CTU-Level Lagrange Multiplier and QP Adaption for Random Access Hierarchical Coding in VVC

Versatile video coding (VVC) adopts a hierarchical coding structure to exploit the inherent temporal rate-distortion (R-D) dependency resulting from inter-prediction, but the scheme does not support at the coding tree unit (CTU)-level. In this paper, we present a CTU-level Lagrange multiplier and quantization parameter (QP) adaptation method in VVC random access (RA) configuration, aimed at enhancing compression performance. Firstly, frames in a group of pictures (GOP) are pre-encoded with different schemes, aiming to collect the coding distortion and motion compensation predicted (MCP) error of frames within the highest layer. Secondly, the distortion effect factor of a $16 \times 16$ block in frames within other layers is estimated based on the coding distortion and MCP error obtained. Finally, the adaptive Lagrange multiplier and QP at the CTU-level are determined by considering the distortion effect factor and the hierarchical frame-level QP setting. Experimental results show that, on the platform of VVC test model (VTM21.2), the proposed algorithm achieves an average 0.9% bit rate saving with a 10% increase in encoding time, under the RA configuration.