Mitigating RDO and MEC Complexity in H.26x Video Codecs using DWT

Résumé fourni par la source

The current rise of audiovisual content demand needs dynamic approaches for content compression. In video compression, the weakness of human visual system (HVS) is exploited to reduce spatial and temporal redundancies, hence reducing inherent bit content in video frame sequences. Nevertheless, processing involved pose new challenges of computation cost in terms of time, power and system design complexity especially in real-time video streaming. Mitigating the trade-off between compression ratios, video quality and computation cost continues to be the core area of research in video coding. In H.26x codecs, motion estimation (ME) and motion compensation (MC) algorithms, sometimes simply called motion estimation and compensation (MEC) are used in inter-frame prediction. Conventionally, MEC involves time consuming searching algorithms in macroblock and sub-block matching. Also, H.26x series employ rate distortion optimisation (RDO) in intra prediction (IP) modes which is repetitive and exhaustive. In this paper, we present two approaches to mitigate MEC and RDO complexities in H.26x codecs: (a) we present a Discrete Wavelet Transform-assisted intra prediction (DWTIP). DWTIP avoids exhaustive and blind evaluation of all possible intra prediction modes by restricting specific modes to a specific macroblock and sub-block size predetermined based on pixel homogeneity levels. This leads RDO relaxation. (b) We also present a DWT-assisted motion estimation and compensation (DWTMEC) for inter-frame prediction. DWTMEC limits the search window to DWT approximation sub-band and reduces the search area, hence reducing MEC computation time. Objective video quality metric is used to compare conventional MEC with DWTMEC and conventional IP with and DWTIP. Results showed that employing DWTMEC along with DWTIP, video coding rates are improved with negligible degradation in video quality.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.