Visual Computing Otago

Visual computing covers a range of disciplines within computer science such as computer graphics, image processing, visualisation, computer vision, and virtual or augmented reality. Our research develops and uses techniques across all of these and more, but most of our work falls into three main themes: Using visual input to make models of the world; Understanding images and the models we build from them using tools from machine learning; Visualising data in the context of the world around us; and Allowing people to interact with data and information in new ways.

Modelling

Images provide a rich source of information about the world, but each image only provides a partial view. By understanding the process of image formation, and combining information from multiple images, we can build up more detailed models of the world. This includes recovering 3D structure, analysing motion and changes over time, and modelling variations in the shape and appearance of objects. More...

Understanding

Geometric models of the world are very useful, but they often lack the semantic information required for specific applications. Techniques from machine learning allow us to augment images or models of the world to provide the ability to understand what we are seeing. More...

Visualisation

Visualisation can be described as the process of converting abstract data into a visual representation that is comprehensible by a human observer, or in other words: making something visible. In our research, we look in particular into how to visualise information in Augmented Reality but also more general in 3D. More...

Interaction

Situated Visualisation of data from geographic information systems (GIS) is exposed to a set of problems, such as limited visibility, legibility, information clutter and the limited understanding of spatial relationships. In this work, we address the challenges of visibility, information clutter and understanding of spatial relationships with a set of dynamic Situated Visualization techniques that address the special needs of Situated Visualization of GIS data in particular for street-view-like perspectives as used for many navigation applications. The proposed techniques use GIS data as input for providing dynamic annotation placement, dynamic label alignment and occlusion culling. More...