在所有靜態影像壓縮技術中,層級樹狀結構分組 (SPIHT)已被証明對於影像壓縮相當有成效;然而,使用既有的技術來做傳輸有以下兩個缺點: 1.若錯誤發生在資料流的一開始就無法還原影像。 2.同時若錯誤發生在中途,錯誤將會蔓延,使整張影像的還原品質變差。 在本篇論文中我們提出一種利用多重描述法的影像傳輸系統。在本架構中,首先,先將影像經過小波轉換得到影像的小波係數並將所得的小波係數取出低頻和高頻部分,同時將高頻的部分做區塊編碼並分為兩群。之後再重新組合回兩個群包含低頻和高頻的描述,其中低頻的部分是兩群都一樣的部分。 再利用SPIHT編碼法則,針對每一群裡的低頻的部分及高頻各區塊部分做獨立的小波係數做編碼並於通道內傳送,而且在本架構的解碼端則看收到從編碼端傳來的資料流的多寡來還原中央合併解碼端及單邊解碼端的影像。 本論文所提出的架構,在無通道錯誤的情況下,中、低碼率的時候可以得到較顯著的還原效果。若在通道傳送時,通道錯誤率愈大則還原的效果會比原本SPIHT在相同通道下傳送效果好。
The objective of the thesis is to present the block-based algorithm for multiple description image transmission. In this algorithm, the wavelet coefficients of the image to be delivered are partitioned into different groups including the low-pass and high-pass subbands. Then, the groups of high-pass subband are divided with block-based algorithm. The wavelet coefficients in each block and the group of low-pass subband are independently encoded using the set partitioning in hierarchical trees (SPIHT) technique. The decoders of the transmission system can reconstruct the image by collecting the SPIHT-encoded bitstreams from any wavelet coefficients. Compare to equal-allocation algorithm, our simulation demonstrates that it can attain comparable performance for lossless channels. Furthermore, Our algorithm outperforms the basic SPIHT when the channels become lossy.