Energy Cost of Coding Omnidirectional Videos using ARM and x86 Platforms

Number of applications involving omnidirectional video's transmission from Unmanned Aerial Vehicle (UAV) are gaining popularity rapidly, ranging from remote surveillance to package delivery. Enabling these applications requires low-latency real-time transmission and a low energy consumption, to achieve a truly immersive experience and a longer flight duration for UAV, respectively. As video encoding and decoding energy costs are two major components of energy consumption in video transmission chain, in this paper, we study the energy cost of encoding and decoding UAV omnidirectional sequences for ARM and x86 based platforms. In addition, we show the difference in energy costs of hardware and software implementations of various video coding standards. The results measure energy cost, coding efficiency and equivalent carbon footprint, for all recent MPEG standards including, Versatile Video Coding (VVC), High Efficiency Video Coding (HEVC) and Advanced Video Coding (AVC), and Alliance for Open Media (AOM) standards, including AOMedia Video 1 (AV1) and VP9. Energy consumption analyses reveal distinct advantages for hardware-based implementations over software-based ones. On the Jetson Orin device, for instance, the AVC and HEVC hardware encoders demonstrate a substantially lower energy consumption, approximately 0.02 Wh. In contrast, hardware implementations at MAC and Dell devices consume 0.24 Wh and 0.16 Wh, respectively. Furthermore, contrasting performance of software implementation at ARM and x86 architectures uncovers significant variations. Remarkably, the VVenC software encoder operates at a pace approximately 4 times faster on arm devices, while, the VP9 and AV1 encoders perform roughly 10 times quicker on x86 devices compared to their speed on the ARM platform.