您的圖書館登入
IP:13.59.210.36
  • 學位論文

在視訊中多移動物體之移除及修補

Multi-moving Objects Removal and Repair in Video

指導教授 : 黃惠俞
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


在快速發展的時代,數位化已經占據生活中的大部分,相機也從原本僅具照相功能的相機已轉變成具多功能的數位相機,攝影功能亦是最常見的。在這項功能中,往往記錄著令人充滿回憶的景象和畫面,而非僅單單的靜態照片。但在攝影之時,難免都會拍到非欲拍攝的景物或物體入畫面破壞原來的美景,基於此,本論文提出一個視訊修補技術,能有效地進行不必要的物體移除,並進行修補之動作,改變視訊內容,讓我們所拍攝的視訊影像可更優美的呈現。 本論文主要是基於多訊框前後具極高的相關特性設計一個可在視頻中以多個訊框為主的多移動物體移除及修補提出有效的修補策略進行處理。首先,先建置粗略的背景訊框,運用此訊框準確找到想要的目標物,對此目標物做擷取之動作,其次再對擷取的目標物做移除,最後,將移除掉之處利用視訊前後相關連特性及區塊為主的填補演算法進行修補處理。本論文修補過程主要是針對結構度做修改,使目標區域之結構部分強度增強,進而改變了優先權,使修補的起始點能夠更準確地落在結構上。

並列摘要


In an era of technical rapid advance, digitized life has occupied human lives; camera function has been upgraded and possessed the multi-functional abilities, such as simile model, auto-focused, object tracking, etc. One of public functions is video module which can capture a sequence of motion data. Hence, this function can capture motion scene and record the sights not only presents a static form. In photography, many inevitable situations that are undesired objects or occulted objects are captured into the video to destroy the original scene. Based on this, the thesis presents a video repair technology to effectively implement unnecessary objects or shelter removal and repair, and change the video content, so let the shot video image be more beautiful presentation. In this thesis, an approach which is a multiframe-based moving objects removal and repair technique in video has been proposed based on high correlation characteristics for multi-frame. The main procedures have the following phases. First, a rough background frame is built based on proposed method. Then, the desired target in the frame can accurately extract using the difference between the current frame and this background frame. Finally, these detected objects can further remove, and the rest of region in the current frame is repaired based on combining the relationships of frames and exemplar-based inpainting method. The repair process in this thesis is focused on the modification of structural degrees, let the intensity of structure property in target area be increase and further change the priority. The starting point for patch can be more accurately on the structure and obtain the good inpainting result.

參考文獻


[6]曾怡君,以背景相減法實現工作空間人物存在判定,國立台北科技大學自動化科技研究所,2012碩士論文。
[1]馬翔毅,使用動態背景補償以偵測與追蹤移動監控畫面之前景物,國立中央大學資訊工程研究所,2007年碩士論文。
[3]R. C. Chang, N. C. Tang, and C. C. Chao, “Application of inpainting technology to video restoration,” in Proc. of the First IEEE Int. Conf. on Ubi-Media Computing, pp. 359-364, 2008.
[5]張穎華,即時行人監控系統,國立中央大學資訊工程研究所,2006年碩士論文。
[7]張述理,用於快速H.264編碼之可適性可變區塊大小提早終止演算法,國立台灣師範大學應用電子科技研究所,2008年碩士論文。

延伸閱讀