A New Adaptive Lagrange Multiplier Selection Model in Hybrid Video Coder Control

The equation λ=cQ2 has been widely used as a parameter selection model in hybrid video coder control with Lagrange rate-distortion optimization, which is called square model (SM) and deduced with an assumption of λ being constant. The coefficient c is obtained from statistic in different video coding recommendations. However, according to the theoretical analysis, λ shouldn't be seen as a constant. A proposition about this idea is proposed and proved by mathematically deducing and experimental justification. Based on the proposition, a new Lagrange multiplier selection model called quadratic-logarithm model (QLM) is deduced and its parameter is adaptively obtained by the scheme proposed in the paper. Finally, QLM is used in H.264 and a new rate-distortion scheme is proposed. The final simulation result shows that compared with SM, the new multiplier selection model can obtain better rate-distortion performance in general sense (0.03dB ave.) without any further extra computation. Therefore, the multiplier model proposed is more feasible and reasonable than the existing ones.