用於低位元率輕量級視訊壓縮的熵編碼模式切換技術

Abstract

這篇論文的主要目標是為低位元率輕量級視訊壓縮設計一種熵編碼模式切換演算法。在低位元率輕量級視訊壓縮中，熵編碼為系統中最花費時間的步驟。而熵編碼採用CABAC平均約需要CAVLC 1.5倍的時間。所以我們發展熵編碼模式切換演算法以達到有效切換模式的目的。我們利用熵編碼模式切換演算法計算在使用位元率控制下，單張畫面使用兩種熵編碼模式會造成的畫質差。當CABAC能比CAVLC能有效提升畫面達0.5dB以上，才使用CABAC模式，反之則使用CAVLC以節省時間。在演算法中，我們同時也為CABAC發展碼率預估演算法而能用更快的速度取得CABAC的壓縮量。從模擬結果顯示能比只用CAVLC模式平均每張畫面能達到0.7dB的畫質提昇，而相對CABAC比CAVLC需要的時間，使用切換演算法平均能節省40%的時間。

The objective of this thesis is to develop a dual-mode entropy coding selection algorithm for the low bit-rate lightweight video compression system. In applications of low bit-rate lightweight video compression system, entropy coding stage costs the most time and the time of CABAC is 1.5 times the time of CAVLC on the average. As a result, we develop a dual-mode entropy coding selection algorithm to choose the better mode for each frame encoded. Dual-mode entropy coding selection algorithm can be used to compute the PSNR difference of one frame which encoded by two kinds of entropy coding mode under rate control. When CABAC increases better PSNR value than CAVLC by 0.5 dB, the algorithm chooses CABAC mode to encode. When the value is below the threshold, the algorithm selects another mode. In our algorithm, we also present a CABAC rate estimation method to save computation complexity. Simulations shows that the proposed algorithm can produce better PSNR value by 0.7 dB than encoding by CAVLC only and save up to 40% additional simulation time than encoding by CABAC only.