Enhanced Cu Partitioning Search Method for Intra Coding in HEVC

Most encoders in High Efficiency Video Coding (HEVC) standard determine the coding unit (CU) partitions in a depth-first full search manner. For intra coding, this method ignores the impact of partitioning on the subsequent blocks, which incurs less accurate prediction and worsens encoding efficiency performance. In this paper, we present an enhanced CU partitioning search method, where the various partitioning types are examined to effectively mitigate such impact. To reduce the additional complexity, we further refine the search candidates, early terminate the search process, and reuse encoding information. Experimental results show that compared with the unmodified HM software, the proposed method achieves a bit rate saving of 0.34% on average and 0.67% at most. Since our method doesn't modify the original definition of HEVC, the encoded bitstreams can be uncompressed by a standard HEVC decoder with no complexity increase.

1. Introduction

Providing a good balance between computational complexity and encoding efficiency, High Efficiency Video Coding (HEVC) is now being chosen by many companies in the video industry. It outperforms the previous video coding standard H.264/AVC by saving nearly 50% bit rate consuming with competitive perceptual quality [1]. Such compression efficiency improvement comes from the newly designed coding tools, such as flexible block partitioning schemes and more precise prediction modes and in-loop filters. Concerning the block partitioning, the quadtree structure in HEVC allows coding units (CUs) to be recursively split into smaller CUs, whose size ranges from pixels to pixels. The leaf node CU can be further split into prediction units (PUs) and transform units (TUs) for prediction and residual coding. In the HEVC reference software HM-16.20 [2], CU partitions are determined by a depth-first full search process, denoted as DFFS in the following. Specifically, the rate distortion cost of each CU is determined by computing and selecting the minimal cost among all possible cases. The computation continues downward until the CU reaches the smallest size. Then, apart from the smallest CU (SCU), the rate distortion cost of each CU is compared to the total full rate distortion cost of its four sub-CUs to decide whether the CU is to be split or not. This is performed upward until the partition of the largest CU (LCU) is finally decided. Figure 1 illustrates the whole DFFS process for a CU.