|dc.identifier.citation||Niall Redmond, 'Influencing user perception using real-time adaptive abstraction', [thesis], Trinity College (Dublin, Ireland). School of Computer Science & Statistics, 2010, pp 143||
|dc.description.abstract||Real-time applications such as games, medical or technical visualisations and urban simulations
can often be highly complex in nature. This can lead to too much visual data
being presented to the user at once, which can make scenes cluttered and difficult to interpret.
Complex scenes can be particularly problematic when it is required to highlight
certain scene data to the user, as is often necessary in visualisations and interactive scenes.
This thesis describes research towards developing a solution for this problem by creating
perceptually optimised approaches using non-photorealistic rendering (NPR).
NPR is a research area within computer graphics which is driven by the desire for
both aesthetic stylisation and effective communication. The majority of existing nonphotorealistic
research focuses on creating stylistic and convincingly artistic renderings.
Other NPR techniques borrow from the artistic world to use stylistic rendering to draw a
user’s attention to certain parts of an image while reducing the impact of less important
areas. This can be done effectively by using multiple levels of abstraction within an image.
We use such techniques in real-time scenes to create renderings which successfully influence
user behaviour in interactive scenes.
In this thesis, we explore effective real-time multi-level abstraction methods for interactive
scenes. We examine a variety of stylistic approaches, to find which ones are suitable
for real-time scenes and which can be altered to create different levels of abstraction. We
propose an adaptive abstraction approach, which can be used across scenes to emphasise
particular objects and influence user perception. A number of non-photorealistic abstraction
techniques exist which can be defined as image-space or object-space approaches. We
implement the most suitable styles for the proposed adaptive abstraction approach, while
using the strengths of both image-space and object-space techniques to retain important
perceptual scene cues.
We investigate how a variety of non-photorealistic styles can affect user perception of
real-time scenes through a number of user experiments. We show how adaptive abstraction
can be used as an effective tool in facilitating user guidance and understanding in
scenes by examining a number of aspects of user perception such as eye-gaze behaviour
and shape perception. We also investigate how adaptive abstraction can affect task performance
in interactive scenes. We present a number of guidelines, learned from perceptual
experiments, on how adaptive abstraction can be best used in varying contexts.
We produce adaptive non-photorealistic styles, which could be smoothly integrated into
a traditionally modelled environment. This allows for the adaptive abstraction approach
to be easily incorporated into any existing application where particular scene data needs
to be highlighted to users. We show how the adaptive abstraction approach can be useful
for a variety of applications, including volume data. Volume visualisation is an example of
a type of application that suffers from an excessive amount of data being simultaneously
presented to a user. This problem can make the clear presentation of volume data a
difficult task. The adaptive abstraction approach suits this type of visualisation as we aim
to simplify complex scenes to focus on particular parts of the scene to make clear, stylised
|dc.publisher||Trinity College (Dublin, Ireland). School of Computer Science & Statistics||
|dc.subject||Computer Studies, Ph.D.||
|dc.subject||Ph.D. Trinity College Dublin||
|dc.title||Influencing user perception using real-time adaptive abstraction||
|dc.type.qualificationname||Doctor of Philosophy (Ph.D.)||
|dc.description.note||TARA (Trinity's Access to Research Archive) has a robust takedown policy. Please contact us if you have any concerns: email@example.com||