分散式視訊編碼(DVC)是ㄧ種新的影像壓縮型態,其主要將計算的負擔從編碼器移到解碼器。在本論文中,我們提出了以小波為基礎的分散式視訊編碼系統。其中關鍵畫面(key frames)使用SPIHT (set partitioning in hierarchical trees)做編解碼,Wyner-Ziv畫面則被分成低頻及高頻兩部分去做編解碼。高頻的部分使用修改後的SPIHT演算法,低頻的部份則是使用分群純量量化(grouped scalar quantization)及渦輪碼(turbo codes)去移除畫面間的冗餘。在論文中,我們也對影響系統效能的許多問題做研究,其中包含了小波轉換的層數、高頻最外層係數的估計及解碼端動作估計的影響。我們也提出了在Wyner-Ziv frame內的碼率分配機制,並且在不同的位元率下測試不同的影像序列來評估整個系統。在可以接受的複雜度,特別是在高位元率的狀況下,本系統也有不錯的圖型品質(PSNR)表現。
Distributed video coding (DVC) is a new paradigm for video compression, which shifts the bulk of the computational burden from the encoder to the decoder. In this thesis, we propose a new wavelet-based DVC system. The key frames follow the prestigious SPIHT (set partitioning in hierarchical trees) encoding and decoding procedures. The Wyner-Ziv frames are separated into the low-frequency subband and the other high-frequency subbands. The high-frequency subbands of Wyner-Ziv frames are coded using the modified SPIHT algorithm. The low-frequency subband uses the grouped scalar quantization and turbo codes to remove its interframe redundancy. We investigate several issues crucial to the proposed system, including level of wavelet transform, estimation of highest-frequency coefficients, and the effects of motion estimation at the decoding side. We also present a bit allocation scheme within the Wyner-Ziv frames. The overall system is evaluated over several test sequences at various bit rates. Good PSNR performance with reasonable complexity is observed, especially at high bit rates.