基於三元網路之360度視訊HEC格式之 VVC畫面間位元率控制;Triplet Network Based VVC Interframe Rate Control for 360-degree Videos with HEC Format

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Communication Engineering > Electronic Thesis & Dissertation > Item 987654321/83786

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/83786

Title:	基於三元網路之360度視訊HEC格式之 VVC畫面間位元率控制;Triplet Network Based VVC Interframe Rate Control for 360-degree Videos with HEC Format
Authors:	陳智圓;Chen, Chih-Yuan
Contributors:	通訊工程學系
Keywords:	360度視訊;HEC;VVC;畫面間編碼;位元率控制;三元網路;損失函數;360 degree video;HEC;VVC;Interframe coding;rate control;Triplet Network;loss function
Date:	2020-07-15
Issue Date:	2020-09-02 17:06:22 (UTC+8)
Publisher:	國立中央大學
Abstract:	360度視訊能帶給觀看者身臨其境的體驗，然而，其存在大量資料的儲存與傳送問題，因此360度視訊編碼的位元率控制(rate control)非常重要。現有的360度視訊編碼方案將360度視訊投影至二維平面後再編碼，其中，複合式等角度立方體投影(hybrid equiangular cubemap, HEC)因幾何失真較為均勻，編碼效能較佳。目前最新視訊編碼標準多功能視訊編碼(Versatile Video Coding, VVC)的參考軟體採用基於R-λ模型之位元率控制(rate control)，其編碼樹單元(coding tree unit, CTU)層的位元分配(bit allocation)，根據前一個已編碼且同一畫面階層(frame level)畫面的同位置(co-located) CTU的更新後參數，為當前CTU進行位元分配，但有時無法達到最佳的分配。因此，本論文提出以三元卷積神經網路(Triplet CNNs Network)為基礎的畫面間編碼之位元率控制演算法，並提出同時考量R-λ模型的參數預測誤差與λ預測誤差之損失函數(loss function)，對HEC格式之CTU進行R-λ模型的參數預測。實驗結果顯示，本論文所提方案相較於VVC的參考軟體VTM-3.0的位元率控制方案與現有考量360度視訊投影格式非均勻特性的CTU層位元率控制方案，平均可達更低的位元誤差，但不降低rate-distortion效能。;360-degree videos can provide viewers immersive experience. However, storage and transmission of the large amount of 360-degree videos is a great challenge. Thus rate control plays an important role for 360-degree video coding. In the existing encoders of 360-degree videos, 360-degree videos are projected onto the two-dimensional image plane before encoding. Among projection formats, hybrid equiangular cubemap (HEC) projection enables better coding performance. R-λ model based rate control is adopted by the video coding standard Versatile Video Coding (VVC) reference software. In coding tree unit (CTU) level bit allocation, the updated parameters of R-λ model of the co-located CTU in the last coded frame at the same frame level are adopted for bit allocation of the current CTU. However, these parameters sometimes may not achieve optimal bit allocation. Therefore, this thesis proposes a Triplet Network based interframe rate control scheme to predict parameters of the R-λ model of CTUs for HEC format. In addition, the loss function that considers both prediction errors of parameters of the R-λ model and λ value is proposed. Experimental results show that the proposed scheme outperforms both the original rate control scheme of the reference software of VVC (VTM-3.0) and an existing rate control scheme that considers the unequal characteristics of 360-degree videos on accuracy of bits mismatch without degradation of rate-distortion performance.
Appears in Collections:	[Graduate Institute of Communication Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	206	View/Open

社群 sharing

Loading...