INCEPT: Intra CU Depth Prediction for HEVC

High Efficiency Video Coding (HEVC) improves the encoding efficiency by utilizing sophisticated tools such as flexible Coding Tree Units (CTUs) partitioning. The Coding Units (CUs) can be split recursively into four equally sized CUs ranging from 64×64 to 8×8 pixels. At each depth level (or CU size), intra prediction via exhaustive mode search was exploited in HEVC to improve the encoding efficiency and result in a very high encoding time complexity. This paper proposes an Intra CU Depth Prediction (INCEPT) algorithm, which limits Rate-Distortion Optimization (RDO) for each CTU in HEVC by utilizing the spatial correlation with the neighboring CTUs, which is computed using a DCT energy-based feature. Thus, INCEPT reduces the number of candidate CU sizes required to be considered for each CTU in HEVC intra coding. Experimental results show that the INCEPT algorithm achieves a better trade-off between the encoding efficiency and encoding time saving (i.e., BDR/∆T) than the benchmark algorithms. While BDR/∆T is 12.35% and 9.03% for the benchmark algorithms, it is 5.49% for the proposed algorithm. As a result, INCEPT achieves a 23.34% reduction in encoding time on average while incurring only a 1.67% increase in bitrate than the original coding in the x265 HEVC open-source encoder.

I. Introduction

The share of video on the Internet traffic is expected to reach 82% by 2022 [1], which raises the need for highly efficient video codecs. The compression efficiency of HEVC has been improved greatly with the adoption of numerous advanced tools [2]. HEVC introduces a more flexible quad-tree coding block partitioning structure and uses the Coding Units (CUs) and Prediction Units (PUs) concept to manage the partition [3]. In HEVC, each video frame is divided into non-overlapping predefined size Coding Tree Units (CTUs), and each CTU is then recursively divided into CUs. CUs can be of various sizes, namely, 64×64, 32×32, 16×16 which and 8×8 pixels, correspond to CU depth levels 0, 1, 2 and 3, respectively. A CU can be further partitioned into PUs and Transform Units (TUs). The partitioning is recursive within a quad-tree hierarchy. PU is obtained by partitioning a CU, and it is used for the prediction. To determine the best coding mode for each PU, all candidate modes are examined by calculating the RD cost. Finally, based on the RD costs, optimal CTU partitioning is obtained. A TU is a square block, which shares the same transform and quantization processes. A TU tree structure has its root at the CU level. The TU shape is always square, and it may take a size from 32×32 to 4×4 samples. Fig. 1 exemplary shows an optimal CTU partitioning into CUs with its corresponding quad-tree structure.