Real-time compositing framework for interactive stereo fMRI displays

This research concentrates on providing high fidelity animation, only achievable with offline rendering solutions, for interactive fMRI-based experiments. Virtual characters are well established within the film, game and research worlds, yet much remains to be learned about which design, stylistic or behavioural factors combine to make a believable character. The definition of believability depends on context. When designing and implementing characters for entertainment, the concern is making believable characters that the audience will engage with. When using virtual characters in experiments, the aim is to create characters and synthetic spaces that people respond to in a similar manner to their real world counterparts. Research has shown that users show empathy for virtual characters. However, uncanny valley effects -- ie dips in user impressions -- can arise: behavioural fidelity expectations increase alongside increases in visual fidelity and vice versa. Often, characters used within virtual environments tend to be of fairly low fidelity due to technological constraints including rendering in real-time (Garau et al. 2003). This problem is addressed here by using non-linear playback and compositing of pre-rendered high fidelity sequences.