一种对R-Lambda码率控制模型的改进算法  

Improvement for R-Lambda Rate Control Model

作　　者：王朋[1] 倪翠[1] 张广渊[1] 李克峰[1] 

机构地区：[1]山东交通学院信息科学与电气工程学院,济南250357

出　　处：《应用科学学报》2017年第3期355-365,共11页Journal of Applied Sciences

基　　金：国家自然科学基金(No.61502277);山东省自然科学基金(No.ZR2015FL017)资助

摘　　要：最新的视频编码标准HEVC的参考代码通常采用rate-Lambda(R-Lambda)模型进行码率控制,但R-Lambda模型为I帧分配的目标码率过多,导致后续视频帧因目标码率不足而重建质量严重下降.为此,对R-Lambda模型进行改进,根据图像的运动趋势和复杂度将I帧划分为运动区域、纹理复杂区域、纹理平坦区域.分配目标码率时,根据人眼视觉系统特性为视觉敏感区域分配更多码率,给非视觉敏感区域分配较少码率,最后利用R-Lambda模型计算相应的QP值.实验结果表明,所提算法在保证重建视频主观质量连续性的同时,提高了重建视频的整体质量.In the reference code of high efficiency video coding, the R-Lambda model is often used for rate control. However, the rate-Lambda （R-Lambda） allocates too much target bits to I-frames, causing serious quality decline in the following frames. To solve the problem, improvement for the R-Lambda model is proposed in this paper. An I-frame is segmented into motion regions, texture-rich regions and smooth regions according to the motion characteristics and complexity. The target bits are allocated to different regions based on the characteristics of human vision system. More target bits are allocated to the visual sensitive regions, and fewer allocated to the visual non-sensitive regions. Finally, the R-Lambda model is used to calculate quantization parameter （QP）. Experimental results show that the proposed algorithm can guarantee uniformity of subjective quality, and improve overall quality of the reconstructed frames.

关 键 词：R-Lambda模型 目标码率 区域划分 重建质量 

分 类 号：TN919.8[电子电信—通信与信息系统]