請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/50345
標題: | 基於空間上的旁資訊生成使用於分散式影像編碼 Spatial-based Side Information Generation for Distributed Video Coding |
作者: | Ji-Ciao Luo 羅際巧 |
指導教授: | 吳家麟(Ja-Ling Wu) |
關鍵字: | 分散式影像編碼,WZ視訊編碼,旁資訊,內插法,超解析度, distributed video coding,Wyner-Ziv video coding,side information,interpolation,super resolution,subsampling input based DVC, |
出版年 : | 2016 |
學位: | 碩士 |
摘要: | 分散式影像編碼已經逐漸被重視於一些應用上。例如:監視系統、行動裝置、雲端等。它有著輕量化的壓縮端,重量化的解壓縮端,不同於傳統的影像編碼(例如:MPEG, JPEG, H.264)的特性。然而他的實作方式,最主流的是DISCOVER團隊提出的架構,一種基於通道編碼還原的機制來達到分散式影像編碼。然而,掌控著編碼效率的最重大因素是在於解壓縮端的旁資訊的好壞。旁資訊產生的夠好,所需要通道編碼還原的位元就越少,並且還原的影像也更加地接近真實未壓縮過訊號。但是現今DISCOVER團隊架構下的分散式影像編碼有個致命的缺點,那就是旁資訊是由時間軸上不同的圖像所生成的。這造成一個嚴重的瓶頸,當影像出現快速動作的畫面時候,旁資訊就難以預測,造成解碼過後的影像品質不好。這篇論文主要提出了一個新的架構,來克服在高速畫面情況下解碼品質不好的情形。提出的架構是基於空間上而非時間軸上的旁資訊影像生成,這使得旁資訊的品質不會受到高速畫面的出現而低落。實驗結果顯示在高速畫面情況下,提出的分散式影像編碼的效能可以有效地勝過DISCOVER團隊所提出的分散式影像編碼。 The novel architectural characteristics of distributed video coding (DVC) are its simple encoder and complex decoder, which are unlike those of conventional video codecs (MPEG, H.264… etc.). The state-of-the-art DVC architecture is realized by using channel coding techniques in which side information (SI) quality plays the major role in overall rate-distortion (RD) performance. If SI quality is good, bit rate and channel decoding iterations are relatively low whereas the corresponding PSNR is relatively high. However, a limitation of current DVC architectures is that the SI generated in the decoder side by using temporal frame interpolation is rather poor for high motion or large group of pictures (GOP) sized sequences. The novel subsampling input based spatial domain SI creation scheme proposed in this study overcomes the above mentioned shortcomings of temporal domain SI creation. In the proposed architecture, SI is spatially interpolated instead of temporally interpolated. Therefore, the quality of SI is independent of the motion and the GOP-size of the input video. Experimental results show that a substantial RD performance gain can be obtained by using the proposed spatial domain SI creation mechanism as compared with conventional temporal domain approaches. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/50345 |
DOI: | 10.6342/NTU201601552 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-105-1.pdf 目前未授權公開取用 | 1.63 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。