Core research areas
With a strong interdisciplinary focus, the Centre for Visual Computing's research activity brings together three areas of expertise – computational, physiological and psychological – in order to make advances in fields as diverse as visual media, biometrics, security and computer gaming.
At the core of our work is visualisation: enabling computers to efficiently perceive, process and understand visual data such as images, videos and complex 3D scenes.
We focus on developing techniques and algorithms that generate and represent complex visual data in the most efficient way – particularly geometric data in 3D and spatio-temporal data in 4D.
For example, we use Partial Differential Equations (PDEs) to model and represent complex data, especially in situations where real-world data can be represented by sets of mathematical functions. Unlike standard geometric data representation methods such as polygonal meshes, the use of PDEs enables data to be visualised at any resolution without the need to store several levels of detail. (See this case study for an example).
This foundational work leads in to further specialised areas such as computer-based simulation, virtual reality and digital imaging.
Machine learning and data mining
Our work in visual computing is closely linked to artificial intelligence and its applications.
This research focus includes pattern recognition, knowledge extraction and data mining, time-series prediction and large-scale analysis of multi-dimensional data. In particular, our expertise and project areas are as follows:
- Large-scale analysis of multi-dimensional data for the purposes of verification, classification and clustering
- Development of automated real-time machine learning systems
- Efficient feature extraction from digital images and feature identification using machine learning algorithms (neural networks, support vector machines, radial basis functions, adaboosting, etc)
- Optimisation and automation of learning algorithms
- Development of time-series prediction systems
- Soft computing and application
Applied digital imaging
How do you measure and analyse images of the Sun's surface to accurately predict space weather? Is it possible to create a accurate 3D face recognition at low cost for security applications? Can optical character recognition be used in the Arabic alphabet?
These are typical of the problems we solve through our research in digital imaging and metrology. Our projects include:
- Automated prediction of solar flares by analysing satellite images in real time
- Automated analysis of digital mammograms and the detection of microcalcifications
- 3D visualisation of segmented regions of interest in 2D images
- Optical character recognition for Arabic writing
It's an area with wide ranging applications: space weather forecasting, security and biometrics, medical imaging, watermarking, visualisation and super resolution.
Computer based simulation
Until recently, product design involved physical experiments and prototypes, which are time-consuming and expensive. We’re driving progress towards design, simulation and product evaluation increasingly taking place in computer based virtual environments.
Techniques such as these allowed the Boeing 777 airliner to be designed, test flown, and altered as required before a single component was manufactured.
Our research in computer-based simulation focuses on two areas: Computer-Aided Design (CAD) and Computer-Aided Analysis (CAA).
Computer aided design
We currently focus our CAD research on the use of Partial Differential Equations (PDEs). PDEs can be used to describe the geometry of complicated objects so that the designer can create and manipulate them in an interactive environment. These techniques are used in developing new CAD software, or integrated into existing CAD packages.
Here’s one application of PDE techniques within CAD: the physical simulation of complex deformable systems in order to generate motion in computer-generated animation. Using PDEs lets us animate realistic objects by performing a series of curve transformations – the curves correspond to the character lines or the object boundaries. Our next challenge involves parameterising shapes to perform realistic animation models (e.g. animation of human figures and cartoon characters) in computer games and video.
Computer aided analysis
One of the major tasks in simulation-based design is to compute the functional properties of the object: for example strength, heat transfer, hydrodynamic or aerodynamic characteristics.
The proper linking of complicated surface geometry to analysis has been identified as a bottleneck in the implementation of simulation-based design, so our present work involves developing effective methods for integrating geometry with analysis for automated optimal design.
Virtual Reality may be widely misunderstood as belonging to fantasy gaming, but our work focuses on dozens of real-world applications.
How about taking a walk through an architect’s model to learn what it feels like to be in the space before it is constructed? Or creating a virtual theatre stage, allowing actors to rehearse separately then come together for the final dress rehearsal?
We have developed expertise in Virtual Reality and Mixed Reality techniques:
Virtual reality is a synthesised 3D environment that lets the user interact with objects directly via visual icons and menus. Normally they provide the user with some degree of immersion, using devices such as a wrap-around screen, a head-mounted display or a completely enclosed environment such as a Cave or a flight simulator.
The objective is to bring the user into a closer relationship with the data represented in visual form so that the user can evaluate changes through interaction, and also do simulations to investigate the properties of models or environments.
Evaluating architects’ designs with a virtual walk-through is just one application. For example, we can also create training environments that use virtual reality to test extreme situations without risking participant safety.
Mixed reality environments combine elements of the real and virtual worlds, where total immersion or complete synthesis do not apply to the same extent.
These techniques have a huge range of applications in science, technology and medicine, and have also been used in novel ways for arts and creative and cultural expression.
They are increasingly being used for global collaborations where all the participants can be brought together for a virtual meeting, and also for distributed learning.
They have also been used successfully to treat psychological conditions such as phobias.