取樣點適應性補償(Sample Adaptive Offset, SAO)是在高效率視訊編碼(High Efficiency Video Coding, HEVC)中的環狀濾波器裡加入的一種新的技術,用來對編碼量化後造成的損失進行取樣點補償。經過此程序處理後,取樣點會更接近原始取樣點,達到提昇編碼效率的目的;然而,SAO須迭代所有補償類型才能得到最佳的補償類型。 本論文提出一種用來快速估測SAO最佳補償類型的方法,並且使SAO達到不錯的效果。快速估測的概念是重複使用參考畫面的資料,主要是以運動向量(Motion Vector)、影像特性等資訊來加速SAO,並且利用鄰近區塊較相似的特性及SAD來減少畫面間預測所造成的錯誤。本論文提出的方法相較於HEVC參考軟體(HM12.0)的原始SAO演算法,減少約27%的時間複雜度,而編碼效率BD-PSNR平均只下降0.03dB,BD-Rate平均只上升0.586%。
Sample adaptive offset (SAO) is a new technique introduced to the HEVC video coding standard for compensating the distortion of quantization. The intensity value of reconstructed samples by SAO will be made closer the original ones. However, the process of SAO must search and iterate all SAO types in order to find the best type with minimal RD cost. In this work, we proposed a fast method to estimate the SAO type as well as its offset values for compensation without sacrificing the RD performance given in full-search SAO. The concept of fast estimation is to reuse the similar data from reference frames. We take advantage of the motion vectors and image characteristics to accelerate the SAO procedure. Besides, we use the characteristics of neighboring blocks and the block matching cost of SAD in target block to reduce the error propagation due to the inter prediction. Compared to the HEVC reference software version 12.0, our proposed method reduces about 27% of SAO encoding time, while with only 0.03dB of BD-PSNR degradation and 0.586% of BD-rate increment in terms of the R-D performance.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。