Statistical analysis on coding unit partition of HEVC encoded HD videos

Abstract

H265/HEVC standard is a new video coding standard. It improves digital video encoder's performance while targeting high-resolution formats. Partitioning a large coding unit (LCU) into one or multiple prediction units (PU) with different modes and sizes based on a given rate distortion criteria is the most complex process in the HEVC encoder. In this work, a statistical study on different video classes for LCU partitioning into different PU sizes for each temporal level has been conducted. The obtained results show a strong correlation among the partitioning process, the temporal level coding, and video classes. This study can be considered as a mile stone for eventual algorithmic encoder optimizations in order to meet real time constraints.