摘要

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 λ of the QP-L method and some functionality about statistic for distortion and rate. Thusly, a heuristic model is presented to determine λ with 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.

全文