YOLO-Based Bitrate Control Algorithm for VVC

Abstract

In this work, an adaptive quantization parameter (QP) selection algorithm to each layer of random access mode of VVC is proposed by referring to the Header-bits of each CTU which is estimated by YOLO. Considering the hierarchical structure of random access coding, different approach is used for upper layer and lower layer. To the upper layer, reduced bits are assigned to CTUs without bounding box generated by YOLO by increasing the value of QP. On the other hand, more bits are assigned to CTUs with bounding box generated by YOLO by decreasing the value of QP. The simulation results show that up to 9.69% and 22.28% are achieved in bitrate reduction in the case of QP value of 22 and 32, respectively.