Complexity Reduction for HEVC
Duration: May - July 2014
Status: Completed
Members: Pranav Sodhani, Nan Hu, Dr. En hui-Yang
Note: The project was carried out during the summer break after my junior year. This 12-weeks long internship at the University of Waterloo was funded by Mitacs Globalink Program.
Overview
The successor to present video coding standard - H.264 is H.265 or more popularly known as HEVC (High Efficiency Video Coding). Jointly developed by the ISO/IEC and the ITU-T, HEVC reduces bit rate by almost 40% over H.264/AVC with the same objective quality but at about 40% encoding complexity overhead. This project aims to develop algorithms which reduce complexity of HEVC while maintaining the bit rate and picture quality. For this, a fast skip mode scheme for HEVC is presented. The proposed algorithm predicts skip mode for depth 0 (64x64) using residual energy information obtained from Rough Motion Estimation (RME). When tested on Class B and Class C video sequences, experiment results demonstrate an encoding complexity reduction of 43% accompanied by a 0.85% increase in BD rate on an average when compared with the HEVC test model HM 14.0.
The skip mode predictions for depth 1, 2 and 3 based on residual energy information from the RME are worthy of further investigations and will be continued as a part of the future work.