Adaptive λ Estimation in Lagrangian Rate-Distortion Optimization for Video Coding

摘要

In this paper, adaptive Lagrangian multiplier λ estimation in Larangian R-D optimization for video coding is presented that is based on the ρ-domain linear rate model and distortion model. It yields that λ is a function of rate, distortion and coding input statistics and can be written as λ(R, D, σ~2) = β(ln(σ~2/D) + δ)D/R + k_0, with β, δ and k_0 as coding constants, σ~2 is variance of prediction error input. λ(R, D, σ~2) describes its ubiquitous relationship with coding statistics and coding input in hybrid video coding such as H.263, MPEG-2/4 and H.264/AVC. The lambda evaluation is de-coupled with quantization parameters. The proposed lambda estimation enables a fine encoder design and encoder control.