Directional Intra Frame Prediction Architecture with Edge Filter and Upsampling for AV1 Video Coding

  • Luiz Neto UFPel
  • Marcel Corrêa UFPel
  • Daniel Palomino UFPel
  • Luciano Agostini UFPel
  • Guilherme Correa UFPel

Resumo


Video compression following the AOMedia Video 1 (AV1) format has become more popular since the bitstream specification launching in 2018. However, although the reference AV1 codec libaom achieves much better compression efficiency in comparison to predecessors, its high computational cost forbids use for real-time compression. To allow for real-time operation, hardware accelerators for the most complex AV1 tools, such as the intra-frame prediction, are required. This paper presents a highly parallelized ASIC solution for the directional intra-frame prediction of AV1, which includes the pre-processing steps of edge filtering and upsampling and the directional intra-frame prediction with support to all nominal angles. The proposed hardware design was synthesized for the TSMC 40nm technology for the target frequency of 476MHz required for real-time operation. This scenario led to a required area of 691.742K gates, a power dissipation of 382.076mW and an energy consumption of 2.047nJ/sample. To the best of the authors’ knowledge, this is the first work published in the literature presenting a hardware solution for the directional intra-frame prediction of AV1, including the pre-filtering stages and the support to all nominal angles.
Palavras-chave: Finite impulse response filters, Streaming media, Hardware, Encoding, Video coding, Computer architecture, Video Coding, AOMedia Video 1, AV1, Hardware Design, Intra Prediction
Publicado
24/08/2020
Como Citar

Selecione um Formato
NETO, Luiz; CORRÊA, Marcel; PALOMINO, Daniel; AGOSTINI, Luciano; CORREA, Guilherme. Directional Intra Frame Prediction Architecture with Edge Filter and Upsampling for AV1 Video Coding. In: SYMPOSIUM ON INTEGRATED CIRCUITS AND SYSTEMS DESIGN (SBCCI), 33. , 2020, Evento Online. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2020 . p. 133-138.