The Optimal Real-Time Video Encoding Scheme for Multi-Camera Panoramic System Using NVENC

Abstract

Today, modern, powerful, and relatively available computer technology allows the proposal of real-time camera systems working with several video streams. There is also a permanent tendency to increase image resolution, what can be useful for example in some image processing methods used in surveillance systems. Given this, an important part of a multi-camera real-time system proposal is optimal and efficient video compression. The modern encoding standards support 4K or 8K video content. One of these standards is H.264 or HEVC (High Efficiency Video Coding) and the popular available encoding technology using GPU (Graphics Processing Unit) acceleration is NVENC (Nvidia Encoding) provided by Nvidia. Many papers deal with the comparison of encoding standards in terms of quality or bitrate savings on testing video sequences [1], [2]. However, this paper is aimed primarily at a comparison of several practical implementations of NVENC technology using different presets of H.264 and HEVC standards. The experiments are mainly focused on testing the encoder’s real-time ability to deal with single video input obtained from the industrial camera using different settings. To do that, some system performance parameters for the continuous monitoring during the NVENC encoding process were chosen. Results in tables I and II are compared, discussed and the best NVENC implementation with encoding scheme is chosen in the context of our briefly introduced novel panoramic system. Finally, the system is fully loaded by the 7 cameras, all streams are encoded using an optimal encoding scheme and a performance limit is established for our hardware configuration.