A Pilot Exploration of Industrial Video Scene Data Embedding using Real-Time MV-HEVC

Video sequences can be coded under Multi-view High Efficiency Video Coding (MV-HEVC) standard efficiently by using simultaneous captured multiple camera views to a single video stream to achieve a real time stereoscopic view. As a successor of Multi-view Video Coding in Advanced Video Coding, MV HEVC performs better compression with bitrate saving up to 50% for multi-view video compared to multiple independent views of earlier coding standard. In this work, we explore the entities of multi-view video compression and decompression processes, investigating the prediction mode manipulation for serving data embedding purpose, and tested on Linux-based Operating System to achieve a near real-time video streaming with data embedding capability. The performance of MV-HEVC with three views streaming are evaluated based on industrial machines views, as a part of Industry 4.0 implementation and compared to the benchmark data sets. The experimental results indicate a minor change in the BjØntegaard-Bitrate and peak signal-to-noise ratio scales, at minimum 0.0039 % and 0.0013dB respectively. Besides, the proposed method also compared to multi-view simulcast HEVC coding in term of execution time, with the average difference ranging from 0.1995 to 3.8839% of additional time required for MV video stream with data embedding application.