Approximate computing based video compression hardware

Official URL: https://risc01.sabanciuniv.edu/record=b2766754_(Table of contents)

Approximate computing trades off accuracy to improve area, power, speed of digital hardware. Many computationally intensive applications such as video encoding and video processing are error tolerant by nature due to the limitations of human visual perception. Therefore, approximate computing can be used to improve area, power, speed of digital hardware implementations of these error tolerant applications. In this thesis, a low error approximate absolute difference hardware is proposed and impact of using approximate circuits in H.264 motion estimation (ME) hardware is assessed. There is a need for more video compression with less quality loss due to significant increase in spatial and temporal video resolutions. Therefore, versatile video coding (VVC) video compression standard is recently developed. It is more computationally complex than previous video compression standards H.264 and high efficiency video coding (HEVC). ME is the most computationally complex part of video compression standards. VVC standard uses affine ME (AME) which is not used in previous video compression standards. AME achieves higher video compression at the expense of much more computational complexity. In this thesis, a novel VVC AME hardware is proposed.