非均勻正交調幅調變下壓縮視訊的資料排列

Abstract

在高解析數位電視廣播系統中，為了降低頻寬的需求，視訊訊號須經壓縮 編碼來減低資料量．然而，壓縮過的視訊訊號較之未壓縮者更容易受到傳 輸雜訊的干擾．因此在數位廣播中，編碼器與傳輸方式須適當設計以降低 雜訊的干擾．在本文中我們探討了非均勻式正交調幅調變的傳輸方式．它 可分解為兩個受到不同程度保護的頻道，而且其傳送的位元數有 1:2 的 關係．為了設計一適合的編碼器，我們修改了目前動態影像壓縮標準 (MPEG) 的編碼方式，將此一標準所編碼的資料依重要性分類，並重新排 列為兩組資料．由實驗結果顯示，以非均勻式正交調幅調變的傳輸方式配 合我們所設計的編碼器比傳統的正交調幅調變與編碼方式較不受雜訊的影 響，而且在設記上更有彈性． The DCT coefficients in transform coding block are of different importance for the image quality. Therefore, during transform, they should be given different pro- tection to ensure that the most important coefficients are correctly received. In the standard multilevel uni- form QAM, all the bits in a symbol are treated as of equal importance, therefore, it is not optimal for the transmission of the transform coded video sequence. The idea of embedded modulation is an extention of the standard QAM. In this modulation,the constellation is partioned into groups such that the modulation distances of the most significant bits inter the groups are in- creased at the sacrifice of the modulation distances of the least significant bits within the group. With such constellation, transmited bits are given unequal pro- tection to fit the characteristics of the transform coe- fficients. For a two level grouping, the original Huffman code sequence is divided into two parts. The first parts consists of the lower order DCT coefficients and motion vectors and the second part consists of the higher order DCT coefficients. In 64 QAM modulation, the 6 bits pattern is formed with 2 bits from the first sequence and 4 bits from the second sequence. With such arrangement, under equal transmission power, most of the bit errors will occur in the higher order coefficients. The consequence will have significant impact on the subsequent error concealment. With standard uniform modulation,each coefficient is subject to equal error probability. While with embedded modulation, error is likely to happen only in the higher order coefficients such that the subsequent error correction will be much simplified. The objective of this study is to simulate the joint source- coding/modulation system and to evaluate its per- formance.