EFFICIENT QP CASCADING IN H.265/HEVC LOW-DELAY PREDICTION

The Rate-Distortion (RD) performance of High Efficiency Video Coding (H.265/HEVC) has been greatly benefited from its hierarchical prediction structures, which adopt different Quantization parameters (Qps) in Group-of-Pictures (Gaps) and hierarchical layers. However, these Qps are conventionally fixed and unable to adapt to dynamic video contents. In this paper, we propose a Qp Cascading (QPC) scheme that adaptively allocates Qps to all Gaps and hierarchical layers in Low Delay (LD) and Low-Delay P (LP) predictions. Firstly, we separate the video frames by a structure of virtual-GOP (vGOP) and explore the inter-and intra-vGOP RD dependencies. Then, these dependencies are utilized to formulate a two-stage Qp cascading scheme: inter-vGOP quantization assignment and intra-vGOP QPC. Comprehensive experiments show that the proposed scheme greatly improves the coding performances of all Common Test Condition (CTC) classes, which outperforms typical fixed and adaptive QPC schemes that have been developed recently.