一個使用多層串流技術之轉碼器架構設計於精細可調層次式到MPEG-1/2/4單一層次式的位元流轉碼應用

Abstract

本篇論文提出一個精細可調層次式到MPEG-1/2/4單一層次式的位元流轉碼器架構，並應用兩個新發展的轉碼技術以改善現有異質環境轉碼的。多層轉碼技術藉由傳送額外的加強層來補償異質性轉碼中的所造成的嚴重錯誤傳遞問題。位元率-失真最佳化模型則在有限的通道頻寬下，提供一個在編碼效能以及傳輸位元率之間最佳平衡點的編碼決策。實驗結果顯示所提出的使用多層轉碼技術的轉碼架構與一般的架構相比較能夠提供非常良好的轉碼效能。整合所提出的冪次位元率-失真模型以及多層轉碼技術，我們所提出的轉碼器架構，與被視為擁有最低轉碼複雜度之一的簡化餘弦轉換域轉碼架構相比較，可達到最多5.8dB的影像訊號雜訊比改善，且幾乎不增加其轉碼複雜度。與另一種擁有最佳轉碼品質的串接餘弦轉換域轉碼架構相比，我們所提出的多層轉碼技術的轉碼架構在低位元率時有相近的視訊品質，在高位元率時有0.2–1.4dB的影像訊號雜訊比損失，但轉碼複雜度上只需其34.17%的計算複雜度。

In this thesis, we proposed a Fine Granularity Scalable (FGS) multi-layer to MPEG-1/2/4 single-layer transcoding framework and used two new techniques, multi-layers transcoding and R-D optimization modeling, to improve the poor performance in heterogeneous transcoding. The multi-layer transcoding technique provided heterogeneous drift error compensation via transmitting an additional enhancement layer. The rate-distortion optimization modeling is used to achieve a better tradeoff point between coding efficiency and transmission bit rate under limited channel bandwidth. The experimental results showed the proposed Multi-layer Simplified DCT-Domain Transcoder (MSDDT) architecture can provide a very good transcoding performance compared to the conventional architectures. With the proposed power-law R-D modeling in the proposed multi-layer transcoding techniques, the proposed framework shows up to 5.8 dB PSNR gains over the SDDT architecture under the similar transcoding complexity. Compared to the Cascaded DCT-Domain Transcoder (CDDT) architecture, the MSDDT architecture has similar PSNR quality at low bit rate and about 0.2–1.4 dB loss in PSNR at high bit rate, but with only keeping 34.17% complexity compared with cascaded decoder-encoder (DEC-ENC) architecture.