Low-Power and High-Throughput Approximated Architecture for AV1 FME Interpolation

Abstract

Modern video encoders like the AOM Video 1 (AV1) implement several complex tools to allow the required high level of compression efficiency. The Fractional Motion Estimation (FME) is one of these tools and in AV1 the FME defines 90 different filters. To handle such complexity, hardware acceleration using approximate computing has become an alternative to be explored. This paper presents an approximate solution for the AV1 FME interpolation filters based on the approximation of the original filter coefficients intending to generate more hardware friendly coefficients. The approximated version was designed in hardware and can achieve real-time interpolation for UHD 8K videos at 30 frames per second, when synthesized using 40nm TSMC standard-cells technology. The designed architecture dissipates 26.79mW which represents more than 80% power reduction when compared to the original precise solution. The approximation implied in a small average coding efficiency degradation of 0.54% in BD-BR. When comparing with related works, this architecture reaches an expressive power reduction (2.1 to 4.8 times) even supporting more complex tools.

Ultra-High Definition AV1 FME Interpolation Architectures Exploring Approximate Computing

“…The work [16] presents an architecture for the AV1 Motion Compensation (MC), the work [17] presents an architecture for the VP9 FME, and the work Digital Object Identifier 10.29292/jics.v17i2.558 [18] presents an architecture for the HEVC FME. The previous works of the authors of this paper, presented in [19] and in [20], show approximate architectures for the AV1 FME and, to the best of the authors' knowledge, these are the first hardware designs in the literature exploring approximation computing for the AV1 FME.…”

“…This work proposes three approximate architectures for the AV1 FME tool, extending the authors' previous work [19] and [20]. The approximations are applied at both algorithmic and data levels.…”

“…The usage of approximate computing in video coding is not a novelty, and this approach has been widely explored, in other related works targeting interpolation filters, like in [16], [17], [18], and [19]. The work [16] presents an architecture for the AV1 Motion Compensation (MC), the work [17] presents an architecture for the VP9 FME, and the work Digital Object Identifier 10.29292/jics.v17i2.558 [18] presents an architecture for the HEVC FME.…”

“…The data approximations reduce the number of filter taps, which leads to reductions in the data required to compute the interpolated samples. The novelties of this paper concerning the previous works [19] and [20], besides the extended discussion about the topic and the developed solutions, are related to the number of taps used in the interpolation filters. The solution in [19] does not reduce the interpolation filter taps and the solution in [20] applies an aggressive reduction in the number of filter taps.…”

“…The novelties of this paper concerning the previous works [19] and [20], besides the extended discussion about the topic and the developed solutions, are related to the number of taps used in the interpolation filters. The solution in [19] does not reduce the interpolation filter taps and the solution in [20] applies an aggressive reduction in the number of filter taps. The new version of the interpolators presented in this paper best balances the power and area gains and the coding efficiency losses reached with the reduction of the filter taps.…”