Parallel Optimization Of Intra Mode Selection In HEVC Using Open MP

Abstract

HEVC aims at providing an efficient video coding algorithm which not only provides high performance, but also improves coding efficiency. It also provides high compression efficiency as compared to its counter parts, but at the cost of an increase in complexity. The increased number of intra prediction modes in HEVC allows better compression and more flexible block representation. The purpose of the thesis is to optimize the mode decision process using Open MP to reduce the encoding time without further increase in complexity. This includes implementing a parallel computing method that will calculate 35 distinct intra prediction modes in parallel and compare them to decide the three best possible modes depending on RD cost values as compared to a single mode in H.264. There are over 11900 different intra modes according to different PB size as shown in Table 5-1 which provides higher level compression but at the cost of increase in the encoding time . Experimental results were compared with the HEVC reference software HM 9.1 for different standard test sequences with respect to various quantization parameters. The proposed algorithm was evaluated using different metrics. These include encoding time, BD- PSNR (Bjontegaard Delta Peak Signal to Noise Ratio) and BD-bitrate (Bjontegaard Delta bitrate), rate distortion (RD curve) graph and percentage reduction in encoding time. There is a 22-57% reduction in encoding time compared to reference software calculations with negligible increase in bitrate and negligible decrease in PSNR.