不依赖于量化参数的视频编码控制优化技术

摘要

目前,混合式视频编码技术普遍采用基于量化参数的Lagrange乘子优化算法(QP-L)实现预测编码的率失真优化计算.文中则提出一种新的率失真优化算法Adapt-L,可以根据视频信号的率失真统计特性,自适应地调整Lagrange乘子的计算.实验及数据分析表明,在QP-L算法的Lagrange乘子和视频信号的率失真统计信息之间存在线性关系；由此,提出一个启发式的Lagrange乘子计算模型,并对该数学模型的合理性进行理论验证.大量实验证明,Adapt-L的编码效率优于基于量化参数的优化算法(QP-L).虽然Adapt-L相对稍为复杂,但其计算不依赖于量化参数,有利于实现高效的码率控制.%Currently a QP-based Lagrange Multiplier (QP-L) method is widely used to achieve Rate-Distortion optimal joint mode selection and motion estimation in hybrid video coding. In this work, a novel cost function for RDO is presented, which is independent of QP and adaptable to the statistics. Firstly, it is shown there be a linear relationship between of the QP-L method and some functionality about statistic for distortion and rate. Thusly, a heuristic model is presented to determineXwith consideration of statistic for distortion and rate, which is after then theoretically verified. It is shown via extensive experiments that the proposed scheme achieves a better coding performance than the traditional QP-based scheme. Moreover, the proposed scheme promises an easy encoding process, due to the fact that it's computation does not involve QP.