In Internet Protocol TeleVision (IPTV) systems, the use of video compression results in a trade-off between random access and bandwidth consumption. Most of the time, priority is given to compression efficiency, resulting in low channel switching performance. In this paper, a Scalable Video Coding (SVC) configuration is proposed, wherein base and enhancement layer are coded in such a way that fast channel switching can be offered without impacting the access network bandwidth consumption. This is achieved by encoding a full quality slow channel switching version of the video stream in the base layer extended with a reduced quality fast switching enhancement layer. Additionally, this configuration offers backward compatibility with H.264/AVC, enabling a gradual upgrade of the IPTV system from H.264/AVC to SVC. When comparing this technique to single layer H.264/AVC, 25.0% bandwidth reduction on the access network is obtained without impacting channel switching speed, but only at the cost of a slightly reduced quality ( -1 dB PSNR) during a short transition period (maximum 48 frames). If a comparison is made with a regular SVC configuration, 39.4% bandwidth reduction on the access network during steady state is obtained. Compared to simulcast H.264/AVC, 32.0% bandwidth reduction on the core network is observed.
IEEE GEM 24 - Improved Deepfake Video Detection Using Convolutional Vision Transformer
Our paper titled "Improved Deepfake Video Detection Using Convolutional Vision Transformer" was accepted at the IEEE CTSoc Conference on Gaming, Entertainment, and Media (GEM) 2024.