Akshay Shashidhara Nagaraghatta
Algorithms and methods for video transcoding.
Nagaraghatta, Akshay Shashidhara
Video transcoding is the process of dynamic video adaptation. Dynamic video adaptation can be defined as the process of converting video from one format to another, changing the bit rate, frame rate or resolution of the encoded video, which is mainly necessitated by the end user requirements. H.264 has been the predominantly used video compression standard for the last 15 years. HEVC (High Efficiency Video Coding) is the latest video compression standard finalised in 2013, which is an improvement over H.264 video compression standard. HEVC performs significantly better than H.264 in terms of the Rate-Distortion performance. As H.264 has been widely used in the last decade, a large amount of video content exists in H.264 format. There is a need to convert H.264 video content to HEVC format to achieve better Rate-Distortion performance and to support legacy video formats on newer devices. However, the computational complexity of HEVC encoder is 2-10 times higher than that of H.264 encoder. This makes it necessary to develop low complexity video transcoding algorithms to transcode from H.264 to HEVC format. This research work proposes low complexity algorithms for H.264 to HEVC video transcoding. The proposed algorithms reduce the computational complexity of H.264 to HEVC video transcoding significantly, with negligible loss in Rate-Distortion performance. This work proposes three different video transcoding algorithms. The MV-based mode merge algorithm uses the block mode and MV variances to estimate the split/non-split decision as part of the HEVC block prediction process. The conditional probability-based mode mapping algorithm models HEVC blocks of sizes 16×16 and lower as a function of H.264 block modes, H.264 and HEVC Quantisation Parameters (QP). The motion-compensated MB residual-based mode mapping algorithm makes the split/non-split decision based on content-adaptive classification models. With a combination of the proposed set of algorithms, the computational complexity of the HEVC encoder is reduced by around 60%, with negligible loss in Rate-Distortion performance, outperforming existing state-of-art algorithms by 20-25% in terms of computational complexity. The proposed algorithms can be used in computation-constrained video transcoding applications, to support video format conversion in smart devices, migration of large-scale H.264 video content from host servers to HEVC, cloud computing-based transcoding applications, and also to support high quality videos over bandwidth-constrained networks.
NAGARAGHATTA, A.S. 2019. Algorithms and methods for video transcoding. Robert Gordon University [online], PhD thesis. Available from: https://openair.rgu.ac.uk
|Jan 28, 2020
|Publicly Available Date
|Jan 28, 2020
|H.264 to HEVC video transcoding; Video transcoding; Split/non-split decisions; Mode mapping; Mode merging
NAGARAGHATTA 2019 Algorithms and methods for video transcoding
Publisher Licence URL
Copyright: the author and Robert Gordon University
You might also like
Pipeline leakage detection and characterisation with adaptive surrogate modelling using particle swarm optimisation.
A low-complexity wavelet-based visual saliency model to predict fixations.
A fuzzy cooperative localisation framework for underwater robotic swarms.