封包網路之雙層視訊編碼器設計

Abstract

本篇論文共涵蓋三個主題。第一部分提出數個改良的雙層架構視訊編碼器 ，用以強化重建之影像品質。其主要應用於非同步傳輸模式網路 (ATM Networks)上之視訊傳送。第二部分探討用以產生最佳重建影像品質的量 化器步階尺度。第三部分比較單層及雙層編碼器對於封包遺失之抵抗力。 在一個採用空間域分解的雙層編碼器中，基本層傳送低解析度之影像；而 增益層則負責補足剩餘的資訊，以提高重建影像之品質。本論文共提出四 種改良的雙層編碼器。其不僅大幅地減少了每張影像所需傳送的位元數， 同時只依基本層封包重建之影像品質亦獲得顯著的改善。此外，我們也提 出許多增益層之編碼法則。其中數種確實能夠進一步地減少所需傳送的訊 息量。一般而言，對於增益層之量化器，若充許較高的輸出位元率可獲得 較小之量化失真。然而高位元率亦導致封包遺失率之提高，從而增加了頻 道失真量。本論文探討了在同時考慮訊號量化及頻道失真的誤差來源下， 如何適當地選取其步階尺度，以達最佳之重建影像品質。無論理論分析或 實際模擬，均驗證我們確實求得了使接收端獲得最小均方誤差訊號之最佳 操作點。最後，我們將雙層及單層編碼器之影像重建強韌性作一比較。實 驗證明當頻道失真發生時，雙層編碼器由於基本層封包獲得保護，確實較 單層編碼器更能抵抗封包遺失造成之失真。 This thesis contains roughly three parts. In the first part, several improved two-layer coding schemes have been proposed to enhance the quality of the reconstructed pictures. In the second part, we look for the optimal quantizer step size that produces the best received picture quality. In the third part, we investigate the packet loss effects on the single-layer and the two-layer codecs. In a two-layer coder with spatial decomposition, the base-layer transmits low-resolution images and the enhancement-layer, the rest information. Three improved schemes on the base layer coder have been proposed. The total bits per frame have been significantly reduced and the pictures have been drastically improved. In addition, several enhancement-layer schemes have also been proposed. The optimal quantizer step size for the enhancement layer has been studied under the consideration of the source-plus-channel distortion. Through both analysis and simulation, we indeed find an optimal operating point that achieves the lowest total MSE at the receiver. Finally, we show that when channel errors occur, two-layer schemes appear to be more robust against cell losses than single-layer schemes.