A Real-Time UHD 4K Hardware for VVC Affine Linear Equation System Solving

The affine prediction is the main innovation in the inter-frame prediction of the Versatile Video Coding (VVC) standard. However, integrating this feature demands substantial computational resources, which makes dedicated hardware mandatory for real-time processing while adhering to mobile device constraints in terms of area and power dissipation. In this context, this paper introduces a new approach for an efficient and high-performance hardware design of the linear equation system solving using the Gauss-Jordan Elimination method, which is a crucial step in the search for affine motion vectors in the VVC standard. By applying our proposed approach, the developed hardware architecture can solve the linear equation system dissipating 299.94mW within an area of 434.97k gates, when processing UHD 4K@60fps videos. When compared to related work, the proposed architecture can reduce the power dissipation by 69.9% and the area requirements by 27%, clearly demonstrating the efficiency of the proposed approach.