高效率視訊編碼之運算複雜度分配與控制;Computational Complexity Allocation and Control for HEVC Encoder

NCU Institutional Repository > 資訊電機學院 > 通訊工程研究所 > 博碩士論文 > Item 987654321/65529

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/65529

題名:	高效率視訊編碼之運算複雜度分配與控制;Computational Complexity Allocation and Control for HEVC Encoder
作者:	謝瑋;Hsieh,Wei
貢獻者:	通訊工程學系
關鍵詞:	HEVC;複雜度分配;複雜度控制;編碼單位;HEVC;complexity control;complexity allocation;coding unit
日期:	2014-07-31
上傳時間:	2014-10-15 17:02:52 (UTC+8)
出版者:	國立中央大學
摘要:	新一代視訊壓縮標準High Efficiency Video Coding (HEVC)為近年JCT-VC團隊所制定，相較於過去的壓縮標準，HEVC提供眾多新壓縮技術來達到更高的編碼效率，使其更適用於高解析度的視訊，相對地也大幅增加編碼端的運算複雜度。然而在行動裝置上受限於電源容量，可用的運算複雜度通常是有限的，因此，降低並控制視訊編碼器的運算複雜度以延長行動裝置的使用時間，並且維持較佳的視訊位元率-失真效能是非常重要的議題。本論文針對HEVC高解析度視訊於運算資源有限的裝置上，提出一套編碼運算複雜度分配與控制機制，包含了畫面層級、編碼單位(Coding Unit, CU)層級之分配與控制，以及預測單位(Prediction Unit, PU)編碼流程改善。首先在畫面層級我們根據每張畫面的量化參數(QP)給予其不同的運算複雜度。接著參考前一張畫面的平均絕對誤差 (Mean Absolute Difference, MAD)將畫面層的運算複雜度分配至最大編碼單位(Largest CU, LCU)層級，並參考空間域與時間域鄰近編碼單位的深度資訊來刪減不必要的編碼單位運算複雜度。在各層編碼單位中，基於分析所得的預測單位編碼增益(Coding Gain, CG)重新修改預測單位編碼流程。本論文所提出的方法可同時達成總體運算複雜度與瞬時運算複雜度限制之應用，實驗結果顯示，在節省40%運算量之目標下，針對總體複雜度與瞬時複雜度之控制，分別可達成平均BD-PSNR僅0.2 dB與0.1 dB損失之效能以及僅有0.3%與0.5%之運算量控制誤差。 ;The latest video compression standard HEVC was established by the Joint Collaborative Team on Video Coding (JCT-VC) and provided various encoding tools to achieve high coding efficiency in comparison to previous standards at the cost of higher computational complexity. However, the allowable computational capability of a portable device for real-time video encoding is generally constrained. Therefore, a complexity control mechanism that well allocates the computational complexity of video encoding under the complexity constraint while maintaining optimal rate-distortion performance is important. Therefore, we propose a computational complexity mechanism for high resolution video on a power-constrained device, including computational complexity allocation and control from frame layer to largest coding unit (LCU) layer, and the rearrangement of prediction unit (PU) encoding flow. In frame layer, we allocate different computational complexity based on the quantization parameter (QP) of each frame. In LCU layer, the mean absolute difference (MAD) of previous frame is used to allocate suitable complexity to each LCU. By referring neighboring and co-located depth information of CU, CU depth 0 is skipped in certain situation. And PU encoding flow is arranged based on the coding gain (CG) analyses. The proposed method could simultaneously achieve the entire complexity constraint (ECC) for a sequence and instant complexity constraint (ICC) for real-time encoding. The experimental results show that under a 60% target computational complexity, the loss of average BD-PSNR is negligible and the complexity control error is no more than 0.5%.
顯示於類別:	[通訊工程研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	383	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....