中圖分類號： TN919.8 文獻標識碼： A DOI：10.16157/j.issn.0258-7998.211453 中文引用格式： 余東航,，李強，聶駿. 基于視頻內(nèi)容特征值的VVC碼率控制算法[J].電子技術(shù)應用,，2021,，47(12)：89-93. 英文引用格式： Yu Donghang，Li Qiang,，Nie Jun. Versatile video coding rate control algorithm based on content feature value[J]. Application of Electronic Technique,，2021，47(12)：89-93.
Versatile video coding rate control algorithm based on content feature value
Yu Donghang,，Li Qiang,，Nie Jun
School of Communication and Information Engineering，Chongqing University of Posts and Telecommunications,， Chongqing 400065,，China
Abstract： Aiming at the problem that the VVC rate control algorithm does not comprehensively consider the actual texture characteristics of the coded frame, and its rate-distortion performance and visual experience need to be improved, a rate control algorithm based on video content-related feature values is proposed. By introducing the gray level co-occurrence matrix, the relevant feature values reflecting the texture complexity of the encoded frame are calculated, which are used to adjust the image weight of the frame layer; based on the R-λ model, the λ parameter of the LCU layer is recalculated to improve the bit allocation of the LCU layer Accuracy. After testing, compared with the adaptive bit allocation algorithm in the low-delay configuration, the algorithm in this paper is closer to the target bit rate, the rate-distortion performance is improved by 0.86%, and the subjective quality of video coding is significantly improved.
Key words : versatile video coding；rate control,；gray-level co-occurrence matrix,；R-λ model；bit allocation
0 引言
多功能視頻編碼(Versatile Video Coding,，VVC)是由ISO/IEC的MPEG和ITU-T的VCEG聯(lián)合制定,，于2020年7月正式發(fā)布的新一代視頻編碼標準。與以前標準相比,，VVC具有更高的壓縮性能和通用性,，可應用于高清和超高清視頻、360°全向視頻,、高動態(tài)視頻范圍和廣色域,、沉浸式媒體等多種應用場景。
VVC碼率控制算法仍沿用了高效視頻編碼(High Efficiency Video Coding,，HEVC)中的R-λ模型[1],。基于該模型的碼率控制算法具有控制效果好,、比特波動小的優(yōu)點,，但對局部紋理復雜邊緣輪廓顯著的視頻，最大編碼單元(Largest Coding Unit,，LCU)層的比特分配并不準確，導致峰值信噪比有所損失,。針對該模型碼率控制算法的不足,，文獻[2]和[3]以LCU層的梯度值衡量編碼區(qū)域的復雜度,，用于指導R-λ模型中LCU的比特分配，提升了R-λ模型的性能,；文獻[4]對LCU層編碼模型參數(shù)做偏差修正處理,，重新計算更新待編碼LCU的權(quán)重分配，提升了LCU層的比特精確度,；文獻[5]根據(jù)視頻幀間相關(guān)性,，參考前一幀R-λ模型更新后計算得到的比特數(shù)預測待編碼幀的復雜度，提高了視頻編碼質(zhì)量,；文獻[6]將I幀消耗大量比特的影響分攤到整個序列,，從而改進了前幾個圖像組(Group of Pictures，GOP)比特分配量,，并在GOP編碼過程中根據(jù)實際消耗的比特不斷更新幀層的λ參數(shù),，提高了后續(xù)幀的比特分配精確度。
