【摘要】：率失真優(yōu)化(RDO)是視頻編碼壓縮中的關鍵技術之一。傳統(tǒng)的率失真優(yōu)化技術使用誤差平方和(SSE)或絕對差和(SAD)度量失真,不能獲得較好的視覺感知質量。本文針對新一代視頻編碼標準高效視頻編碼(HEVC)幀內編碼RDO,提出了一種基于結構相似度(SSIM)度量失真的RDO方法。首先,根據(jù)HEVC編碼樹形單元(CTU)結構設計了基于SSIM的失真計算方法;然后,提出了一種基于量化器推導的碼率-量化步長(R-Δ)模型和一種基于統(tǒng)計分析的失真-量化參數(shù)(DSSIM-QP)模型,用于求取拉格朗日乘數(shù);最終,使用多QP優(yōu)化方法求取模型參數(shù)。實驗表明,相對于HEVC傳統(tǒng)的RDO,針對全I幀(即所有幀都為幀內編碼)固定QP編碼和多QP優(yōu)化編碼,在相同的SSIM條件下,碼率分別平均降低8.4%和13.9%左右,同時編碼復雜度分別平均增加約3%和2%。
[Abstract]:Rate-distortion optimization (RDO) is one of the key techniques in video coding and compression. Traditional rate-distortion optimization techniques use (SSE) or absolute error and (SAD) metric distortion, which can not achieve better visual perception quality. In this paper, a novel RDO method based on structural similarity (SSIM) metric distortion is proposed for the new generation of efficient video coding (HEVC) intra coding. Firstly, according to the (CTU) structure of HEVC coding tree unit, the distortion calculation method based on SSIM is designed, then a rate-quantization step size (R- 螖) model based on quantizer and a distortion-quantization parameter (DSSIM-QP) model based on statistical analysis are proposed. It is used to obtain Lagrange multiplier, and finally, multiple QP optimization method is used to obtain model parameters. The experiments show that compared with the traditional HEVC, the fixed QP coding and multi-QP optimization coding for all I frames (that is, all frames are intra coding) can reduce the code rate by about 8.4% and 13.9%, respectively, under the same SSIM conditions. At the same time, the coding complexity increases by about 3% and 2% respectively.
【作者單位】： 寧波大學信息科學與工程學院;
【基金】：國家自然科學基金(U1301257,61171163,61271270,61271021,61311140262)資助項目
【分類號】：TN919.81

【參考文獻】

相關期刊論文 前3條

1 楊春玲;肖冬琴;;基于SSE和SSIM的H.264幀內預測模式選擇改進算法[J];電子與信息學報;2011年02期

2 陳云善;高慧斌;蘇宛新;王e，

本文編號：2172660