Multi-Layer Features Fusion Model-Guided Low-Complexity 3D-HEVC Intra Coding

Three-dimensional (3D) video with depth information is essential for many applications in the consumer electronics industry. The 3D-high efficiency video coding (3D-HEVC) is the latest 3D video coding standard. Nonetheless, it utilizes various complex coding techniques to create extra intermediate views for better representation of 3D videos, which imposes significant challenges for real-time 3D video applications. Specifically, the high complexity of 3D-HEVC intra coding could be a significant barrier to the adoption of 3D video in consumer electronics. Therefore, in this research, a low-complexity 3D-HEVC intra coding technique is proposed. Firstly, we perform a complexity analysis of 3D-HEVC intra coding. Secondly, we develop a multi-layer features fusion (MLFF) model to estimate the optimal coding tree unit (CTU) depth and prediction unit (PU) mode. Thirdly, to improve the model’s prediction accuracy, we incorporate two external features into the model: the quantization parameter (QP) and texture complexity. Finally, we embed the MLFF model into the 3D-HEVC test platform. The experimental results demonstrate that the suggested method can effectively reduce the 3D-HEVC intra coding time with a small amount of rate-distortion (RD) performance loss while maintaining the subjective quality of the synthesized view.