Tech Papers 2021: AI-based saliency-aware video coding uses deep learning to identify visually important areas, optimizing video compression by enhancing quality in salient regions while lowering bitrate elsewhere
Tech Papers 2021: This paper discusses advanced quantization techniques in modern video codecs, showing how algorithms like Spatio-Temporal Adaptive Quantization (STAQ) and Local QP Refinement (LQR) significantly improve compression efficiency and video quality
Tech Papers 2021: This paper discusses how MPAI is developing AI-based data coding standards, to enable interoperable, efficient, and innovative AI applications
Tech Papers 2021: This paper discusses the promises and limitations of AI in real-time video encoding, showing that while it can enhance encoder performance, its complexity and cost often outweigh the benefits unless simplified and optimised approaches are used
Tech Papers 2021: Applying AI to video compression offers advantages like improved quality-of-experience (QoE), lower bitrate and processing needs, and increased density. This is achieved through applications such as Dynamic Encoding Style (DES), Dynamic Resolution Encoding (DRE), and Dynamic Frame Rate Encoding (DFE).
Tech Papers 2021: An open-solution system uses multiple Kinect v2 sensors connected to a single computer to perform real-time 3D reconstruction of dynamic scenes.
Tech Papers 2021: The report details new tools for broadcasters to capture and produce immersive 6-DoF audiovisual experiences by using light field video and spatial audio, which are then composited in a game engine.
Tech Papers 2021: This paper presents a lightweight CNN for subtle facial expression analysis in images and videos, enabling applications like expression-based search and actor summaries.
Tech Papers 2021: This paper explores correcting quantization artifacts in SDR and HDR content, evaluating debanding methods and CNN-based approaches to reduce visible false contours during HDR conversion.
Tech Paper 2021: This paper explores using a Colour Appearance Model (CAM) with Chromatic Adaptation Transform (CAT) for more accurate HLG-to-SDR down-mapping, producing natural results compared to traditional tone-mapping methods.
