AVISION Audio and visual attention models applied to 2D and 3D audio-visual content

摘要

Within this paper a highly flexible approach for audio and visual attention modeling is presented. The developed system aims to be widely adaptable for different application scenarios within multimedia processing and coding. Possible use cases are presented and their influence on the system concept is shown. Furthermore the development of an attention model within the EU-funded research project DIOMEDES is described. This project focuses on developing a system for hybrid delivery of 3D stereoscopic and multi-view content to the homes through multiple transmission paths. The attention model, which is based on the framework presented, is used to enhance the content encoding process. This publication gives an overview over system design aspects as well as algorithms used for attention modeling.