Complexity Reduction In H.264 Encoder Using Open Multiprocessing

Abstract

H.264 video standard developed by Joint Video Team has proven dramatic improvements in bit-rate efficiency, compression ratio, video quality and error resilience. But, all this is achieved at the expense of more than four times of the computational complexity due to various new features including quarter-pixel motion estimation with variable block sizes and multiple reference frames, adaptive directional intra-prediction, integer transformation based on discrete cosine transform, alternative entropy coding mode Context-based Adaptive Variable Length Coding (CAVLC) or Context-Based Adaptive Binary Arithmetic Coding (CABAC), in loop de-blocking filter etc. This thesis aims at reducing the encoding time for the video sequence while maintaining the same quality and compression efficiency using parallel processing approach. Entire video sequence is divided into four groups having one I frame each. In this thesis, the original JM software code, written in serial manner is enhanced in such a way that the encoding of all individual parts of the original video is implemented in parallel using the threads that are managed by Open Multiprocessing runtime system which shows more than 60% encoding time reduction.JM 18.0 reference software is used in this thesis for implementing H.264 video encoder. The reference software along with manual is available at http://iphome.hhi.de/suehring/tml. The software manual includes information about the H.264 encoder and decoder input parameters, syntax, compilation issues, and additional information with regards to the best usage and configuration of the software.