具有粒子濾波器與虛擬攝影機運動軌跡分析之視訊影像穩定法

Video Stabilization with Robust Particle Filter and Virtual Camera Movement Trajectory Analysis.

駱學穎(Hsueh-Ying Lo)

指導教授 ： 林道通

國立臺北大學/電機資訊學院/資訊工程學系/碩士(2016年)

摘要

攝影機的震動補償是一項重要的技術，因為移動式攝影機的使用在現今社會上是相當普及的，但並非所有可攜式攝影機均有搭載防震系統，所以需要靠後製來降低震動的產生。連續性的圖像震動補償，須要有準確的畫面匹配，使得影像看起來較為舒適與穩定，這種類型的校正可以被看作是精確度的問題。然而，校正後要確保整體畫面的流暢度。在本文中，我們提出了一個視訊穩定技術，利用區塊對應進行估算實際的相機移動路徑，並利用高斯濾波器運算出減少震動的虛擬平滑移動路徑，加以修補震動影像，使其成為較為穩定的影片。我們的工作重點是發開一套簡便的運算法，減少演算的複雜度，並確保輸出影像後的穩定程度。

關鍵字

動量資訊分析 ； 相機震動 ； 相機動量預估 ； 動量偵測 ； 影像防震

並列摘要

The camera stabilization and vibration compensation is an important technology. Nowadays,
the use of portable cameras in is widespread, but not all portable cameras are equipped
with stabilization system. To reduce the vibration we need apply the post-production procedure.
The video vibration compensation relies on the accurate registration between two
frames. In order to make the video looks more comfortable and stable, this type of correction
may be considered as a problem of image matching. However, we also need to ensure that
video fluency upon the correction.
In this thesis, we propose a video stabilization technology, using block matching algorithm
to estimate the actual movement trajectory of the camera, and using a Gaussian filter
operation to calculate the smoothly virtual movement trajectory, whereby repair vibration
video. The goal of our work is to develop a simple algorithm to reduce the complexity of
operations and to ensure the stability of the output video.

並列關鍵字

motion vector analysis ； camera vibration ； camera motion estimation ； motion vector ； video stabilization

參考文獻

[1] H. Jin A. Agarwala B.M. Smith, L. Zhang. Light field video stabilization. In 2009 IEEE

[2] C.H. Chen C.H. Chen, C.W. Hsu. Video stabilization by feature-block motion estimation.

Applications, pages 557–560, 2008.

[3] P.K. Biswas M. Okade. Fast video stabilization in the compressed domain. In 2012

IEEE International Conference on Multimedia and Expo, pages 1015–1020, 2012.