Video based reconstruction system for mixed reality environments supporting contextualised non-verbal communication and its study

This Thesis presents a system to capture, reconstruct and render the three-dimensional form of people and objects of interest in such detail that the spatial and visual aspects of non-verbal behaviour can be communicated.The system supports live distribution and simultaneous rendering in multiple locations enabling the apparent teleportation of people and objects. Additionally, the system allows for the recording of live sessions and their playback in natural time with free-viewpoint.It utilises components of a video based reconstruction and a distributed video implementation to create an end-to-end system that can operate in real-time and on commodity hardware.The research addresses the specific challenges of spatial and colour calibration, segmentation and overall system architecture to overcome technical barriers, the requirement of domain specific knowledge to setup and generate avatars to a consistent high quality.Applications of the system include, but are not limited to, telepresence, where the computer generated avatars used in Immersive Collaborative Virtual Environments can be replaced with ones that are faithful of the people they represent and supporting researchers in their study of human communication such as gaze, inter-personal distance and facial expression.The system has been adopted in other research projects and is integrated with a mixed reality application where, during a live linkup, a three-dimensional avatar is streamed to multiple end-points across different countries.

Level dc:type.qualificationlevel

Doctoral (Level 8)

Year dc:date.issued

2026

Author dc:creator

• Fairchild, AJ

Language dc:language

en

Identifier

oai:salford-repository.worktribe.com:1402217

OAI identifier oai:identifier

oai:salford-repository.worktribe.com:1402217