VVC Based Rate Control Using SKIP CTU Predictor

Versatile Video Coding (VVC) enables high coding gain on 360-degree equiangular cubemap (EAC) projection format. Nevertheless, the R-$\boldsymbol{\lambda}$ model based rate control of VTM cannot accurately estimate $\boldsymbol{\lambda}$ of SKIP CTUs so that the bitrate error is increased. Thus this paper proposes a SKIP CTU predictor for VVC interframe rate control. Different from the existing SKIP block ratio based rate control scheme, the predictor runs before encoding the current CTU. It predicts whether the current CTU has high percentage of SKIP CUs or not instead of predicting whether the optimal mode of a CU is SKIP mode or not. The value of CTU-level $\boldsymbol{\lambda}$ of the predicted SKIP CTU is estimated using $\boldsymbol{\lambda}$ values in spatially adjacent CTUs to decrease the bitrate error. The lightweight neural network based predictor has only 199 parameters. Bit starvation in later coded frames is reduced by constraining frame-level quantization parameters in intra frames. Tests show that the proposed scheme outperforms state-of-the-art rate control methods on 360-degree 8K and 6K EAC videos on bitrate accuracy.