(3.229.120.26) 您好!臺灣時間:2021/04/10 23:20
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:王瓊海
研究生(外文):Wang, Chiung-Hai
論文名稱:基於移動向量補償及利用人類視覺特性之立體視訊編碼
論文名稱(外文):Perceptual stereo video coding based on motion compensation
指導教授:周俊賢周俊賢引用關係
指導教授(外文):Chiung-Hsien Chou
學位類別:碩士
校院名稱:大同工學院
系所名稱:電機工程學系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:1997
畢業學年度:85
語文別:中文
論文頁數:75
中文關鍵詞:立體視訊編碼人類視覺系統臨界可視失真面
外文關鍵詞:stereo video codinghuman visual systemJND profile
相關次數:
  • 被引用被引用:0
  • 點閱點閱:103
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:1
由於顯像技術的快速發展及結合3-D視訊之多媒體應用所引起的廣泛討
論與興趣,3-D立體視訊的產生、傳送以及顯像技術已越來越受重視。因
為3-D立體視訊需大量的資料以建立起立體視覺效果,故立體的視訊信號
在傳送及儲存前必須予以壓縮編碼。本論文即針對如何有效地壓縮立體視
訊信號提出一可行的編碼架構。由於左眼視訊與右眼視訊的高相關性,我
們利用列交錯或行交錯來將左右視像結合為一合成之圖像,並利用內含於
此合成圖像之高相關性提升編碼效率。經由位移向量估測法,估算合成視
訊圖框中之區塊的位移向量,可以有效的去除二相鄰之合成圖像在時間上
的多餘量。 為了確保高壓縮比,源編碼的架構必需配合人類的視覺特性
。經由一人類視覺模型,吾人可估計對應於合成圖像的臨界可視失真面,
用以評估移動向量預估誤差信號之視覺重要性,並藉以做為適應性編碼之
依據。此外,吾人亦設計一區斷演算法用以挑選出與立體視覺效果有關的
信號。根據此等信號的位置調整對應之JND信號的大小以維持重建視訊的
立體效果。我們提出這個立體影像編解碼的目標效能是能在不超過0.2
bpp的位元率下,仍能達到32 dB的PSNR及38 dB的PSPNR的立體視訊品質。

With the speedy progress in the display technologies and
the fast growing interest in the multimedia that integrates 3-D
video information, the technology of generating, transmitting
and displaying 3-D stereoscopic video signals has recently
received intensive attention. Since 3-D stereoscopic video
display needs a tremendous amount of data to generate the
stereoscopic visual effect, the compression of stereo video
sequences is required before transmission and storage. This
thesis is concerned with the efficient data compression of
stereo video sequences. The high correlation inherent in a
stereo pair is fully utilized by merging the left and right
views of the video frame through row-wise or column-wise
interleaving. The temporal redundancy between the composite
video frames is effectively removed by the motion estimation
that predicts the interframe displacement vectors of the fixed
blocks. In order to guarantee a high compression
efficiency, the design of source coding algorithm is matched to
the characteristics of human visual perception. A composite
just-noticeable-distortion (JND) profile corresponding to the
composite video frame is derived from a perceptual model to
evaluate the perceptual significance of the prediction error for
adaptive source coding. A full-search algorithm is performed to
segment the prediction error image into non-overlapping blocks
that contain the maximal possible number of perceptually
important signals. The overlapping blocks will be sorted in
terms of the number of perceptually important signals so as to
determine the coding order for each non-overlapping blocks. A
segmentation algorithm is designed to locate the stereo-relevant
signals where the composite JND profile is modified for
preserving the quality of stereo visual effect. The performance
of the proposed stereo video codec have an average PSNR of 32 dB
and average PSPNR of 38 dB at a bit rate of no more than 0.2 bpp
with an acceptable stereo video quality.

QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔