- 學位論文
基於移動視差補償之立體視訊畫框插補研究
Stereoscopic Video Construction Based on Motion and Disparity Compensated Frame Interpolation
摘要
近年來隨著科技的進步，加速了平面顯示技術的突破，次世代的3D立體顯示技術也在科技進步的輔佐下正式的邁入量產階段，在顯示設備及其顯示技術的提供能與時並進，方能給予這個次世代的顯示系統一個完整的市場空間，同時能夠引領世界的視訊系統一個重大的革命。傳統立體影像擷取系統一般藉由雙鏡頭攝影機，同步擷取不同角度下之影像對，但雙鏡頭下不僅需要增加硬體上的成本，且影像處理系統的複雜度也與鏡頭數成正比，此外雙鏡頭攝影機需以兩眼的平均距離置放(約6.5公分)，因此體積較傳統的單鏡頭攝像機來得大，較不適合做在體積較小的裝置，如Webcam、手機、PDA。 立體顯示系統在同時間能夠顯示兩張不同的影像讓左右兩眼接收，使人們在兩眼分別看到不同的影像而產生深度感，進而使得使用者有更為真實的視覺效果，隨著立體視覺顯示器之製造技術日漸成熟，雙視角以及多視角視訊技術也愈來愈受矚目，由於其所需處理的資料量倍增，因此立體視訊編碼的壓縮技術也更為重要，立體視訊有著傳統的單視角視訊編碼技術所沒有的特性，如果使用傳統的單視角視訊編解碼技術分別針對各個通道的影像做處理，在其極高的運算複雜度倍增下，將會大幅降低編碼的效率。 本論文中提出一個基於單鏡頭立體影像擷取系統之立體視訊技術，在單鏡頭立體影像擷取系統下，其產生的影像資訊在同一時間點中只會擷取一張影像，於次一時間點擷取同水平線上另一具有角度差之影像，其畫面更新率相對只有傳統雙鏡頭立體影像的一半，藉由在此所提出之畫框插補演算法，在顯示端仍能在同時間顯示兩張具有視角差的影像，在演算法部分，利用傳統單視角視訊技術加上考慮了立體視訊特有的視差關係，藉由移動估計與視差估計所計算出的移動資訊，預測出在同一時間點上，未被擷取之另一角度的影像資訊，本研究的目的是希望能利用單鏡頭立體影像擷取系統的特性，藉由此論文提出的畫框插補演算法，在顯示端獲得和雙視點立體影像系統相同的顯示效果，模擬結果顯示出，比較一般單視角視訊技術所內插還原的影像，本論文所提出的方法在鏡頭移動的情形下約可提升PSNR 3~5dB，而在鏡頭固定下更可提升約6~9dB。
並列摘要
Stereo images can make users sense depth perception by showing two images to each eye simultaneously. It gives users a vivid information about the scene structure. With the technologies of stereoscopic video capture device and 3D-TV display device getting more and more mature, the importance of the stereo content will rise in the near future. Under this trend, stereo video system draws more and more attention. However, to build up a stereoscopic video system, many design challenges, such as bad coding efficiency, high computational complexity, and hardware architecture implementation, etc., must be overcome. In single-lens stereoscopic camera system, the frame rate of each channel only half conventional stereo video format. In this thesis, based on single lens stereoscopic camera system, a motion and disparity compensated frame interpolation (MDCFI) is proposed to overcome the overheads of conventional two lens stereoscopic camera. Pass by proposed MDCFI to up-conversion the frame rate of single-lens stereo video format, which obtained the video format same to conventional stereo video format. In algorithm, using the conventional video technique and considered the parallax characteristic of stereoscopic video system. Calculated the information of movement base on the motion and disparity estimation, and the un-captured video information which was on the other side at that particular time can be predicted. The purpose of this research is to utilize the characteristic of single-lens stereoscopic camera system and combining the algorithm of frame interpolation proposed in this paper, to obtain an equal displays outcome as dual-lens stereoscopic video system. The simulation results has proved that comparing the images from conventional algorithm of frame interpolation technique, the images using the proposed method in this paper has the PSNR improved 3~5dB while camera moves and the PSNR improved about 6~9dB while the camera was fixed. As a result, the proposed algorithm significantly increased PSNR performance with less complexity compared to the conventional algorithm.
參考文獻
延伸閱讀
- 楊沛哲（2015）。線上雙狹縫彩色差動共焦量測系統與校正技術之精進研究〔碩士論文，國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU.2015.10954
- 游智宇（2010）。抬頭顯示器上目標重疊視訊之位移追蹤補償機制〔碩士論文，大同大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0081-3001201315110197
- Chang, W. S. (2007). 利用圖場的動作補償內插方法之分散式視訊編碼 [master's thesis, National Tsing Hua University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0016-1411200715101521
- Luo, S. J. (2013). Structural Stereoscopic 3D Image Editing Taking into Account the Depth Information [doctoral dissertation, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU.2013.00410
- 王迺岳（2002）。Introduction to Video Formats in Stereoscopic Displays。光學工程，(79)，31-39。https://doi.org/10.30011/OE.200209.0004