使用長程參考幀之H.264錯誤恢復影像編碼

Abstract

由於封包遺失以及潛在的錯誤擴散，導致透過錯誤傾向環境來傳輸壓縮後的影像，是件困難的議題。在本篇論文中，提出了基於多重參考幀的錯誤恢復架構，利用具錯誤恢復能力的幀(ER-frame)作為參考幀之一，並採用R-D最佳化 (Rate-Distortion Optimization)決定ER-frame作為參考幀的方格個數與位置。為了減少時間複雜度，提出了一個加速演算法，藉由參考幀選擇(reference frame selection)以及跳過不必要的參考幀(reference frame skipping)來達到加快移動偵測的效果。實驗結果顯示，在移動偵測中利用ER-frame作為參考幀可以增強錯誤恢復的程度，採用RDO也使得所提出的方法具有適應不同錯誤程度的環境以及不同視訊影像的能力，而所提出的加速演算法，大大減低了計算量，而只承受一些可被忽略的視訊品質衰減。

Transmission of compressed video over error-prone channels presents many challenges due to packet loss and potential error spread effects. In this thesis, an error-resilient scheme based on multiple reference frames is proposed, which employs error-resilient frames (ER-frame) as part of the reference frames and adopts a rate-distortion optimized (RDO) technique to choose the number and location of blocks referring to the ER-frames. To reduce time complexity, a fast algorithm is further proposed, which speeds up motion estimation process by using the techniques of reference frame selection, and reference frame skipping. Experimental results show that employing ER-frames in motion estimation can enhance error-resilient performance. The RDO technique adopted makes the proposed scheme adaptive to various channel loss rates and video sequences. And with the proposed fast algorithm, computational cost can be reduced significantly with negligible performance loss.