Analysis of the Capabilities of Embedded Systems in Intraprediction Video Coding

With the increased use of mobile devices, and embedded systems, battery life has become of vital importance. Thus, in the video coding field, energy consumption has joined compression efficiency and computational complexity as a key parameter. Energy consumption is one of the most important factors to consider in any video-intensive application due to its relationship to the autonomy of these devices. This article aims to carry out a performance and temperature analysis of two embedded systems (Raspberry Pi 4 and Jetson Nano), which are used in multiple applications and fields, including video encoding, and a consumption study of different encoders of various video standards executed on them. The analysis will consider parameters such as the compression rate, the video quality using an objective metric (PSNR), and the computing time over an intrascenario. In this study, three software encoders and three hardware encoders implementing different video standards have been analyzed: x264 (H.264/AVC), x265 (HEVC), rav1e (AV1), omxh264enc (H.264/AVC), nvv4l2h264enc (H.264/AVC), and nvv4l2h265enc (HEVC). Finally, the results demonstrate that real-time video streaming using hardware encoders is only achievable for the full-HD resolution format.